Re: how to compare Oracle vs SQL-Server?
Date: 1996/08/12
Message-ID: <01bb8803$d43e9050$30d7aec7_at_nittany>
Michael,
Here's an IDC whitepaper comparing cost of ownership between Oracle and SQL
Server. It was sponsored by MS, put IDC should be relatively independant...
Copyright International Data Corporation, 1996
Analysts: Chris Christiansen and Dan Kusnetzky
An IDC White Paper Sponsored by Microsoft
Executive Summary
As the overlap increases between Microsoft SQL Server and Oracle7 Server,
database selection is becoming more complicated. Microsoft’s SQL Server
product increasingly competes with Oracle, under 50 concurrent users. The
overlap of Microsoft SQL Server and Oracle7 Server is a function of both
products’ expansion. Oracle7 Server is moving in two directions—upward
(larger databases and more concurrent users) and downward into the
Microsoft SQL Server area.
Meanwhile, Microsoft SQL Server is moving upwards. The evolution of robust
hardware and operating systems in the PC LAN area is propelling the
Microsoft SQL Server database product into production environments. Our
research shows that in surveyed sites, the types of applications running on
both database environments are roughly similar (see Table 1).
Using IDC’s Cost-to-Use methodology, we attempted to better understand the
costs of these two products by studying three metrics:
· Cost per site · Cost per server · Cost per database size
IDC Opinion
IDC believes that Microsoft SQL Server and Oracle’s Oracle7 Server
increasingly compete for some of the same customers. While Oracle7 Server’s
base contains more large installations, it and SQL Server overlap below 50
concurrent users in production environments. Although we believe Oracle7
Server offers higher functionality in some areas, Microsoft SQL Server is
making significant progress in reliability, scalability, and availability.
Our Cost-to-Use research shows that for under 50 concurrent users,
Microsoft SQL Server five-year ownership costs are 44% lower on a per-site
basis and equal on a per-server basis.
Table 1 Database Applications in Production (multiple responses allowed)
Oracle7 Server (Count) Oracle7 Server (%) Microsoft SQL Server (Count) Microsoft
SQL Server (%)
Accounting/financial 30 58 28 52 Sales/customer tracking 11 21 17 31 Other 14 27 17 31 Documentation/file management 11 21 13 24 Decision support 11 21 12 22 Data warehousing 14 27 12 22 Executive information systems (EIS) 10 19 9 17 Transaction processing 10 19 9 17 Order processing 15 29 8 15 Customer service 6 12 7 13 Manufacturing 7 13 2 4 Branch office automation 3 6 2 4 Retail 1 2 0 0 Total 52 100 54 100
Source: International Data Corporation, 1996 We note the following findings specific to the class of sites surveyed: · Microsoft SQL Server is 44% less expensive on a per-site basis when overall costs (hardware, software, network, support staff, installation, and training) are considered over a five-year period. · The biggest cost differences were in the areas of software, help desk/user support, and applications development. · On a per-server basis, costs are equal because Oracle sites contained more servers than Microsoft sites.
· Average Oracle7 Server database sizes were larger than those of Microsoft SQL Server. On a Cost-to-Use per gigabyte (GB) of database storage basis, Microsoft held a clear lead in sites with databases under 1GB. In the 1GB to 10GB range, the two databases were roughly equal. Above 10GB, Oracle7 Server showed an advantage over Microsoft SQL Server.
Research Highlights
IDC developed the Cost-to-Use (CtU) methodology to measure “real-worldâ€
computing costs (hardware, software, networking, support staff,
installation, and training). Continuing IDC’s Cost-to-Use research on
client/server, we recently studied U.S. five-year ownership costs in two
database environments, Microsoft SQL Server and Oracle7 Server. IDC
obtained average configuration and support staffing from phone interviews
with 106 respondents (54 for Microsoft and 52 for Oracle). These results
were normalized to configurations based on 43 concurrent users. Within the
boundaries of this study, we believe these results represent a fair and
accurate representation of typical cost structures in each database
environment.
IDC’s survey studied costs and did not probe on other aspects of value that
are frequently critical to purchase decisions. Undoubtedly, buyers should
assess database products based on the application, workload, and
application development skill sets. Nonetheless, we believe that our
Cost-to-Use metric represents a meaningful input in the overall buying
decision.
Moreover, this analysis focuses on the comparison between Microsoft SQL
Server and Oracle7 Server. The study does not take into account Oracle’s
Workgroup Server database. This product was explicitly designed to compete
at the low end of the market. It is likely that its costs would be less
because of packaging (i.e., pricing and functionality).
