Windows Server 2003 – SAN | SQL Server Performance Forums

SQL Server Performance Forum – Threads Archive

Windows Server 2003 – SAN

Future Environment:
2 node IBM x445 4 CPU 2.0 GHZ 4GB Memory each server.
1 Controller IBM FastT700 with 2 EXP700.
26 FC-HDD 74GB 15K in RAID 1 + 2 Hot Spare
Tape and so on… I need to configure this cluster and I have 3 questions: 1.Do I need to configure "Zone" and LUN MAsking" even if I have only 2 servers or this only apply when there are more than 2 Servers on the same storage device? 2.I have plenty of space so is it a good thing to put SQL Server DB on one array, SQL Server Logs on another array and SQL Server Backups on a third array? 3.It is sufficient to have only 2 Hot Spare or do you think I need more? Thank you for your time.[8D]
Kind regards.
Franco
2. it is still important to have logs on a separate set of disks, 2 disks in RAID 1 is sufficient unless you must keep a very large amount of logs.
one big reason for this is during checkpoints, SQL Server will flood the disk queue for the data disks with writes such that it may take several seconds to clear up.
if your log is on the same disk set, what should normally be a <1ms write to the log becomes a >1sec write, meaning your transaction rate takes a nose dive until the checkpoint completes.
on the backup partition,
the theory is that a sequential read from one disk set writing to another disk set runs at the full combined sequential throughout of the lesser disk set, and presumably a read followed by a write to then same disk set cannot achieve the full sequential throughput.
i have actually seen very good backup transfer rates with just a single disk set,
test this out on your system, with 26 disks, you ought to be able to get several hundred MB/sec, of course, you will need parallel FC adapters, and probably multiple backup files 3. i really think the incidence of disk failures is sufficiently low that 1 or 2 hot spares is adequate.
I am assuming that your 26 disk in RAID 1 is actually RAID 10 or whatever IBM calls it, such that you can actually have multiple disk failures so long as they are not adjacent disks?
a question to ask is where is the server located with respect to your personnel, if its in your office building, then on a disk failure, you call ibm support, they fedex a new disk, you replace it the next day.
if the server is a remote location, then you might want a few extra hot spares so that you don’t need to replace failed drives right away,
final issue is end of life, most high-end drives are rated for a 5-year life span. i would be more concerned about higher than normal failure rate in the first week (infant mortality), and after 4 years of continuous use.

Thank you very much for your suggestions.
I have another question related arrays:
Is better to have multiple arrays or multiple partitions within the arrays?
Please consider the following: The idea is to create separate arrays for each drive letter.
In fact I have to plan these resources:
Arrays:
1 SQL Server databases
2 SQL Server transaction logs
3 SQL Server backups
4 SQL Server tempdb
5 Oracle database
6 Oracle redo logs
7 Oracle archived redo logs
8 Oracle Exports and Backup (RMAN)
9 File system
10 Ca-BrightStore arcServe Backup for Windows v.11
11 Quorum disk (for the cluster) Every Array would have a different letter.
I have plan to also have Hot Spare drive of 73GB each. Do you think this is a good design?
Please advise.
Kind regards. Franco
]]>