Skip navigation.

A Wider View: Experimentations and ruminations on Oracle performance analysis

Syndicate content
Oracle Database performance tuning and analysis experimentation and ruminations.Craig Shallahamer, President/Founder, OraPubhttp://www.blogger.com/profile/04109635337570098781noreply@blogger.comBlogger95125
Updated: 17 min 38 sec ago

What Is That Light-Green Oracle Database CPU Wait Time?

Tue, 2015-02-10 20:10
This page has been permenately moved. Please CLICK HERE to be redirected.

Thanks, Craig.
What Really Is That Light-Green Oracle Database CPU Wait Time?

Have you ever wondered what that light-green "cpu wait time" really means in Oracle Enterprise Manager? It's what I call, the "gap" time. The "gap" time is the "missing" or the "leftover" time when DB Time does not equal the DB CPU (foreground process CPU consumption) plus the non-idle wait time. And, it happens more often than you might think.
If you have ever noticed that the database time seems too large, then you need to read this article. And, if you really want to know what the light-green "cpu wait time" in your OEM charts is, then you need to read this article. It's that good.

If you're serious about Oracle performance tuning and analysis, you'll want to know I just posted my complete 2015 public training schedule. It's on the main OraPub.com page HERE. Remember, alumni receive a 50% discount...crazy I know.
My Experiment Shows...
My experiment shows a strong relationship between the "gap" time and operating system CPU utilization. This means that a significant portion of the "gap" time is Oracle foreground processes sitting in the CPU run queue ready to consume CPU. This CPU run queue time is not part of DB CPU but it part of DB Time. So, when the CPU run queue time increases, so does DB Time and so does the "gap" time. And I have the data to show it! And you can run the same experiment yourself.
Let me put this another way. Most of the DB Time "gap" is Oracle foreground processes waiting in the operating system CPU run queue so they can eventually and truly consume CPU.

This is really important: When an Oracle foreground process is not consuming CPU but is sitting in the CPU run queue, Oracle Active Session History (ASH) facility records the session sample state as "CPU" and if the Oracle process is a foreground process (not a background process) Oracle's time model records this time as DB Time but not DB CPU. So in both the ASH and time model cases, someone must do some math to calculate this "cpu wait time".

But that name... "cpu wait"!
CPU Wait Time Is A Lousy Name
"CPU wait time" is a lousy name. Why? Mainly because it has caused lots of confusion and speculation. The name would be more appropriately called something like, "cpu queue time." Three reasons come to mind.
First, wait time means something special to Oracle DBAs. To an Oracle DBA anything associate with a "wait" should have a wait event name, a wait occurance, the time should be instrumented (i.e., measured) and should be recorded in the many wait interface related views, such as v$system_event or v$session.
Second, from an Oracle perspective the process is truly "on cpu" because the process is not "waiting." Remember, an Oracle session is either in one of two states; CPU or WAIT. There is no third choice. So the words "CPU Wait" are really confusing.
Third, from an OS perspective or simply a non-Oracle perspective, the Oracle process is sitting in the CPU run queue.
I'm sure in some Oracle Corporation meeting the words "cpu wait" were considered a great idea, but it has caused lots of confusion. And I'm sure it's here to stay.
What Does This "CPU WAIT" Look Like In OEM?
In OEM, the "cpu wait" is a light green color. I grabbed a publically available screenshot off the internet and posted it below. Look familiar? 

OK, so it's really easy to spot in OEM. And if you've seen it before you know EXACTLY what I'm referring to.
What Is CPU Wait Time?
First, let's review what we do know.
1. DB CPU is true Oracle foreground process CPU consumption as reported by the OS through a system call, such as getrusage.
2. CPU Wait time is derived, that is, somebody at Oracle wrote code to calculate the "cpu wait" time.
3. CPU Wait time is a lousy name because it causes lots of confusion.
4. CPU Wait time is shown in OEM as a light green color. DB CPU is shown as a dark/normal green color.
Second, I need to define what I'll call the DB Time "gap." This is not error and I am not implying something is wrong with database time, that it's not useful or anything like that. All I am saying is that sometimes DB Time does not equal DB CPU plus the non-idle wait time. Let's put that in a formula:
DB Time = DB CPU + non Idle Wait Time + gap
Really, What Is CPU Wait Time?
Now I'm ready to answer the question, "What is CPU WAIT time?" Here is the answer stated multiple ways.
"CPU Wait" time is Oracle foreground process OS CPU run queue time.
I ran an experiment (detailed below) and as the OS CPU utilization increased, so did the DB Time "gap" implying that the gap is CPU run queue time or at least a significant part of it.
I ran an experiment and there was a strong correlation between OS CPU utilization and the DB Time "gap" implying that the gap is CPU run queue time.
I ran an experiment and using queuing theory I was able to predict the "gap" time implying that the gap is CPU run queue time. (Whoops... sorry. That's what I'll present in my next post!)
So I'm very comfortable stating that when DB Time is greater than Oracle process CPU consumption plus the non-idle wait time, it's probably the result of Oracle foreground process CPU run queue time.

Yes, there could be some math problems on Oracle's side, there could be uninstrumented time (for sure it's happened before), the operating system could be reporting bogus values or a host of other potential issues. But unless there is an obvious wrong value, I'm sticking with the experimental evidence.
Now I'm going to show the experimental "evidence" that is, that the DB Time "gap" time correlates with the OS CPU utilization.
Let The Data Drive Our Understanding
You can download all the data collection scripts, raw experimental data, Mathematica notepad files, graphic files, etc HERE in a single zip file.

You should be able to run the experiment on any Linux Oracle test system. All you need is a logical IO load and for that I used my free opload tool which, you can download HERE.
The experiment placed an increasing logical IO load on an Linux Oracle 12c system until the operating system CPU utilization exceeded 90%. The load was increased 18 times. During each of the 18 loads, I gathered 31 three minute samples. Each sample contains busy time (v$osstat), idle time (v$osstat), logical IO (v$sysstat "session logical reads"), non-idle wait time (v$system_event where wait_class != 'Idle'), DB CPU (v$sys_time_model), background cpu time (v$sys_time_model), database time (v$sys_time_model DB time) and the sample time (dual table current_timestamp).
The CPU utilization was calculated using the "busy idle" method that I blog about HERE. This method is detailed in my Utilization On Steroids online video seminar.
The workload is defined as the logical IOs per second, lio/s.
Below is a table summarizing the experimental data. The times shown are the averages. If you look at the actual raw experimental data contained in the analysis pack, you'll notice the data is very consistent. This is not suprising since the load I placed should produce a very consistent workload.
Do you see the gaps? Look closely at load 18. The DB Time is 8891.4 seconds. But the sum of DB CPU (996.8 seconds) and the non-idle wait time (2719.2) seconds only equals 3716.0. Yet DB Time is 8891.4. So the "gap" is 5175.3 which is DB Time (8891.3) minus DB CPU (996.8) minus the non-idle wait time (2719.2).

Note: Load 11 and 12 where excluded because of a problem with my data collection. Sorry.

While we can numberically see the DB Time "gap" increase as the CPU utilization increases, check out the graphic in the next section!

The Correlation Between CPU Utilization And DB Time Gap
We can numerically and visually see that as the CPU utilization increases, so does the DB Time "gap." But is there a strong mathematical correlation? To determine this, I used all the experimental samples (except load 11 and 12). Because there was 17 different workloads and with each workload I gathered 31 samples, the correlation comprises of something like 527 samples. Pretty good sample set I'd say.

The correlation coefficient is a strong 0.891. The strongest is 1.0 and the weakest is 0.

Graphically, here is the scatterplot showing the relationship between the CPU utilization and the workload.

Don't expect the DB Time "gap" and OS CPU utilization correlation to be perfect. Remember that DB Time does not include Oracle background process CPU consumption, yet it is obviously part of the OS CPU utilization.

Summary
My experiment indicated the light-green "CPU wait time" is primarily Oracle foreground process operating system CPU run queue time. This is DB Time "gap" time.