Based on the survey data, IDC chose the following hardware and software as
representative of typical system configurations:
· Microsoft’s SQL Server running on two Compaq ProLiant 1500 (133MHz
Pentium) systems with Microsoft’s Windows NT Server operating systems.
· Oracle’s Oracle7 Server running on three HP 9000 Model E25 systems with
HP-UX (Unix) operating systems.
The survey data showed that the average Oracle7 Server site contained three
servers compared with two systems for SQL Server. The difference in server
counts places the Oracle7 Server software at a cost disadvantage to
Microsoft SQL Server. However, staffing costs were unaffected because the
data is based on the average number of support personnel at the
respondents’ sites.
Hardware, software, and network configurations often evoke debate. Even if
all three of these cost elements were equal for Microsoft SQL Server and
Oracle7 Server, the results would remain largely unchanged. Our results
indicate that five-year staffing expenditures (typically 50% to 70% of
total five-year costs) severely diminish the importance of all other
expenditures. As such, buyers should focus most of their attention on
staffing costs.
This study is based on concurrent users, as derived from the survey.
(Oracle7 Server databases in the sample supported an average of 48
concurrent users, with Microsoft SQL Server databases supporting 38
concurrent users.) Concurrent users (or the average number of users that
are simultaneously active on the database at any given time) is an
appropriate criterion for transaction processing environments. This study’s
results has limited applicability to non-production environments (e.g.
decision support).
Key Findings
Although Microsoft SQL Server is often perceived as a decision support
platform versus Oracle7 Server’s production reputation, IDC survey results
show:
· Microsoft SQL Server was used for mainstream applications in roughly the
same way as Oracle7 Server. Order processing was the only notable
exception, with Oracle7 Server in use twice as often as SQL Server, as
Table 1 indicates.
· Microsoft SQL Server is much less expensive to use (on a per site basis)
than Oracle7 Server in production database environments with less than 50
concurrent users.
· Oracle7 Server sites contained more database servers (3.00) than
Microsoft SQL Server sites (1.73). We had expected to see fewer Oracle7
Servers per site because of a tendency to consolidate database servers.
This difference caused IDC to calculate costs in two ways: per site and per
server. When IDC calculated per-site costs, Microsoft SQL Server was 44%
less expensive than Oracle7 Server. However, when we normalized the data on
a per-server basis, Oracle7 Server and Microsoft SQL Server were equal in
cost. These figures are based on five-year costs, as shown in Figure 1.
· In the software category, Oracle7 Server was more expensive because of
license fees, software support, and HP-UX (HP’s Unix operating system).
Software accounts for only 1% of Microsoft SQL Server costs and 4% of
Oracle7 Server total expenditures.
· Microsoft SQL Server requires fewer personnel in the user support/help
desk area. Oracle7 Server’s costs were significantly greater than SQL
Server’s (by more than twice). This single cost item accounted for 18% of
Microsoft SQL Server costs, but consumed 28% of Oracle7 Server sites’
expenditures. This represents the single largest expenditure for both
vendors.
· The applications development category also showed a large variance
between the two vendors. Microsoft SQL Server requires fewer developers,
and its costs were much less than those of Oracle7 Server. This item was
Microsoft SQL Server’s fourth largest cost (14% of total) and Oracle7
Server’s second greatest expenditure (25% of total).
· Total internal staffing expenditures (excluding hardware, software,
network, external installation charges, and training) accounted for 74% of
Microsoft SQL Server’s costs versus 79% for Oracle7 Server. Essentially,
this means that Microsoft SQL Server requires fewer personnel to support
the same-sized user population.
As always, these results should be used with caution. We believe that
Oracle7 Server sites are more mature and tended to have more support
personnel. The data also indicates that Oracle7 Server sites supported
larger databases. (In our study, Oracle7 Server databases were roughly
19.3GB versus Microsoft SQL Server’s average of 5.4 GB.)
Even though large database sizes don’t imply heavy workloads, we believe
that the surveyed Oracle7 Server sites support larger workloads. This
supposition may account for Oracle7 Server’s greater number of servers per
site.
Based on the survey data, we believe this report’s findings are largely
applicable to sites with 10 to 50 concurrent users. Because surveyed sites
had an average of 43 concurrent users, we believe our results are less
relevant to environments with 50 to 100 concurrent users. This survey only
a found a few sites with over 100 concurrent users for both database
products. Therefore, this study has no applicability to sites with more
than 100 concurrent users.
Analysis of Site- and Server-Based Cost-to-Use
Many users now focus their acquisition strategies on individual server
purchases. Users increasingly buy computer equipment and employ support
personnel at the business-unit level. Many times, they add capacity by
getting another server. Traditional IT departments that formerly controlled
“site-wide†purchases now focus on high-end systems and don’t participate
in many workgroup, departmental, or even divisional server purchases. For
these reasons, we also analyzed our cost data on a per-server basis.
