March 8, 2005 at 3:34 pm
I have set up a new Dell Poweredge 2650 with Dual Xeon 3.2 GHz Procs, 1GB Ram, RAID 1 for OS and Log files and RAID 5 for MDF file and NIC Teaming. Our current server -uh desktop used as a server- is a single 3GHz proc with 1GB Ram, 1 HD and about 20 users and a two websites connecting the database all day long. I have been testing the new server in anticipation of faster responses to our VB app that our employees use. When I run the same queries on both servers the new server is not always faster but is actually slower sometimes and it doesn't have more then one application accessing the DB.
The database is current to the database on the production server and I have performed general maintaince on the new server (defrag etc). Is this just not that much of an improvement in resources to have way more blazing speeds then the other "server"? Are there any utlities that I can run to test the performance?? Thanks in advanced for any reponses!
March 8, 2005 at 4:14 pm
Since the desktop had only one drive, there could be no RAID, so the server RAID configuration will definitely slow performance.
Did the desktop have an IDE Drive ? If so, be aware that IDE drives perform write caching which will speed up performance but does not guarantee recovering from crashes. SCSI drives typically do not have write caching but do guarantee recovering from crashes.
Regarding utilities, see "SQLIO Disk Subsystem Benchmark Tool" at http://www.microsoft.com/downloads/details.aspx?FamilyID=9a8b005b-84e4-4f24-8d65-cb53442d9e19&DisplayLang=en
SQL = Scarcely Qualifies as a Language
March 8, 2005 at 4:15 pm
First I can guarantee the Server will outperform the desktop overall. But it could be factors of the cahce being populated already, distance data travels across network, network bandwidth. Try performing the queries on each box locally and see what the performance is like. Also check the execution plan, can change for reasons around hardware that are sometimes unexpected and just need a little tweaking.
March 9, 2005 at 2:37 am
Make sure that you have tempdb sized correctly and that SQL is using the memory. Also ensure that all database options are the same, to avoid file autogrowth and fragmentation.
March 9, 2005 at 3:42 am
try gathering server, io and client statistics in query analyzer on both the server and desktop. This will show that the production server generally will be faster than the desktop.
Also check the network interface on the server. Is it a teaming / failover card? We had some problems with this configuration. Changing the card to only teaming or only failover helped.
hth
JP
March 9, 2005 at 6:44 am
Definitely try Antares' suggestion to run the querries locally on both machines and compare performance. I had a similar problem and it took me a while to figure it out. I was sure it was the network but all setups and diagnostics looked correct.
I eventually found that the switch I was connecting to had a low blocking speed. Essentially, even though I had a 100mb FULL DUPLEX connection without any errors, the switche's internal bus was overloaded and latency was being experienced.
Network issues can really affect performance...
March 9, 2005 at 10:44 am
The previous post about the SQL cache is something not to ignore. Fix any network problems first.
After you have eliminated the network as a possibility, try running your tests again, but take the time as your benchmark the "second" time you run the app, after the data has been cached. The first test will be disk I/O bound. The second test will more closely reflect the actual real-world performance.
So long, and thanks for all the fish,
Russell Shilling, MCDBA, MCSA 2K3, MCSE 2K3
March 9, 2005 at 3:14 pm
Thanks everyone for the replies! Now I've got some more great reading material. I have been tooling around with this all day long and will continue to explore the other options mentioned here. I have however greatly improved the network performance by enableing Full Duplex for my Nic Team. They were set at Half Duplex and when I accessed our VB app located on the server it would take way longer than the desktop. Now it seems up to speed. These are Broadcom gigabit nics running broadcom's control suite. They are set for Smart Load Balancing with Failover -if anyone was wondering! Thanks again!
Viewing 8 posts - 1 through 7 (of 7 total)
You must be logged in to reply to this topic. Login to reply