My experiment also showed the "gap" time is highly correlated with CPU utilization. Which means, as the CPU utilization increases, so does the "gap" time.

If there are Oracle Database instrumentation bugs or a host of other potential problems, that will also affect the "gap" time.

If you want a more complete and detailed DB Time formula is would be this:

DB Time = DB CPU + Non Idle Wait Time + gap time

In my next post, I'll show you how to calculate the gap time based on queuing theory!

Thanks for reading!

Craig.








Categories: DBA Blogs

How To Approach Different Oracle Database Performance Problems

Tue, 2015-02-03 19:32
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.How To Approach Different Oracle Database Performance ProblemsJump Start Your Oracle Database Tuning Effort
Every Oracle Database Administrator will tell you no two performance problems are the same. But a seasoned Oracle DBA recognizes there are similarities...patterns. Fast problem pattern recognition allows us to minimize diagnosis time, so we can focus on developing amazing solutions.

I tend to group Oracle performance problems into four patterns. Quickly exploring these four patterns is what this article is all about.


You Can Not Possibly List Every Problem And Solution
When I teach, some Oracle Database Administrators want me to outline every conceivable problem along with the solution. Not only is the thought of this exhausting, it's not possible. Even my Stori product uses pattern matching. One of the keys to becoming a fantastic performance analyst is the ability quickly look at a problem and then decided which diagnosis approach is the best. For example, if you don't know the problem SQL (assuming there is one) tracing is not likely to be your best approach.

The Four Oracle Database Performance Patterns
Here are the four performance patterns I tend to group problems into.

The SQL Is Known
Many times there is a well know SQL statement that is responsible for the poor performance. While I will always do a quick Oracle Time Based Analysis (see below) and verify the accused SQL, I will directly attack this problem by tuning with SQL specific diagnostic and tuning tools.

But... I will also ask a senior application user, if the users are using the application correctly. Sometimes new applications users try and use a new application like their old application. It's like trying to drive a car with moving your feet as you are riding a bicycle... not going to work and it's dangerous!

Business Process Specific
I find that when the business is seriously affected by application performance issues, that's when the "limited budget" is suddenly not so limited. When managers and their business's are affected they want action.

When I'm approached to help solve a problem, I always ask how the business is being affected. If I keep hearing about a specific business process or application module I know two things.

First, there are many SQL statements involved. Second, the problem is bounded by a business process or application. This is when I start the diagnostic process with an Oracle Time Based Analysis approach which, will result in multiple solutions to the same problem.

As I teach in my online seminar How To Tune Oracle With An AWR Report, user feel performance through time. So, if our analysis is time based we can create a quantitative link between our analysis and their experience. If our analysis creates solutions that reduce time, then we can expect the user experience to improve. This combined with my "3 Circle" approach yields spot-on solutions very quickly.

While an Oracle Time Based Analysis is amazing, because Oracle does not instrument CPU consumption we can't answer the question, "What's Oracle doing with all that CPU?" If you want to drill into this topic check out my online seminar, Detailing Oracle CPU Consumption: The Missing Link.

It's Just Slow
How many times have I experienced this... It's Just Slow!


If what the user is attempting to explain is true, the performance issue is affecting a wide range of business processes. The problem is probably not a single issue (but could be) and clearly the key SQL is not know. Again, this is a perfect problem scenario to apply an Oracle Time Based Analysis.

The reason I say this is because an OTBA will look at the problem from multiple perspectives, categorize Oracle time and develop solutions to reduce those big categories of time. If you also do Unit Of Work Time Based Analysis, you can an even anticipate the impact of your solutions! Do an OraPub website search HERE or search my blog for UOWTBA.
Random Incident That Quickly Appears And Vanishes
This is the most difficult problem to fix. Mainly because the problem "randomly" appears and can't be duplicated. (Don't even bother calling Oracle Support to help in this situation.) Furthermore, it's too quick for an AWR report to show it's activity and you don't want to impact the production system by gathering tons of detailed performance statistics.

Even a solid Oracle Time Based Analysis will likely not help in this situation. Again, the problem is performance data collection and retention. The instrumented AWR or Statpack data does not provide enough detail. What we need step-by-step activity...like a timeline.

Because this type of problem scares both DBAs and business managers, you will likely need to answer questions like this:

  • What is that blip all about?
  • Did this impact users?
  • Has it happened before?
  • Will it happen again?
  • What should we do about it?

The only way I know how to truly diagnose a problem like this is to do a session-level time-line analysis. Thankfully, this is possible using the Oracle Active Session History data. Both v$active_session_history and dba_hist_active_sess_history are absolutely key in solving problems like this.

ASH samples Oracle Database session activity once each second (by default). This is very different than measuring how long something takes, which is the data an AWR report is based upon. Because sampling is non-continuous, a lot of detail can be collected, stored and analyzed.

A time-line type of analysis is so important, I enhanced my ASH tools in my OraPub System Monitor (OSM) toolkit to provide this type of analysis. If you want to check them out, download the OSM toolkit HERE, install it and read the osm/interactive/ash-readme.txt file.

As an example, using these tools you can construct an incident time-line like this:

HH:MM:SS.FFF User/Process Notes
------------ ------------- -----------------
15:18:28.796 suspect (837) started the massive update (see SQL below)

15:28:00.389 user (57) application hung (row lock on TM_SHEET_LINE_EXPLOR)
15:28:30.486 user (74) application hung (row lock on TM_SHEET_LINE_EXPLOR)
15:29:30.??? - row locks becomes the top wait event (16 locked users)
15:29:50.749 user (83) application hung (row lock on TM_SHEET_LINE_EXPLOR)

15:30:20.871 user (837) suspect broke out of update (implied)
15:30:20.871 user (57) application returned
15:30:20.871 user (74) application returned
15:30:20.871 user (83) application returned

15:30:30.905 smon (721) first smon action since before 15:25:00 (os thread startup)
15:30:50.974 user (837) first wait for undo - suspect broke out of update
15:30:50.974 - 225 active session, now top event (wait for a undo record)

15:33:41.636 smon (721) last PQ event (PX Deq: Test for msg)
15:33:41.636 user (837) application returned to suspect. Undo completed
15:33:51.670 smon (721) last related event (DFS lock handle)

Without ASH seemingly random problems would be a virtually impossible nightmare scenario for an Oracle DBA.
Summary
It's true. You need the right tool for the job. And the same is true when diagnosing Oracle Database performance. What I've done above is group probably 90% of the problems we face as Oracle DBAs into four categories. And each of these categories needs a special kind of tool and/or diagnosis method.

Once we recognize the problem pattern and get the best tool/method involved to diagnosis the problem, then we will know the time spent developing amazing solutions is time well spent.

Enjoy your work!

Craig.


Categories: DBA Blogs

I Have Lots Of Oracle Database Server Power But Performance Is Slow/Bad

Mon, 2015-01-19 19:14
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.I Have Lots Of Oracle Database Server Power But Performance Is Slow/Bad
Oracle Database parallelism and serialization is what we as Oracle Database Administrators live and die for. You have a screaming fast Oracle Database system and there is lots of computing power available.

But performance is unacceptable; users are screaming, the phone is ringing, and those fancy dashboards are flashing like it's Christmastime.

