firstnetsupport
asked on
Troubleshooting/Benchmarking SBS 2011
Hi I have been having some issues with installing and configuring SBS 2011 on a Fujitsu TX100 S3p server using RAID 1 with 2 x 500GB Harddrives.
It turns out there is something wrong with the drives and Fujitsu support are replacing those, but i have 2 x spare drives 1TB each from another Fuji server and i wanted to ask what the best benchmark utility would be to test the system with?
seeing as i have had problems with the discs that would be a major part of the test suite for me.
The 1TB Drives are working ok but i just want some clarity on how the system should be performing compared with other rollouts of similar spec.
It turns out there is something wrong with the drives and Fujitsu support are replacing those, but i have 2 x spare drives 1TB each from another Fuji server and i wanted to ask what the best benchmark utility would be to test the system with?
seeing as i have had problems with the discs that would be a major part of the test suite for me.
The 1TB Drives are working ok but i just want some clarity on how the system should be performing compared with other rollouts of similar spec.
Based on my experience RAID 1 and SATA are anathema. Meaning they don't belong in a server anymore especially an SBS 2011 Standard server.
Depending on the number of users 16GB-24GB of RAM is a good start.
4 SAS spindles would be RAID 5 via hardware RAID, 1GB flash or battery backed cache.
6-8 SAS spindles would be RAID 6 via hardware RAID, 1GB flash or battery backed cache.
One of our blog posts explaining why we don't do SATA in servers with links to Microsoft's statement that SATA does not belong in servers (finally!): http://bit.ly/1dnoqs3 (MPECS Inc. Blog: Repeat After Me: SATA Does Not Belong In Servers Part Deux)
Depending on the number of users 16GB-24GB of RAM is a good start.
4 SAS spindles would be RAID 5 via hardware RAID, 1GB flash or battery backed cache.
6-8 SAS spindles would be RAID 6 via hardware RAID, 1GB flash or battery backed cache.
One of our blog posts explaining why we don't do SATA in servers with links to Microsoft's statement that SATA does not belong in servers (finally!): http://bit.ly/1dnoqs3 (MPECS Inc. Blog: Repeat After Me: SATA Does Not Belong In Servers Part Deux)
ASKER
I agree with your comments, but my question relates to testing performance of the system as it stands can you recommend how to best test the system?
Start performance testing with Task Manager -> Performance then Resource Monitor. That will tell you a lot. If you want more info you can use Performance Monitor also free from MS.
As far as third party suites, I have never used one because, imo, they aren't necessary. Phillip and I have given you the criteria, and the tools above will show you what is going on with your server.
As far as third party suites, I have never used one because, imo, they aren't necessary. Phillip and I have given you the criteria, and the tools above will show you what is going on with your server.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
I agree to Philip, there is no single tool which can give you performance bench-marking for entire system.
There are tools which comes as collection of various other simulation tools which may help you. such as dacapo , or Login VSI . I had personally used both of them and found quite effective. Since Login VSI is not a free tool however you may download it for testing purpose.
Dacapo is a free tool work on Java and it provides you information based on the response time.
IOMeter and SQL IO stress test , jetstress testing tool and many more are used for bench-marking the disk IO response time which is quite crucial in exchange , sql and many other activity.
There are tools which comes as collection of various other simulation tools which may help you. such as dacapo , or Login VSI . I had personally used both of them and found quite effective. Since Login VSI is not a free tool however you may download it for testing purpose.
Dacapo is a free tool work on Java and it provides you information based on the response time.
IOMeter and SQL IO stress test , jetstress testing tool and many more are used for bench-marking the disk IO response time which is quite crucial in exchange , sql and many other activity.
Not disagreeing with any of the suggestions or tools, but pointing out that we are dealing here with Small Business Server with a MAX of 75 users. Stress testing Exchange for what is more likely to be 5-50 users won't tell us very much except maybe we might have spent more on the storage sub-system, but how many are going to replace controller and drives to achieve a .001 second speed increase in the delivery of email?
Yes, SATA drives are, at best, sluggish from the console and if loading large files over the network or hitting it with 100s of SQL transactions per minute would not be satisfactory. But for the average small business network it will take about 1 second longer to load a word document over the LAN than with SAS or SSD drives. But if you have hundreds of SQL transactions per minute one should have, and probably would have, researched the storage sub system ahead of time.
I think the whole exercise is futile. If things are slow.. first make sure you have RAM, then limit the amount used by Exchange and SQL (search or start a new thread if you don't know how), then use the built in MS tools, then replace the drives or the controller or both.
Yes, SATA drives are, at best, sluggish from the console and if loading large files over the network or hitting it with 100s of SQL transactions per minute would not be satisfactory. But for the average small business network it will take about 1 second longer to load a word document over the LAN than with SAS or SSD drives. But if you have hundreds of SQL transactions per minute one should have, and probably would have, researched the storage sub system ahead of time.
I think the whole exercise is futile. If things are slow.. first make sure you have RAM, then limit the amount used by Exchange and SQL (search or start a new thread if you don't know how), then use the built in MS tools, then replace the drives or the controller or both.
SBS is most dependent on RAM, then Disk I/o, then CPU. You should have at least 16 GB RAM and 24 is better, 32 is best, since that is all the this version of Windows will address. For disks, SSD is fastest, 15K RPM SAS is next, 10K SAS is next and SATA is a poor 4th. For RAID.. raid on the controller is better than software raid, a discrete controller is better than the pseudo controllers built into some motherboards.
500 and 1 TB imply SATA, which will give you the slowest performance.