With all the software upgrade choices made, it is time to look at the server hardware. This section is a deep dive comparing the technical merits of five servers deemed capable of running Oracle 11g database with good performance.
Read the other sections of this guide on servers for Oracle upgrades
Requires Free Membership to View
Oracle database upgrades: the in-place vs. migration upgrade decision
Choosing the right server hardware is all about choosing the right software
The best of the Oracle 11g-ready servers
Can Sun shine running Oracle 11g?
Several vendors, such as Dell, HP, IBM, Lenovo and SuperMicro offer servers that meet or exceed the needs of a departmental Oracle database, each of those vendors' products has its pros and cons.
Tale of the tape among Oracle 11g server contenders
| Server: | Dell Power Edge T410 | HP Proliant ML330 G6 | IBM System X3400 M2 | Lenovo ThinkServer | SuperMicro 2021M-U2R+ |
|---|---|---|---|---|---|
| Processors | (2) Intel Xeon E5504 2Ghz | (2) Intel Xeon E5504 2Ghz | (2) Intel Xeon E5504 2Ghz | (2) Intel Xeon E5504 2Ghz | (2) AMD 1.70GHz Opteron 2344HE |
| Ram | (4) 1Gbyte 1066MHz, RDIMMs | (2) 2Gbyte 1066MHz, RDIMMs | (2) 2GB PC2-5300 ECC 667MHz DIMM | (4) 1 GB 667 Mhz ECC DDR2 FBDIMM | (4) 1 GB 667 Mhz DDR2 ECC FBDIMM |
| Storage | (2) 500GB 7.2k RPM Serial ATA 3Gbps Hard Drives | (2) 500GB 3G SATA 7.2K Hot Plug Hard Drives | (2) 500GB 7.2K RPM 3.5" Simple-Swap SATA Drives | (2) 500 GB SATA 7.2K 3.5'' 3Gbps Hard Drives | (4) 250Gb SATA2 SATA 7.2K 3.5'' 3Gbps Hard Drives |
| Optical Drive Peripherals | DVD R/W USB Keyboard & Mouse | DVD R/W USB Keyboard & Mouse | DVD Rom USB Keyboard & Mouse | DVD Rom USB Keyboard & Mouse | DVD Rom USB Keyboard & Mouse |
| Networking | On Board Gigabit Ethernet | Gigabit Dual Port Network | Integrated Gigabit Ethernet | Single Port Gigabit Ethernet | Integrated Gigabit Dual Port Ethernet |
| Networking | DVD R/W USB Keyboard & Mouse | DVD R/W USB Keyboard & Mouse | DVD Rom USB Keyboard & Mouse | DVD Rom USB Keyboard & Mouse | DVD Rom USB Keyboard & Mouse |
| Power Supply | Card Redundant Dual Power Supplies | Adapter Redundant Dual Power Supplies | Single 670 Watt Power Supply | Single 750 Watt Power Supply | Redundant Dual Power Supplies |
| Form Factor | Tower Case | Tower Case | 5U Tower Case | 4U Tower Case | 2U Rack Mount |
| Direct Price | $2424 | $2488 | $2662 | $2159 | $2251 |
Administrators will find it pretty straightforward to use online configuration tools from the respective vendors to customize a server from the leading manufacturers to meet their needs. We took a look at five servers; all customized in a similar fashion, offering similar features and coming in at similar price points. At first blush, the servers selected all seem very much alike, but upon closer inspection of the specifications, some differences can be revealed, which will affect performance.
Performance can be compared by installing an operating system and running a benchmarking program. We chose to use the 64-bit edition of Windows Server 2008R2 (Release Candidate) and the 64-bit Edition of Performance Test V7.0 from PassMark software. Benchmarks prove to be a useful way to determine the value of a system.
Faster, higher, stronger: How Oracle 11g servers stack up
| Server: | Dell Power Edge T410 | HP Proliant ML330 G6 | IBM System X3400 M2 | Lenovo ThinkServer | SuperMicro 2021M-U2R+ |
|---|---|---|---|---|---|
| CPU Mark | 3921.3 | 3955 | 3877.1 | 3866.2 | 3472.1 |
| Memory Mark | 672.4 | 681.1 | 529 | 541 | 455 |
| Disk Mark | 422 | 453 | 407 | 417 | 459 |
| PassMark | 1207.3 | 1258.7 | 1182.2 | 1176.1 | 997.3 |
CPU performance, memory performance and disk performance are the key performance metrics for determining how well a server may handle a database. Our benchmarks show that a pair of 2 GHz Intel's Xeon CPUs will outperform a pair of 1.70Ghz AMD Opteron CPUs by about 14%. But a pair of 1.70Ghz AMD Opetrons costs about 40% less than a pair of Intel's 2Ghz Xeons, showing that while the AMD may not lead in speed, it does lead in value.
In the real world, unless a server is running at close to 100% utilization, most benchmarks will prove to be academic at best at determining how well a server will fit the performance needs of an organization. Administrators will find that CPU performance and memory performance may very well have little impact on the performance of a SQL database, which is affected more by disk I/O and network bandwidth than an operating system may be.
The simplest way to boost the performance of a database may come from just using faster drive technology, such as Serial Attached SCSI, RAID arrays or a high performance Storage Area Network. But, those technologies can significantly increase prices.
For the most part, our benchmarks showed that SATA drives of equal capacities performed within a few percentage points of each other. On the other hand memory speeds varied significantly, simply because different types of RAM were used in the tested servers. Faster servers used 1066 MHz memory modules, while the slower performers used 667 Mhz modules.
For those pursuing the Oracle upgrade path, The HP Proliant ML330 G6 arguably offers the best price-performance attributes, while the SuperMicro 2021M-U2R+ proves to be the best value. Both the HP and the SuperMicro servers offer redundant power supplies and hot swap-able hard drives, as well as dual gigabit Ethernet ports. That brings the units into the realm of high availability and adds an extra bit of peace of mind for the harried administrator.
The Dell Power Edge T410 is a real contender here, simply because Dell frequently runs specials, which can shave several hundred dollars off the purchase price. The IBM and Lenovo units are also good choices when one considers that the systems come with three-year warranties and onsite service. That just goes to show, that any one of the tested servers will be more than adequate to meet the needs of a busy departmental Oracle server and the final choice will ultimately be based upon corporate policy and personal preferences.
This was first published in July 2009

Join the conversationComment
Share
Comments
Results
Contribute to the conversation