What is going on?! What can I do about it?! That's what this post is about.
Learn By Doing
Back in December on the third day of my Oracle Performance Firefighting class, I had each student bring in an AWR report from one of their systems that was giving them problems. (I keep my classes small, giving each student time to do their analysis and time for us to talk about it as a class.)
Get this:Half of the systems had a similar "problem." I think it's important every DBA understands this "problem" because it's more common than most people believe.
There was plenty of computing power and the key SQL statement they cared about was a batch job. What was the core problem? The quick answer is "serialization" that is, a lack of parallelism. Exploring this using a very large production system AWR report and coming up with solutions is what this posting is all about.
Serialization Is Death
In Oracle systems, serialization is death and parallelism is life. Follow this line: business, end user, application designer, DBA, Oracle Database Kernel Architects (or whatever their title is), OS Administrators, OS designers, CPU designers and IO subsystem designers all have something in common. They work hard to parallelize tasks. Just one example: Oracle is designed to have multiple background and foreground processes running parallel.
But all this parallelization effort can be wasted and minimized if a process turns into a serial work stream (at any level; Oracle, OS, business, etc.). The result is "slowness" because the wall time increases.
Available Power And Slowness Equals Opportunity
When I tune Oracle Database systems, I look for opportunities. And each of my solutions will specifically target an opportunity. When I see unused power and complaints of slowness, I look for ways to increase parallelism. Why? Because having available power combined with slowness likely means a serialization limitation exists.

For sure serialization may be necessary. Two examples come to mind; Oracle database memory serialization control (think: latch and mutex) and business rules.

But if I can find a way to increase performance by using up available power by increasing parallelism, I'll likely be able to turn a slow serialization situation into a screaming fast parallelization situation!
How To Recognize A Serially Constrained System
It's easy to recognize a serially constrained system. Ask yourself these two questions. First, is there available CPU or IO power? Second, are there complaints of application "slowness." If the answer to both of these questions is "Yes" then there is likely a serialization issue. Furthermore, the general solution is to use the available resources to our advantage. That is, find areas to increase parallelization, which will use the available resources and improve performance.
If you have the power, use it! What are you saving it for?
(There may be a very good answer to the "saving" but I'll save that for another article.)
Can I Be Out Of CPU And Be Serially Constrained?
Yes. An Oracle Database system can be serially constrained and be out of OS resources. A great example of this is when there is a raging Oracle memory serialization issue. If you see both significant Oracle latching or mutex wait time combined with a raging CPU bottleneck, you likely have a serialization issue... an Oracle Database memory structure access serialization issue.

So, while available power on a "slow" system likely means we have a serially constrained system there are situations in Oracle with a raging CPU bottleneck that also means there is likely a serialization issue.

Find Out: Is There Available CPU Power?
Here Is A Real Life Situation. To simplify, I'm going to focus on only instance number one. Look at instance number one in the below picture.

The above AWR report snippet shows RAC node #1 OS CPU utilization at 15%. This means that over the AWR report snapshot interval, the average CPU utilization was 15%. I never initially trust an AWR report for calculated results. Plus it's good practice to do the math yourself. If you use the super fast busy-idle method I have outlined in THIS POST and detailed in my online seminar, Utilization On Steroids, the utilization calculates to 16% ( 0.5/(0.5+2.7)=0.16 ). So the AWR Report's 15% for CPU "% Busy" looks to be correct.
Clearly with an average CPU utilization of 15%, we have an opportunity to use the unused CPU power to our advantage.
Find Out: Is There Available IO Power?
I am looking for fast IO responsiveness. That is, a low response time. A great way to get a quick view of IO subsystem responsiveness is to look at the average wait time for the event, db file sequential read.
The wait event, db file sequential read is the time it takes to read a single block synchronously. I like to call it a pure IO read call: a) what time is it? b) make the IO call and wait until you get it, c) what time is it? d) calculate the delta and you have the wait time...and the IO read call response time! If you want more details, I wrote about this HERE, which includes a short video.
For our system, let's figure out the single block IO subsystem read response time. Using the same AWR report, here is a screen shot of the Top Time Events.

Again, I'm just going to focus on the first instance. If you look closely (middle right area), you'll see for instance number one, the average db file sequential read time wait time is 2.22ms. That's fast!
There is no way a physical spinning disk is going to return a block in 2.22ms. This means that many of Oracle's single block read calls are be satisfied through some non-Oracle cache. Perhaps an OS cache or an IO subsystem cache. We can't tell, but we do know the block was NOT an Oracle's buffer cache because the db file sequential wait means the block was not found in Oracle's buffer cache.
A single block synchronous IO read call with an average of 2.22ms means there is available IO read capacity and probably available write capacity as well. Again, just like with the OS CPU subsystem, we have unused power that we will try and use to our advantage.
At this point, I will assume there is also plenty of memory and network capacity available. So, the bottom line is we have a "slow" system combined with available CPU and available IO power. Wow! That is a great situation to be in. I call this, "low hanging fruit."
Real Life: Looking For The "Slow" SQL
At the top of this post, I mentioned that in my Firefighting class in each of the "serialization" cases, there was a key SQL statement that was part of a larger batch process. Keep in mind, that at this point in the analysis I did NOT know this. All I knew was that users were complaining and there was plenty of CPU and IO resources.
Usually, in this situation there is a relatively long running process. There could be lots of quick SQL statement involved, but usually this is not the case. And I'm hoping there is a key long running SQL statement that can be parallelized.
Long running can roughly be translated into "high elapsed time." I've have written a number of articles about elapsed time (search my blog for: elapsed time) and even have a free tool with which, you can gather to get more than simply the average elapsed time. And I have online seminars that touch on this subject: Tuning Oracle Using An AWR Report and also, Using Skewed Performance Data To Your Advantage. So there are lots of useful resources on this topic.
In the AWR report, I'm going to look closely at the SQL Statistics, in particular the "SQL ordered by Elapsed Time (Global)." What I really want is the statistics only for instance one, that is, not global. But that's all I have available. Plus the DBAs will/should know if the key SQL statement(s) are run on instance one. Here's the report.


In the report above, look at the elapsed times (second column on the left). Now looking right, find the "Execs", that is, the executions column. The execution column is the number of completed executions within this AWR snapshot range. If the executions is zero, this means the SQL did not complete during the snapshot interval, that is before the ending snapshot.
If you're wondering, these top elapsed time SQL statements are involved in batch processing. When I look at this, I see opportunity, fruit waiting to be harvested!
And I love this: Every DBA in the class in this situation said, "Oh! I know about this SQL. It's always causing problems." Now it's time to do something about it!
Real Life: Putting This All Together
We have identified available CPU and IO capacity. And we have identified THE elapsed time SQL statement. While I'm a pretty laid back kind of guy, at this point I start to apply some pressure. Why? Because the users are complaining, we have identified both an opportunity, the cause of the problem and the general solutions.
There are two general solutions:
1. Do less work. You want to empty a candy dish faster? Then start with less candy in the dish! If you want a SQL statement to run faster, tune the SQL so it touches less blocks.
2. Do the same amount of work, but group the work and run each group at the same time. This is parallelization! This is why the total elapsed time will not decrease (it will probably increase a little) but the wall time will likely decrease... and dramatically! Here is a LINK to posting that contains a short video demonstrating the difference between elapsed time and wall time.
How To Parallelize (in summary)
There are many different ways to parallelize. But the goal is the same: use the available resources to reduce wall time (not necessarily the elapsed time). Perhaps the application can be redesigned to run in parallel streams. But that can take a very long time and be a real hassle. But in many cases, it's the best long term solution.
If you are short on time, are licensed for Oracle Parallel Query and the SQL has been optimized (oh boy... how many times have all heard that before), you likely can use Oracle PQ. And of course, even if the SQL is not optimized, you can still run PQ and performance may be fantastic.
By the way, adding faster IO disks or more IO disks (what is a "disk" is nowadays anyways) will likely NOT work. Remember the IO subsystem is performing wonderfully.
Thanks for reading and enjoy the mystery of your work!
Craig.
Categories: DBA Blogs

Do The LGWRs Always Sleep For The Full Three Seconds?

Mon, 2015-01-12 16:01
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.Do Oracle Database LGWRs (10g, 11g, 12c) Always Sleep For The Full Three Seconds?
Back in June I wrote (included a video) about the Oracle Database log writer(s) "3 second sleep rule." That's the rule we were all taught by our instructors when we started learning about Oracle yet never really knew if it was true. In that post, I demonstrated Oracle Database log writer background processes are normally put to sleep for three seconds.

In this post, I want to answer a related but different question.

