I'm currently hosting a game server with the current specification E1230 32GB Ram 500Gbx4 SATA Windows 2008 server SQL Server 2012 Databse size is 10Gb Number of players 1k

  • The problem at first I had I/O bottleneck with 2x1Tb as I was using the default SQL configuration and all the files .mdf and .ldf were stored in default C:/SQLSERVER
  • Then I added 2 more 1 Tb disks and I isolated .mdf and .ldf and TempDB locations. Problem has been fixed and I/O is performing well.
  • Now I have the opportunity to change the disks system to 4x 300GB SAS in HW RAID 10, the problem I'm limited to 4 Disks 1 Array only, so I will have to store mdf, ldf and tempdb files in the same array and not isolated.
  • Will I have I/O issue with this configuration? it will be better than 4x single HDD's which every physical HDD contains mdf, ldf, and tempdb files which I already use at the moment without issue or performance could be worse with shared 4x SAS 300GB RAID 10?
  • The answer here is going to partially depend on the actual hardware being used. What's the RAID card model? Does it have a BBU? What's the model number of the drives? 7.2k, 10k or 15k? 3Gbps or 6Gbps SAS? If 6, do the backplane and RAID card support 6? Have you measured the throughput of the existing set of drives? Have you compared that throughput to what could be provided by the hardware RAID? – Charles Aug 20 '12 at 4:59
  • Whatever configuration you end up with, ad an extra couple of drives as hot spares. – John Gardeniers Aug 20 '12 at 5:10
  • Not sure about the RAID card model but I know that it has BBU. Drives are 10k disks 6GBps. Not sure if the RAID card support 6. Here is a screenshot "s9.postimage.org/70dqzzdgf/ioperf.png"; of the current setup in action with 4x500GB, mdf, ldf and tempdb files isolated on a physical drive.What is worrying me with the Raid 10 array that I will not be able to isolate the mdf, ldf and tempdb files which may cause poor read/write performance. – bissa Aug 20 '12 at 8:39

If the database size is only 10 GB, a RAID1 array of two small SSDs makes far more sense than anything you can do with spinning disks. Because this is a database server, write endurance may be an issue with MLC flash drives. Be sure to use either SLC flash drives (e.g. Intel 313 series) or MLC flash drives that let you track wear percentage via SMART (e.g. Intel 320 series).

To answer your original question: you can expect 500-700 read IOPS or 250-350 write IOPS from your 4-disk RAID10 array, depending on unknowns including whether your SAS drives operate at 10K or 15K rpm. Your old 7.2K rpm SATA drives probably gave you around 75 random read/write IOPS each. In other words, your performance probably would improve when switching to the 4-disk RAID10 array, particularly if you keep your logs on a separate physical volume, perhaps taking advantage of a mirrored pair of the old 7200 rpm drives for this purpose.

Your RAID1 array of two SSDs could deliver something on the order of 4,000 to 20,000+ IOPS depending on the model of SSD selected, but only if your RAID controller can keep up.

  • Appreciate your reply, I'm only limited to 4 disks in the server chassis. Here is a screenshot "s9.postimage.org/70dqzzdgf/ioperf.png" of the current setup in action with 4x500GB, mdf, ldf and tempdb files isolated on a physical drive.What is worrying me with the Raid 10 array that I will not be able to isolate the mdf, ldf and tempdb files. – bissa Aug 20 '12 at 8:31
  • That's quite a limitation! If cost is a driving concern, how about connecting an external eSATA enclosure for the old drives? Keep in mind that the logs are meant to be segregated not only for performance reasons, but also for disaster recovery purposes. – Skyhawk Aug 20 '12 at 21:03
  • I was thinking is this configuration, as I'm limited to 4 HDD's only in the server, I already ordered the SAS disks with RAID10 HW controller from the host provider but I can change the RAID configuration anytime. Looks RAID 10 4x300GB (1 partition 600GB) will not be ideal for storing OS, mdf, ldf and all the files or? So I was thinking to change the configuration to: C:/ OS (300GB SAS 1 Disk) D:/ MDF database files (RAID 1) (300GB SAS 2 Disks) E:/ LDF files and tempfb files (300GB SAS 1 Disk) ..... Will this configuration work better than the RAID 10 ARRAY? – bissa Aug 21 '12 at 0:06
  • Don't do that. You want RAID for fault tolerance. If you are limited to 4 physical disks, try the RAID 10 with battery backed write caching enabled. Consider solid state drives for your next upgrade. Your only other viable option is two RAID1 arrays, one for OS/logs and the other for data. How many write operations per second are you seeing on your existing log drive? Your performance data showed bytes/sec but not operations/sec. – Skyhawk Aug 21 '12 at 5:51
  • Thank you Miles, host provider confirmed with me that the controller will be LSi 9260 with BBU so I'm going to enable the write caching and the disks will be 15k RPM. I will test first the 4 Disks ARRAY RAID 10 with all the files stored on it and I will see how the performance going to be. – bissa Aug 22 '12 at 0:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.