SQL Server Performance Forum – Threads Archive

optimal configuration on HP DL 380

Hello, as a part of internation company we have such policy that oblige us to use Raid 5 + spare on all servers where you need to protect data.
Confiration: 4-6 HDD SCSI 10-15 kRPM, 2-4 GB RAM, smartarray 5xxx 64-128(256) k Cache with battery back up,
Windows 2000 server standard, MS SQL 2000 standard, windows write chache is disabled (it is done by the driver).
But it seems to me that raid5 on that server is not very fast comparing to ordinary IDE disks without array, especially for writes.
our information system (OLTP) has both read/writes, but the problem is more in writes This year we will buy new server (DL 380 G4) What could help me as for hardware?
-is is disk array with more chache (HP provides also model with 256 k Cache) ?
-is it creating two arrays (but I have to buy extension , because DL 380 allows only 6 disk and in this case I would need at least 8.
-more memory (4 GB)? – but it has no sence for writes, isn’t it ?
-or another recomendation ? Thanks for you advice Martin

for a maximum of 6 disk including spare, i would suggest 3 disks raid5 for data, 2 disks raid1 for log, 1 disk spare.
without spare, try 4 disk raid5 data, 2 disk log.
since you do have battery backup, enable the write cache on the controller (not the OS)

Have you looked at the counters to verify the problem is in the array though? You’re using words like "seems", which bothers me somewhat. [<img src=’/community/emoticons/emotion-1.gif’ alt=’

‘ />] What do the disk read and write queue lengths look like? Look at both logical and physical disk counters. Tell us what you’re seeing on the server. How are you comparing it to ordinary IDE disks without array? How do you know the problem is more in writes? What does your memory consumption, processor, etc look like? Are you having latency on the network side when trying to write data? Are you trying to write large amounts of data at a time, or is the slowtime visible in small, individual writes? Also, I agree with Joe on the write cache. VERIFY that the batter backup works though before you do this. You also need to verify that the shutdown process works properly. Test it thoroughly before you do this. Power outages are a really, really bad time to find out these items didn’t work and you just corrupted all your databases. MeanOldDBA [email protected] When life gives you a lemon, fire the DBA.

Dear Martin, If speed is what you need you can… Buy a DL380 G4 (2x) 3.6 Mhz (if you have money buy special CPU’s Xeon with Extra Level3 cache) If you are rich a ML580 4 CPU’s…
Go for the 4 disks with 15000K 72 Gb SCSI 3.2 and 2 disks with 15000K 36 Gb SCSI 3.2
8 Gb RAM
SmartArray 6xxx
Make array RAID 1 use 2 disks 36 Gb [For OS & PageFile]
Make array RAID 1 use 2 disks 72 Gb [For DataFile & FileBackup] [Format NTFS 8kb clusters]
Make array RAID 1 use 2 disks 72 Gb [For LogFile] or… Idem but only 4 disks (if you have tons of money)
Make array RAID 1 use 2 disks 36 Gb [For OS]
Make array RAID 1 use 2 disks 72 Gb [For PageFile OS & SQL CE Virtual Directory in case of replication over IIS with Mobile Devices with SQLCE]
2 Fiber [example Emulex] cards connected to a SAN [1 used for Full Backup Server & 1 for access to example 3 LUN’s [1 for DataFile… on mid range disks on SAN, 1 for LogFile on high end disks on SAN and 1 for FileBackup on low end disks on SAN]
As the SAN system has internal RAID5 config… data is secure and normally if your company can afford a SAN [HP or HITACHI] it’s mirrored to a other site by hardware or software [example VERITAS] Use the Compaq/HP Insight Manager, nice free tool, to send you traps… For SQL use Ent. Edition settings… use AWE… min. 2Gb – max. 5Gb
If LinkedServers are used make them run out of process [MemToLeave limits]. You can also fine-tune the read/write settings of the RAID controller Greetingzzz, KITSS

]]>