Do Oracle Database log writer background processes ALWAYS sleep for the full three seconds? Our initial response would likely be, "Of course not! Because what if a foreground process commits during the three second sleep? The log writer(s) must wake up." That would make sense.

But, is this really true and what else could we learn by digging into this? I created an experiment to check this out, and that is what this post is all about.

The Experiment
In my June post I demonstrated the Three Second Rule. You will see this again below. But in this experiment we are looking for a situation when one of the 12c log writers wakes BEFORE their three second sleep.

You can download the experimental script I detail below HERE.

This is really tricky to demonstrate because of all the processes involved. There is a the Oracle foreground process and in 12c, there are multiple log writer background processes. Because this is experiment follows a timeline, I needed to gather the process activity data and then somehow merge it all together in a way that we humans can understand.

What I did was to do an operating system trace ( strace ) each process ( strace -p $lgwr )  with the timestamp option ( strace -p $lgwr -tt ) sending each process's the output to a separate file ( strace -p $lgwr -tt -o lgwr.txt ). This was done to all four processes and of course, I needed to start the scripts to run in the background. Shown directly below are the log writer strace details.

lgwr=`ps -eaf | grep $sid | grep lgwr | awk '{print $2}'`
lg00=`ps -eaf | grep $sid | grep lg00 | awk '{print $2}'`
lg01=`ps -eaf | grep $sid | grep lg01 | awk '{print $2}'`

echo "lgwr=$lgwr lg00=$lg00 lg01=$lg01"

strace -p $lgwr -tt -o lgwr.str &
strace -p $lg00 -tt -o lg00.str &
strace -p $lg01 -tt -o lg01.str &

Once the log writers were being traced, I connected to sqlplus and launched the below text in the background as well.

drop table bogus;
create table bogus as select * from dba_objects where object_id in (83395,176271,176279,176280);
select * from bogus;
commit;
exec dbms_lock.sleep(2.1);

exec dbms_lock.sleep(2.2);
exec dbms_lock.sleep(2.3);
update bogus set object_name='83395' where object_id=83395;
exec dbms_lock.sleep(3.1);
update bogus set object_name='176271' where object_id=176271;
exec dbms_lock.sleep(3.2);
update bogus set object_name='176279' where object_id=176279;
exec dbms_lock.sleep(3.3);
update bogus set object_name='176280' where object_id=176280;
exec dbms_lock.sleep(3.4);
commit;
exec dbms_lock.sleep(3.5);
update bogus set object_name='89567' where object_id=89567;
exec dbms_lock.sleep(3.6);
commit;
exec dbms_lock.sleep(3.7);
exit;

Once the sqlplus session was connected,

sqlplus system/manager @/tmp/runit.bogus &
sleep 2

I grabbed it's OS process id and started an OS trace on it as well:

svpr=`ps -eaf | grep -v grep | grep oracle$sid | awk '{print $2}' `
echo "svpr=$svpr"

strace -p $svpr -tt -o svpr.str &

Then I slept for 30 seconds, killed the tracing processes (not the log writers!):

sleep 30

for pid in `ps -eaf | grep -v grep | grep strace | awk '{print $2}'`
do
echo "killing pid $pid"
kill -2 $pid
done

Then I merged the trace files, sorted them by time, got rid of stuff in the trace files I didn't want to see and put the results into a final "clean" file.

merge=/tmp/strace.merge.bogus
rm -f $merge
for fn in lgwr lg00 lg01 svpr
do
cat ${fn}.str | awk -v FN=$fn '{print $1 " " FN " " $2 " " $3 " " $4 " " $5 " " $6 " " $7 " " $8 " " $9}' >> $merge
done

ls -ltr $merge
date
cat $merge | sort > /tmp/final.bogus

cat /tmp/final.bogus | grep -v times | grep -v getrusage | grep -v "svpr lseek" | grep -v clock_gettime | grep -v gettimeofday | grep -v "svpr read" | grep -v "svpr write" > /tmp/final.bogus.clean

The amazing thing is... this actually worked! Here is the output below:

19:11:41.981934 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {2, 200000000}) =
19:11:42.859905 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:11:43.986421 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:11:44.186404 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {2, 300000000}) =
19:11:44.982768 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:11:45.860871 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:11:46.499014 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 100000000}) =
19:11:46.989885 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:11:47.983782 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:11:48.861837 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:11:49.608154 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 200000000}) =
19:11:49.993520 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:11:50.984737 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:11:51.862921 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:11:52.817751 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 300000000}) =
19:11:52.997116 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:11:53.985784 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:11:54.863809 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:11:55.998974 lgwr open("/proc/41955/stat", O_RDONLY) = 19
19:11:55.999029 lgwr read(19, "41955 (ora_pmon_prod35) S 1 4195"..., 999) =
19:11:55.999075 lgwr close(19) = 0
19:11:55.999746 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:11:56.127326 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 400000000}) =
19:11:56.986935 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:11:57.864930 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:11:59.003212 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:11:59.531161 svpr semctl(7503875, 16, SETVAL, 0x7fff00000001) = 0
19:11:59.531544 lgwr semctl(7503875, 18, SETVAL, 0x7fff00000001) = 0
19:11:59.532204 lg00 pwrite(256, "\1\"\0\0\311\21\0\0\354\277\0\0\20\200{\356\220\6\0\0\r\0\0\0\367^K\5\1\0\0\0"..., 2048, 2331136) = 2048
19:11:59.532317 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {2, 480000000}) =
19:11:59.532680 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {0, 100000000}) =
19:11:59.537202 lg00 semctl(7503875, 34, SETVAL, 0x7fff00000001) = 0
19:11:59.537263 lg00 semctl(7503875, 16, SETVAL, 0x7fff00000001) = 0
19:11:59.537350 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:11:59.538483 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {2, 470000000}) =
19:11:59.540574 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 500000000}) =
19:12:00.865928 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:12:02.011876 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:12:02.537887 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:12:03.050381 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 600000000}) =
19:12:03.866796 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:12:05.014819 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:12:05.538797 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:12:06.657075 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 700000000}) =
19:12:06.867922 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:12:08.017814 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:12:08.539750 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:12:09.868825 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}

There is a lot of detail in the above output. I'm only going to make a few comments that pertain to the objectives of this post.

Oracle is using the semaphore call semtimedop to sleep. The beauty of this call, is it allow the process to be woken (that is, signaled) by another process! Keep that mind as you follow the timeline.

Here we go:

19:11:41.981934. Notice the server process' "2, 2" and later the "2,3" and "3, 1" and "3, 2"? This is the result of the dbms_lock.sleep commands contained in the sqlplus script!

19:11:42.859905. Notice lg01 and the other log writer background processes always have a "3, 0" semtimedop call? That is their "3 second sleep."

Look at the first few lgwr entries. I've listed them here:

19:11:43.986421
19:11:46.989885
19:11:49.993520
19:11:52.997116

Notice anything strange about the above times? They are all just about 3 seconds apart of from each other. That's the 3 second sleep in action. But that's not the focus of this post. So let's move on.

Read this slow: I want to focus on just one part of the output which, is shown below. Notice the server process is sleeping for 3.4 seconds. If you look at the sqlplus script (near the top of this post), immediately after the 3.4 second sleep the server process issues a commit. Therefore, because the 3.4 sleep starts at 19:11:56.1 and I'm expecting to see some log writer activity in 3.4 seconds. This would be at 19.11.59.5. This could occur in the middle of the log writer 3 second sleep, which means we will likely see a log writer kick into action before their 3 second sleep completes! Let's take a look.