Figure 1 contrasts the per-site results with the per-server figures. The
per-server numbers are based on dividing the per-site figures by the
average number of servers per site, as indicated by the survey’s
respondents. (Dividing the $/month/user figures in the per-site columns by
two Microsoft and three Oracle servers does not yield the per-server costs
because of rounding and differences in staffing salaries by job
description.) When cost per server is measured, Microsoft SQL Server and
Oracle7 Server costs are equal, as shown in Figure 1.
Users should note that per-server calculations assume that every user
accesses every server. This assumption may not be true in all situations.
If a select portion of users only log onto one specific server, the
per-server cost calculation would differ sharply.
Figure 1
Database Cost-to-Use
Costs for 43 Concurrent Users over Five Years
1 Includes training (IS/DBA only); staff costs for management, backup
management, database performance monitoring, and all other database
administration and operations; and internal and external installation costs
2 Includes hardware, software, and networking
Source: International Data Corporation, 1996
It is not unreasonable to expect that the architectural complexity of the
Oracle7 Server environment is greater than the Microsoft SQL Server site.
We believe that the per-server costs somewhat normalize this complexity,
and users should consider the figures that are most applicable to their
individual situations.
As shown in Figure 1, the per site results indicate that the Microsoft SQL
Server exhibited the lowest cost per month per user. Microsoft SQL Server
costs $422/month/user versus Oracle7 Server’s $608, a $186 difference. This
graphic clearly shows the role of two staff categories (user support/help
desk and applications development). As our experience indicates, the
salaries for these personnel are the costliest component of most IT
managers’ budgets.
We believe Microsoft SQL Server is extremely cost competitive with Oracle7
Server because of its integrated functionality. For example, Microsoft
developed the operating system, Windows NT Server, the database, and the
Windows client software. Because Oracle7 Server concentrates on database
software and does not control server or client systems software
development, the lower degree of integration and the resulting higher
support costs are not surprising. (Both databases support an equally wide
array of clients, but Oracle7 Server supports more server platforms.)
· Hardware (per site)—As shown in Table 2, hardware costs (including
warranty and maintenance) were roughly the same despite the difference in
the number of servers. Microsoft SQL Server was configured with two
servers, and Oracle7 Server configurations contained three servers.
However, if we configured only two HP servers, the resulting change of a
few dollars would have little effect.
Servers are not the largest hardware cost item. At $38.08 for both vendors, client PCs account for 65–75% of hardware costs. (Both vendors’ client configurations are identical.) If the client costs were subtracted, hardware for both vendors shrinks to 3–4% of total CtU expenditures (from 12% for Microsoft SQL Server and 10% for Oracle).
Table 2 Cost and Percentage Breakdown of Database CtU by Site and Server
(costs for 43 users over five years)
Per Site Per
Server
SQL Server Oracle7 Server SQL Server Oracle7 Server $/month/ user % of Total $/month/ user % of Total $/month/ user % of Total $/month/ user % of Total Hardware 51 12 59 10 29 13 20 9 Software 6 1 26 4 3 1 9 4 Networking 5 1 6 1 3 1 2 1 Staff: management 68 16 73 12 34 15 28 12 Staff: user support/ Help desk 75 18 168 28 39 17 63 28 Staff: backup management 22 5 19 3 12 5 7 3 Staff: DB perf. monitoring 13 3 20 3 $7 3 8 4 Staff: all other DB. admin. and operations 65 15 30 5 33 15 12 5 Staff: Applications Development 58 14 155 25 30 13 58 25 Staff: installation (internal) 12 3 12 2 7 3 4 2 Staff: installation (external) 38 9 29 5 22 10 10 4 Training: IS/DBA only 9 2 11 2 5 2 4 2 Total? 422 99 608 100 224 98 225 100 % Change 44% 0%
Note Dividing the $/month/user figures in the per-site columns by two Microsoft and three Oracle servers does not yield the per-server costs because of rounding and differences in staffing salaries by job description. Totals may not be exact due to rounding. Source: International Data Corporation, 1996 · Software (per site)—These costs (including monthly support costs) were not a significant portion of total costs (1% for Microsoft SQL Server and 4% for Oracle). The difference was due to Oracle7 Server’s and HP-UX’s higher license and support charges. We configured Oracle7 Server on HP-UX because most survey respondents ran Oracle7 Server on midrange Unix systems.
· Staff: management (per site)—At roughly 16% of the total for Microsoft SQL Server and 12% for Oracle, this category was Mic Received on Mon Aug 12 1996 - 00:00:00 CEST