RE: Open Source Database Questions

From: Matthew Zito <mzito_at_gridapp.com>
Date: Tue, 9 Jun 2009 18:17:08 -0400
Message-ID: <C0A5E31718FC064A91E9FD7BE2F081B1021A34C6_at_exchange.gridapp.com>



Moreso the former, in that they say it's compliant with TPC-C spec X.Y, and they have some bright people working on the project, but no one from the TPC organization have put their official stamp of approval. Presumably they have re-implemented the same schema, query requirements, keying, etc. as the TPC-C, but like many open-source projects, I certainly haven't dug into the source code and looked, nor am I aware of those who have.  

The other thing that "audited" TPC-C/H/E results offer is that there's a whole slew of rules above and beyond the spec itself that the solution has to abide by. You have to provide validated, real-world pricing, you have to disclose all of your configuration information, you have to physically have someone from one of the two auditing firms there while you do the runs, you have to pass recoverability tests, etc. etc, all of which are incredibly involved and make you expose a fair bit of the secret sauce of how you do things.  

This is why many startups/more proprietary/non-corporate funded organizations don't bother to deal with TPC results. The way they look at it, it's just a "speeds and feeds" argument, and most people today buy on functionality, manageability, supportability, etc. etc., and the opportunity cost to do one of these audited engagements is very significant.  

So, things like jTPCC are totally useless for comparing against audited TPC-C results, since it's only a small piece of the puzzle. But, if you want to arbitrarily benchmark your database against two different storage arrays, or two different OSes on the same hardware, etc. etc. - i.e. jTPCC always against other jTPCC results, you can feel confident you're getting something that roughly approximates and is probably at least very very close, the same TPC-C spec that is considered the "gold standard" for OLTP workloads.  

Matt  


From: Niall Litchfield [mailto:niall.litchfield_at_gmail.com] Sent: Tuesday, June 09, 2009 6:08 PM
To: Matthew Zito
Cc: Richard.Goulet_at_parexel.com; regdba_at_yahoo.com; Oracle-l Subject: Re: Open Source Database Questions  

so does that mean that jtpcc scores are non-audited in the 'smart people verify it and show how it's done, but we don't pay them' sense, or in the 'we published a score and called it jtpcc' sense. If it's repeatable and clear then in my book it counts. If it's the same class as "8 out of 10 cat owners say their cats prefer it" as the second approach is well, chalk one up to big business.  

Niall

On Tue, Jun 9, 2009 at 4:14 PM, Matthew Zito <mzito_at_gridapp.com> wrote:  

The open-source database options at this point - postgres, mysql, ingres, etc. all have totally valid use cases in the enterprise today. Basically, Postgres is the most stable, as Dick says, similar concepts to Oracle, but not nearly as flexible as MySQL and with fewer features. MySQL offers a couple of different kinds of replication and clustering, different types of tablespaces, interchangeable storage engines, etc., but is comparably stable to Postgres. Ingres is a little more nichey, but definitely workable.

The big differentiation is the amount of community knowledge - there's a lot of good information on tuning, configuring MySQL out in the wild, and if you wanted to purchase commercial support for MySQL there's a couple of different avenues. For PostgreSQL your best bet is to go get EnterpriseDB, which claims Oracle compatibility, though my understanding is that "compatibility" is relative.

The reason that they're using JTPCC is that it costs money to have an "audited" TPC-C run, and non-trivial amounts. Consequently, a lot of open-source projects can't afford or don't want to try to compete with certified benchmarks, so instead they use non-audited reimplementations of the various benchmarks.

Matt

--

Matthew Zito
Chief Scientist
GridApp Systems
P: 646-452-4090
mzito_at_gridapp.com
http://www.gridapp.com <http://www.gridapp.com/>

-----Original Message-----

From: oracle-l-bounce_at_freelists.org on behalf of Goulet, Richard Sent: Tue 6/9/2009 11:06 AM
To: regdba_at_yahoo.com; Oracle-l
Subject: RE: Open Source Database Questions

Peter,

        From past experience (which had the same motives) take a serious look at PostGreSql(http://www.postgresql.org/). The latest versions have a lot in common with Oracle, though not everything, so it's a lot easier to transition from one to the other.

Dick Goulet
Senior Oracle DBA
PAREXEL International

-----Original Message-----

From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Peter Barnett Sent: Tuesday, June 09, 2009 10:53 AM
To: Oracle-l
Subject: Open Source Database Questions

Because of the current economy we are being encouraged to consider open source databases.

One of the vendors has referenced a JTPCC benchmark which is an open source Java version of TPCC. Has anyone heard of JTPCC? Is it in any way truly comparable to TPCC? Keep in mind that I am skeptical of TPCC so this may be a loaded question.

Is anyone currently using open source databases in production? Our typical database is approximately 200G but the range is 50G to 3.5T. One of our typical databases has approximately 200 distinct users at any one time. Any thoughts on a sweet spot for open source databases?

This does all come down to money! I am pretty sure that we will have an open source database inhouse by year end. The major question is how best to use it.

Thanks,

Pete Barnett
Database Technologies Lead
Regence      

--

http://www.freelists.org/webpage/oracle-l

--

http://www.freelists.org/webpage/oracle-l

--

Niall Litchfield
Oracle DBA
http://www.orawin.info

--

http://www.freelists.org/webpage/oracle-l Received on Tue Jun 09 2009 - 17:17:08 CDT

Original text of this message