19:11:56.127326 svpr semtimedop(7503875, {{34, -1, 0}}, 1, {3, 400000000}) =
19:11:56.986935 lg00 semtimedop(7503875, {{18, -1, 0}}, 1, {3, 0}) =
19:11:57.864930 lg01 semtimedop(7503875, {{19, -1, 0}}, 1, {3, 0}) =
19:11:59.003212 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {3, 0}) =
19:11:59.531161 svpr semctl(7503875, 16, SETVAL, 0x7fff00000001) = 0
19:11:59.531544 lgwr semctl(7503875, 18, SETVAL, 0x7fff00000001) = 0
19:11:59.532204 lg00 pwrite(256, "\1\"\0\0\311\21\0\0\354\277\0\0\20\200{\356\220\6\0\0\r\0\0\0\367^K\5\1\0\0\0"..., 2048, 2331136) = 2048
19:11:59.532317 lgwr semtimedop(7503875, {{16, -1, 0}}, 1, {2, 480000000})

We can see the server process 3.4 second sleep starting at time 19:11:56.1 and we can see the sleep end and the server process' next command begin at the expected time of 19:11:59.5. Next in the trace file output is result of the commit. The commit results in the wake of both the lgwr and lg00 background processes.

But notice the lgwr background process started one of its 3 second sleeps at 19:11:59.0 which means it doesn't want to wake until 19:12:02.0. But look at when the lgwr process woke up. It woke up at 19.11.59.5 which is clearly before the expected time of 19:12:02.0. What you just noticed was the lgwr background process was signaled to wake up before its three second sleep completed.

But why did the lgwr need to be woken up? Because the server process' redo must be immediately written.

But it gets even better because the lgwr background process doesn't do the redo write! The lgwr process signals the lg00 process to do the write, which we can see occurs at time 19:11:59:5. Wow. Amazing!

What We Can Learn From This
Personally, I love these kinds of postings because we can see Oracle in action and demonstrating what we believe to be true. So what does all this actually demonstrate? Here's a list:

  1. We can see the 12c log writers involved. Not only lgwr.
  2. All log writer background process initiate a sleep for the default three seconds. I have seen situations where it is not three seconds, but it appears the default is three seconds.
  3. The server process signals the lgwr process to write immediately after a commit is issued.
  4. The server process signals the lgwr process to write using a semaphore.
  5. The log writers (starting in 12c) can signal each other using semaphores. We saw lgwr signal the lg00 background process to write.
  6. The server process was performing updates over 10+ a second period, yet its redo was not written to disk until it committed. This demonstrates that ALL redo is not flushed every three seconds. (This is probably not what you learned... unless you joined one of my Oracle Performance Firefighting classes.)
  7. The log writers while normally put to sleep for three seconds, can be woken in the middle for an urgent task (like writing committed data to an online redo log).

I hope you enjoyed this post!

Thanks for reading,

Craig.
Categories: DBA Blogs

Can A Background Process Impact A Foreground Process And Its Database Time?

Mon, 2014-12-29 19:03
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.Can A Background Process Impact A Foreground Process And Its Database Time?
Have you ever heard someone say, "Background processes do not impact foreground processes because they run in the background and in parallel with foreground processes." I've heard this hundreds of times!

While doing some performance research I came across a great example of how an Oracle Database background process can directly and significantly impact a foreground process.

The above quote represents a masterfully constructed lie; it contains both a lie and a truth. The mix of a truth and a lie make understanding the reality of the situation difficult. In this post, I'll explain the truth, delve into the lie and relate it all to foreground process database time.

By the way, I am in no way saying there is something wrong with or incorrect about DB Time. I want to ensure this is clear from the very beginning of this post.

Just so there is no confusion, an Oracle foreground process is sometimes also called a server process or a shadow process. These can terms can be used interchangeably in this post.

The Truth
Clearly background and foreground processes operate in parallel. I don't think any DBA would deny this. As I frequently say, "serialization is death and parallelism is life!" A simple "ps" command will visually show both Oracle background and foreground processes at work. But this in no way implies they do not impact each other's activity and performance.

In fact, we hope they do impact each other! Can you imagine what performance would be with the background processes NOT running in parallel?! What a performance nightmare that would be. But this where the "no impact" lie lives.

The Lie
Most senior DBAs can point to a specific situation where Oracle cache buffer chain latch contention affected multiple foreground sessions. In this situation, foreground sessions were franticly trying to acquire a popular cache buffer chain latch. But this is a foreground session versus foreground session situation. While this is example is important, this post is about when a background process impacts a foreground process.

Have you every committed a transaction and it hangs while the foreground process is waiting on "log file switch (checkpoint incomplete)" or even worse "log file switch (archiving needed)" event? All the foreground process knows is that its statement can't finish because a required log switch has not occurred because a checkpoint is incomplete. What the server process does not know is the checkpoint (CKPT), the database writer (DBWR) and the log writer (LGWR) background processes are involved. There is a good chance the database writer is frantically writing dirty buffers to the database (dbf) files so the LGWR can safely overwrite the associated redo in the next online redo log.

For example, if a server process issued a commit during the checkpoint, it will wait until the checkpoint is complete and the log writer has switched and can write into the next redo log. So, while the log writer background processes is probably waiting on "log file parallel write" and the database writer is burning CPU and waiting on "db file parallel write", the foreground processes are effectively hung.

This is a classic example of how a background process can impact the performance of a foreground process.

A Demonstration Of The Lie
Here's a quick demonstration of the above situation. On an existing database in my lab, I created two 4MB redo logs and dropped all the other redo logs. I started a DML intensive workload. According to the alert.log file, the redo logs where switching every couple of seconds! Take a look at this:
$ tail -f /home/oracle/base/diag/rdbms/prod30/prod30/trace/alert*log
Thread 1 cannot allocate new log, sequence 2365
Checkpoint not complete
Current log# 4 seq# 2364 mem# 0: /home/oradata/prod30/redoA1.log
Mon Dec 29 11:02:09 2014
Thread 1 advanced to log sequence 2365 (LGWR switch)
Current log# 5 seq# 2365 mem# 0: /home/oradata/prod30/redoA2.log
Thread 1 cannot allocate new log, sequence 2366
Checkpoint not complete
Current log# 5 seq# 2365 mem# 0: /home/oradata/prod30/redoA2.log
Thread 1 advanced to log sequence 2366 (LGWR switch)
Current log# 4 seq# 2366 mem# 0: /home/oradata/prod30/redoA1.log
Thread 1 cannot allocate new log, sequence 2367
Checkpoint not complete
Current log# 4 seq# 2366 mem# 0: /home/oradata/prod30/redoA1.log
Thread 1 advanced to log sequence 2367 (LGWR switch)
Current log# 5 seq# 2367 mem# 0: /home/oradata/prod30/redoA2.log
Thread 1 cannot allocate new log, sequence 2368
Checkpoint not complete
Current log# 5 seq# 2367 mem# 0: /home/oradata/prod30/redoA2.log
Mon Dec 29 11:02:20 2014

Obviously not what you want to see on a production Oracle system! (But my guess many of you have.)

Using my OSM realtime session sampler tool (rss.sql - related blog posting HERE) I sampled the log writer every half a second. (There is only one log writer background process because this is an Oracle 11g database, not an Oracle Database 12c system.) If the log writer session showed up in v$session as an active session, it would be picked up by rss.sql.  Both "ON CPU" and "WAIT" states are collected. Here is a sample of the output.


It's very obvious the log writer is doing some writing. But we can't tell from the above output if the process is impacting other sessions. It would have also been very interesting to sample the database writer also, but I didn't do that. To determine if the background processes are impacting other sessions, I needed to find a foreground session that was doing some commits. I noticed that session 133, a foreground process was busy doing some DML and committing as it processed its work. Just as with the log writer background process, I sampled this foreground process once every 0.5 second. Here's a sample of the output.


Wow. The foreground process is waiting a lot for the current checkpoint to be completed! So... this means the foreground process is being effectively halted until the background processes involved with the checkpoint have finished their work.
This is a great example of how Oracle background processes can impact the performance of an Oracle foreground process.

But let's be clear. Without the background processes, performance would be even worse. Why? Because all that work done in parallel and in the background would have to be done by each foreground process AND all that work would have to be closely controlled and coordinated. And that, would be a performance nightmare!
DB Time Impact On The Foreground Process
Just for the fun of it, I wrote a script to investigate DB Time, CPU consumption, non-idle wait time and the wait time for the "log file switch wait (checkpoint incomplete)" wait event for the foreground process mentioned above (session 133). The script simply gathers some session details, sleeps for 120 seconds, again gathers some session details, calculates the differences and displays the results. You can download the script HERE. Below is the output for the foreground process, session 133.
SQL> @ckpttest.sql 133

Table dropped.

Table created.

PL/SQL procedure successfully completed.

CPU_S_DELTA NIW_S_DELTA DB_TIME_S_DELTA CHECK_IMPL_WAIT_S
----------- ----------- --------------- -----------------
2.362 117.71 119.973692 112.42

1 row selected.

Here is a quick description of the output columns.

  • CPU_S_DELTA is the CPU seconds consumed by session 133, which is the time model statistic DB CPU.
  • NIW_S_DELTA is the non-idle wait time for session 133, in seconds.
  • DB_TIME_S_DELTA is the DB Time statistic for session 133, which is the time model statistic DB Time.
  • CHECK_IMPL_WAIT_S is the wait time only for event "log file switch (checkpoint incomplete)" for session 133, in seconds.

Does the time fit together as we expect? The "log file switch..." wait time is part of the non-idle wait time. The DB Time total is very close to the CPU time plus the non-idle wait time. Everything seems to add up nicely.

To summarize: Oracle background processes directly impacted the database time for a foreground process.

In Conclusion...
First, for sure Oracle foreground and background processes impact each other...by design for increased performance. Sometimes on real production Oracle Database systems things get messy and work that we hoped would be done in parallel must become momentarily serialized. The log file switch example above, is an example of this.

Second, the next time someone tells you that an Oracle background process does not impact the performance of a foreground process, ask them if they have experienced a "log file switch checkpoint incomplete" situation. Pause until they say, "Yes." Then just look at them and don't say a word. After a few seconds you may see a "oh... I get it." look on their face. But if not, simply point them to this post.

Thanks for reading and enjoy your work!

Craig.




Categories: DBA Blogs

Is Oracle Database Time Correct? Something Is Not Quite Right.

Mon, 2014-12-22 19:17
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.Is Oracle Database Time Correct? Something Is Not Quite Right.
Oracle Database tuning and performance analysis is usually based on time. As I blogged HERE, the Oracle "database time" statistic is more interesting than simply "time spent in the database." It is the sum of CPU consumption and non-idle wait time. And Elapsed Time is the sum of all the database time related to perhaps a session or a SQL statement execution. However...
If you do the fundamental math, you'll notice the numbers don't always add up. In fact, they rarely match. In this posting, I want to demonstrate this mismatch and I want you to see this on your systems!

I'll include experimental data from a number of different systems, the statistical analysis (including pictures) and provide a tool you can download for free from OraPub.com to check out the situation on your systems.
Checking DB Time Math
DB Time is defined as "time spent in the database." This is the sum of Oracle process CPU consumption plus non-idle wait time. Usually we don't derive DB Time. The Time Model Statistics view v$sys_time_mode contains the DB Time statistic. But this view also contains the DB CPU statistic. Since there is no sum of non-idle wait time, most people just assume everything is fine.
However, if you run the simple query below on your system, it could look something this:
SQL> l
1 select db_time_s, db_cpu_s, tot_ni_wt_s
2 from (select value/1000000 db_time_s from v$sys_time_model where stat_name = 'DB time' ),
3 (select value/1000000 db_cpu_s from v$sys_time_model where stat_name = 'DB CPU' ),
4* (select sum(TIME_WAITED_MICRO_FG)/1000000 tot_ni_wt_s from v$system_event where wait_class != 'Idle' )
SQL> /

DB_TIME_S DB_CPU_S TOT_NI_WT_S
---------- ---------- -----------
330165.527 231403.925 119942.952

1 row selected.
If you add up the DB CPU and the total non-idle wait time, the value is 351,346.877. Woops! 351K does not equal 330K. What happened on my Oracle Database 12c (12.1.0.2.0)? As I have demonstrated in this POSTING (which contains videos of this) and in my online seminar training HERE, many times DB Time does nearly equal DB CPU plus the non-idle wait time. But clearly in the above situation something does not seem quite right.

Checking DB Time On Your Systems
To demonstrate the possibility of a DB Time mismatch, I created a simple plsql tool. You can download this free tool or do an OraPub.com search for "db time tool". The tool, which is easily configurable, takes a number of samples over a period of time and displays the output.


Here is an example of the output.

OraPub DB Time Test v1a 26-Sep-2014. Enjoy but use at your own risk.
.
Starting to collect 11 180 second samples now...
All displayed times are in seconds.
.
anonymer Block abgeschlossen
..........................................................................
... RAW OUTPUT (keep the output for your records and analysis)
..........................................................................
.
sample#, db_time_delta_v , db_cpu_delta_v, tot_ni_wait_delta_v, derived_db_time_delta_v, diff_v, diff_pct_v
.
1, 128,4, 128,254, ,103, 128,357266, ,043, 0
2, 22,014, 3,883, 17,731, 21,614215, ,399, 1,8
3, 1,625, 1,251, ,003, 1,253703, ,371, 22,8
4, 13,967, 12,719, 1,476, 14,194999, -,228, -1,6
5, 41,086, 41,259, ,228, 41,486482, -,4, -1
6, 36,872, 36,466, ,127, 36,593884, ,278, ,8
7, 38,545, 38,71, ,137, 38,847459, -,303, -,8
8, 37,264, 37,341, ,122, 37,463525, -,199, -,5
9, 22,818, 22,866, ,102, 22,967141, -,149, -,7
10, 30,985, 30,614, ,109, 30,723831, ,261, ,8
11, 5,795, 5,445, ,513, 5,958586, -,164, -2,8
.
The test is complete.
.
All displayed times are in seconds.

The output is formatted to make it easy to statistically analyze. The far right column is percent difference between the reported DB Time and the calculated DB Time. In the above example, they are pretty close. Get the tool and try it out on your systems.

Some Actual Examples
I want to quickly show you four examples from a variety of systems. You can download all the data in the "analysis pack" HERE. The data, for each of the four systems, contains the raw DB Time Test output (like in the section above), the statistical numeric analysis output from the statistical package "R", the actual "R" script and the visual analysis using "smooth histograms" also created using "R."

Below is the statistical numeric summary:


About the columns: Only the "craig" system is mine and other are real production or DEV/QA systems. The statistical columns all reference the far right column of the DB Time Test Tool's output, which is the percent difference between the reported DB Time and the calculated DB Time. Each sample set consists of eleven 180 second samples. The P-Value greater than 0.05 means the reported and calculated DB Time differences are normally distributed. This is not important in this analysis, but gives me clues if there is a problem with the data collection.

As you can easily see, two of the system's "DB Times" difference is greater than 10% and one of them was over 20%. The data collected shows that something is not quite right... but that's about it.

What Does This Mean In Our Work?
Clearly something is not quite right. There are a number of possible reasons and this will be focus of my next few articles.

However, I want to say that even though the numbers don't match perfectly and sometimes they are way off, this does not negate the value of a time based analysis. Remember, we not trying to land a man on the moon. We try diagnosing performance to derive solutions that (usually) aim to reduce the database time. I suspect that in all four cases I show, we would not be misled.

But this does highlight the requirement to also analysis performance from a non-Oracle database centric perspective. I always look at the performance situation from an operating system perspective, an Oracle centric perspective and an application (think: SQL, processes, user experience, etc.) perspective. This "3 Circle" analysis will reduce the likelihood of making a tuning diagnosis mistake. So in case DB Time is completely messed up, by diagnosing performance from the other two "circles" you will know something is not right.

If you want to learn more about my "3-Circle" analysis, here are two resources:
  1. Paper. Total Performance Management. Do an OraPub search for "3 circle" and you'll find it.
  2. Online Seminar: Tuning Oracle Using An AWR Report. I go DEEP into an Oracle Time Based Analysis but keeping it day-to-day production system practical.
In my next few articles I will drill down into why there can be a "DB Time mismatch," what to do about it and how to use this knowledge to our advantage.

Enjoy your work! There is nothing quite like analyzing performance and tuning Oracle database systems!!

Craig.





Categories: DBA Blogs

Changing The Number Of Oracle Database 12c Log Writers

Mon, 2014-12-08 22:51
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.Changing The Number Of Oracle Database 12c Log Writers
In an Oracle Database 12c instance you will likely see multiple log writer (LGWR) background processes. When you first start the Oracle instance you will likely see a parent and two redo workers. This is a very big deal and something many of us have been waiting for - for many years!

While I'm excited about the change, if I can't control the number of LGWRs I could easily find myself once again constrained by the lack of LGWRs!

So, my question is how do I manipulate the number of LGWRs from the default. And what is the default based on? It's these types of questions that led me on this quest. I hope you enjoy the read!


Serialization Is Death
Multiple LGWRs is great news because serialization is death to computing performance. Think of it like this. A computer program is essentially lines of code and each line of code takes a little bit of time to execute. A CPU can only process N lines of code per second. This means every serial executing program has a maximum through capability. With a single log writer (LGWR) background process the amount of redo that can be processed is similarly constrained.

An Example Of Serialization Throughput Limitation
Suppose a CPU can process 1000 instructions per millisecond. Also, assume through some research a DBA determined it takes the LGWR 10 instructions to process 10 KB of redo. (I know DBAs who have taken the time to figure this stuff out.) Given these two pieces of data, how many KB of redo can the CPU theoretically process per second?

? KB of redo/sec = (1000 inst / 1 ms)*(10 KB redo / 10 instr)*(1000 ms / 1 sec)* (1 MB / 1000 KB) = 1000 KB redo/sec

This is a best case scenario. As you can see, any sequential process can become a bottleneck. One solution to this problem is to parallelize.

Note: Back in April of 2010 I posted a series of articles about parallelism. If you are interested in this topic, I highly recommend you READ THE POSTS.

Very Cool! Multiple 12c LGWRs... But Still A Limit?
Since serialization is death... and parallelism is life, I was really excited when I saw on my 12c Oracle instance by default it had two redo workers in addition to the "parent" log writer. On my Oracle version 12.0.1.0.2.0 Linux machine this is what I see:
$ ps -eaf|grep prod40 | grep ora_lg
oracle 54964 1 0 14:37 ? 00:00:00 ora_lgwr_prod40
oracle 54968 1 0 14:37 ? 00:00:00 ora_lg00_prod40
oracle 54972 1 0 14:37 ? 00:00:00 ora_lg01_prod40

This is important. While this is good news, unless Oracle or I have the ability to change and increase the number of LGWR redo workers, at some point the two redo workers, will become saturated bringing us back to the same serial LGWR process situation. So, I want and need some control.

Going Back To Only One LGWR
Interestingly, starting in Oracle Database version 12.0.1.0.2.0 there is an instance parameter _use_single_log_writer. I was able to REDUCE the number LGWRs to only one by setting the instance parameter _use_single_log_writer=TRUE. But that's the wrong direction I want to go!

More Redo Workers: "CPU" Instance Parameters
I tried a variety of CPU related instance parameters with no success. Always two redo workers.

More Redo Workers: Set Event...
Using my OSM script listeventcodes.sql I scanned the Oracle events (not wait events) but was unable to find any related Oracle events. Bummer...

More Redo Workers: More Physical CPUs Needed?
While talking to some DBAs about this, one of them mentioned they heard Oracle sets the number of 12c log writers is based on the number of physical CPUs. Not the number CPU cores but the number of physical CPUs. On a Solaris box with 2 physical CPUs (verified using the command, psrinfo -pv) upon startup there was still on two redo workers.

$ psrinfo -p
2
$ psrinfo -pv
The physical processor has 1 virtual processor (0)
UltraSPARC-III (portid 0 impl 0x14 ver 0x3e clock 900 MHz)
The physical processor has 1 virtual processor (1)
UltraSPARC-III (portid 1 impl 0x14 ver 0x3e clock 900 MHz)

More Redo Workers: Adaptive Behavior?
Looking closely at the Solaris LGWR trace file I repeatedly saw this:

Created 2 redo writer workers (2 groups of 1 each)
kcrfw_slave_adaptive_updatemode: scalable->single group0=375 all=384 delay=144 r
w=7940

*** 2014-12-08 11:33:39.201
Adaptive scalable LGWR disabling workers
kcrfw_slave_adaptive_updatemode: single->scalable redorate=562 switch=23

*** 2014-12-08 15:54:10.972
Adaptive scalable LGWR enabling workers
kcrfw_slave_adaptive_updatemode: scalable->single group0=1377 all=1408 delay=113
rw=6251

*** 2014-12-08 22:01:42.176
Adaptive scalable LGWR disabling workers

It looks to me like Oracle has programed in some sweeeeet logic to adapt the numbers of redo workers based the redo load.

So I created six Oracle sessions that simply inserted rows into a table and ran all six at the same time. But it made no difference in the number of redo workers. No increase or decrease or anything! I let this dml load run for around five minutes. Perhaps that wasn't long enough, the load was not what Oracle was looking for or something else. But the number of redo workers always remained at two.

Summary & Conclusions
It appears at instance startup the default number of Oracle Database 12c redo workers is two. It also appears that Oracle has either already built or is building the ability for Oracle to adapt to changing redo activity by enabling and disabling redo workers. Perhaps the number of physical CPUs (not CPU cores but physical CPUs) plays a part in this algorithm.

While this was not my research objective, I did discover a way to set the number of redo workers back to the traditional single LGWR background process.

While I enjoyed doing the research for this article, it was disappointing that I was unable to influence Oracle to increase the number of redo workers. I sure hope Oracle either gives me control or the adaptive behavior actually works. If not, two redo workers won't be enough for many Oracle systems.

All the best in your Oracle performance endeavors!

Craig.


Categories: DBA Blogs

The Perfect Gift For The Oracle DBA: Top 5 DBA T-Shirts

Mon, 2014-12-01 17:26
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.The Perfect Gift For The Oracle DBA: Top 5 DBA T-Shirts
It's that time of year again and I can already hear it, "Dad, what do you want for Christmas?" This year I'm taking action. Like forecasting Oracle performance, I'm taking proactive action.

Like most of you reading this, you have a, let's say, unique sense of humor. I stumbled across the ultimate geek website that has an astonishing variety of t-shirts aimed at those rare individuals like us that get a rush in understanding the meaning of an otherwise cryptic message on a t-shirt.

I picked my Top 5 DBA Geek T-Shirts based on the challenges, conflicts and joys of being an Oracle DBA. With each t-shirt I saw, a story came to mind almost immediately. I suspect you will have a similar experience that rings strangely true.

So here they areā€”the Top 5 T-Shirts For The Oracle DBA:
Number 5: Change Your Password
According to Slash Data the top password is now "Password".  I guess the upper-case "P" makes people feel secure, especially since last years top password was "123456" and EVERYBODY knows thats a stupid password. Thanks to new and improved password requirements, the next most popular password is "12345678". Scary but not surprising.

As Oracle Database Administrators and those who listened to Troy Ligon's presentation last years IOUG conference presentation, passwords are clearly not safe. ANY passwords. Hopefully in the coming years, passwords will be a thing of the past.


Number 4: Show Your Work
Part of my job as a teacher and consultant is to stop behavior like this: I ask a DBA, "I want to understand why you want to make this change to improve performance." And the reply is something like one of these:

  1. Because it has worked on our other systems.
  2. I did a Google search and an expert recommended this.
  3. Because the box is out of CPU power, there is latching issues, so increasing spin_count will help.
  4. Because we have got to do something and quick!

I teach Oracle DBAs to think from the user experience to the CPU cycles developing a chain of cause and effect. If we can understand the cause and effect relationships, perhaps we can disrupt poor performance and turn it to our favor. "Showing your work" and actually writing it down can be really helpful.

Number 3: You Read My T-Shirt
Why do managers and users think their presence in close proximity to mine will improve performance or perhaps increase my productivity? Is that what they learn in Hawaii during "end user training"?

What's worse is when a user or manager wants to talk about it...while I'm obviously in concentrating on a serious problem.

Perhaps if I wear this t-shirt, stand up, turn around and remain silent they will stop talking and get the point. We can only hope.

Number 2: I'm Here Because You Broke Something
Obnoxious but true. Why do users wonder why performance is "slow" when they do a blind query returning ten-million rows and then scroll down looking for the one row they are interested in.... Wow. The problem isn't always the technology... but you know that already.

Hint to Developers: Don't let users do a drop down or a lookup that returns millions or even thousands or even hundreds of rows... Please for the love of performance optimization!


Number 1 (drum roll): Stand Back! I'm Going To Try SCIENCE
One of my goals in optimizing Oracle Database performance is to be quantitative. And whenever possible, repeatable. Add some basic statistics and you've got science. But stand back because, as my family tells me, it does get a little strange sometimes.

But seriously, being a "Quantitative Oracle Performance Analyst" is always my goal because my work is quantifiable, reference-able and sets me up for advanced analysis.


So there you go! Five t-shirts for the serious and sometimes strange Oracle DBA. Not only will these t-shirts prove and reinforce your geeky reputation, but you'll get a small yet satisfying feeling your job is special...though a little strange at times.

All the best in your Oracle performance endeavors!

Craig.
Categories: DBA Blogs

Off May Not Be Totally Off: Is Oracle In-Memory Database 12c (12.1.0.2.0) Faster?

Mon, 2014-11-17 21:56
This page has been permanently moved. Please CLICK HERE to be redirected.

Thanks, Craig.Off May Not Be Totally Off: Is Oracle In-Memory Database 12c (12.1.0.2.0) Faster?
Most Oracle 12c installations will NOT be using the awesome Oracle Database in-memory features available starting in version 12.1.0.2.0. This experiment is about the performance impact of upgrading to 12c but disabling the in-memory features.

Every experiment I have performed comparing buffer processing rates, clearly shows any version of 12c performs better than 11g. However, in my previous post, my experiment clearly showed a performance decrease after upgrading from 12.1.0.1.0 to 12.1.0.2.0.

This posting is about why this occurred and what to do about it. The bottom line is this: make sure "off" is "totally off."

Turn it totally off, not partially off
What I discovered is by default the in-memory column store feature is not "totally disabled." My experiment clearly indicates that unless the DBA takes action, not only could they be a license agreement violation but a partially disabled in-memory column store slightly slows logical IO processing compared to the 12c non in-memory column store option. Still, any 12c version processes buffer faster than 11g.

My experiment: specific and targeted
This is important: The results I published are based on a very specific and targeted test and not on a real production load. Do not use my results in making a "should I upgrade decision." That would be stupid and an inappropriate use of the my experimental results. But because I publish every aspect of my experiment and it is easily reproducible it is a valid data point with which to have a discussion and also highlight various situations that DBAs need to know about.

You can download all my experimental results HERE. This includes the raw sqlplus output, the data values, the free R statistics package commands, spreadsheet with data nicely formatted and lots of histograms.

The instance parameter settings and results
Let me explain this by first showing the instance parameters and then the experimental results. There are some good lessons to learn!

Pay close attention to the inmemory_force and inmemory_size instance parameters.

SQL> show parameter inmemory

NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
inmemory_clause_default string
inmemory_force string DEFAULT
inmemory_max_populate_servers integer 0
inmemory_query string ENABLE
inmemory_size big integer 0
inmemory_trickle_repopulate_servers_ integer 1
percent
optimizer_inmemory_aware boolean TRUE

SQL> show sga

Total System Global Area 7600078848 bytes
Fixed Size 3728544 bytes
Variable Size 1409289056 bytes
Database Buffers 6174015488 bytes
Redo Buffers 13045760 bytes

In my experiment using the above settings the median buffers processing rate was 549.4 LIO/ms. Looking at the inmemory_size and the SGA contents, I assumed the in-memory column store was disabled. If you look at the actual experimental result file "Full ds2-v12-1-0-2-ON.txt", which contain the explain plan of the SQL used in the experiment, there is no mention of the in-memory column store being used. My assumption, which I think is a fair one, was that the in-memory column store had been disabled.

As you'll see I was correct, but only partially correct.

The parameter settings below are when the in-memory column store was totally disabled. They key is changing the default inmemory_force parameter value from DEFAULT to OFF.

SQL> show parameter inmemory

NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
inmemory_clause_default string
inmemory_force string OFF
inmemory_max_populate_servers integer 0
inmemory_query string ENABLE
inmemory_size big integer 0
inmemory_trickle_repopulate_servers_ integer 1
percent
optimizer_inmemory_aware boolean TRUE
SQL> show sga

Total System Global Area 7600078848 bytes
Fixed Size 3728544 bytes
Variable Size 1291848544 bytes
Database Buffers 6291456000 bytes
Redo Buffers 13045760 bytes

Again, the SGA does not show any in-memory memory space. In my experiment with the above "totally off" settings, the median buffers processing rate was 573.5 LIO/ms compared to "partially off" 549.4 LIO/ms. Lesson: Make sure off is truly off.

It is an unfair comparison!
It is not fair to compare the "partially off" with the "totally off" test results. Now that I know the default inmemory_force must be changed to OFF, the real comparison should be made with the non in-memory column store version 12.1.0.1.0 and the "totally disabled" in-memory column store version 12.1.0.2.0. This is what I will summarize below. And don't forget all 12c versions showed a significant buffer processing increase compared to 11g.

The key question: Should I upgrade?
You may be thinking, if I'm NOT going to license and use the in-memory column store, should I upgrade to version 12.1.0.2.0? Below is a summary of my experimental results followed by the key points.

1. The non column store version 12.1.0.1.0 was able to process 1.1% more buffers/ms (median: 581.7 vs 573.5) compared to to "totally disabled" in-memory column store version 12.1.0.2.0. While this is statistically significant, a 1.1% buffer processing difference is probably not going to make-or-break your upgrade.

2. Oracle Corporation, I'm told, knows about this situation and is working on a fix. But even if they don't fix it, in my opinion my experimental "data point" would not warrant not upgrading to the in-memory column store version 12.1.0.2.0 even if you are NOT going to use the in-memory features.

3. Visually (see below) the non in-memory version 12.1.0.1.0 and the "totally off" in-memory version 12.1.0.2.0 samples sets look different. But they are pretty close. And as I mentioned above, statistically they are "different."

Note for the statistically curious: The red color 12.1.0.1.0 non in-memory version data set is highly variable. I don't like to see this in my experiments. Usually this occurs when a mixed workload sometimes impacts performance, I don't take enough samples or my sample time duration is too short. To counteract this, in this experiment I captured 31 samples. I also performed the experiment multiple times and the results where similar. What I could have done was used more application data to increase the sample duration time. Perhaps that would have made the data clearer. I could have also used another SQL statement and method to create the logical IO load.
What I learned from this experiment
To summarize this experiment, four things come to mind:

1. If you are not using an Oracle Database feature, completely disable it. My mistake was thinking the in-memory column store was disabled when I set it's memory size to zero and "confirmed" it was off by looking at the SGA contents.

2. All versions of 12c I have tested are clearly faster at processing buffers than any version of 11g.

3. There is a very slight performance decrease when upgrading from Oracle Database version 12.1.0.1.0 to 12.1.0.2.0.

4. It is amazing to me that with all the new features poured into each new Oracle Database version the developers have been able to keep the core buffer processing rate nearly at or below the previous version. That is an incredible accomplishment. While some people may view this posting as a negative hit against the Oracle Database, it is actually a confirmation about how awesome the product is.

All the best in your Oracle performance tuning work!

Craig.




Categories: DBA Blogs