• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Server speed issue

S

spinelli

Gawd
Joined
Feb 19, 2004
Messages
527
Ok here is the story,

I just built 4 new (and supposedly faster) servers for my company and I am showing hardly and speed increase over our workstations. I can not seem to pinpoint a bottleneck and wondered if you all could help.

Here are the configs....all run 2k3 server and sql

Workstations

Supermicro X5DAE mobo
4gb pc2700
2x Xeon 3.0Ghz 1mb cache
2x 73gb cheetah 15k U320
1x 37gb cheetah 75k U320
NO RAID


Server

Supermicro X5DAE mobo
4gb pc2700
2x Xeon 3.2 Ghz 2mb cache
2x Adaptec 2200S 128mb cache RAID card
2 RAID 0 Arrays one on each card with 5x 73GB 15K Cheetahs on each

When the programmers here run their tests they only show a 2-3 minute decrease in time over the workstations and that boggles my mind.

I never built any servers before but I would think by the specs the servers should stomp the snot outta the workstations.

Any help/suggestions on what my bottleneck is here would be appreciated
 
Well im no expert, but really how much difference in speed do u expect? Thats only and extra 200 mhz per processor, hardly noticable IMHO
 
I wouldda thought the raid would have been the speed....didnt expect much from the procs alone
 
Originally posted by spinelli
I wouldda thought the raid would have been the speed....didnt expect much from the procs alone
Click to expand...

Since both drive configs are fast, you weren't going to see that much of a difference. SQL is a memory and processor intensive task, and you didn't really do anything to change that.

I can't believe a company actual approved the purchase of such machines, considering the cost. Your company must have too much money (unlikely) or morons making the decisions.

In the future, you should do your research before making an upgrade like that. You'd piss off a lot of people at most companies spending that kind of money with no results.
 
Originally posted by djnes

I can't believe a company actual approved the purchase of such machines, considering the cost. Your company must have too much money (unlikely) or morons making the decisions.
Click to expand...
considering they're using 4 GB of RAM, dual Xeons, and multiple u320 drives in EACH MACHINE, workstation AND server, I'd argue they DO have too much money. I can't imagine what could justify such hardware.

Can't believe he's running suicide RAID on a server...

also, I'm not sure how you're benchmarking these machines, but the speed of your network might well be the bottleneck...
 
Originally posted by jagec
also, I'm not sure how you're benchmarking these machines, but the speed of your network might well be the bottleneck...
Click to expand...

Or if the queries aren't written very well, it may not matter what hardware it's on.
 
the second one would be correct djnes.

My boss ordered the parts....I put em together /shrug

as far as suicide raid...these servers are used just for tests, not production. Each test takes about 4 hours and we run em all day. Speed is more important then the data which is duplicated on other boxes. I did offer to my boss to switch all the new servers to raid 5 instead of 0 and got shot down quickly. so raid 0 it is

Nope network speed is more then adequate...anyway all the data for the tests is sent to one of the arrays prior to starting a test run.

BTW we are a financial firm who basically tries to predict the stock market using different methods....all of em cpu intensive..thus the dualies on workstations as well as servers
 
Originally posted by spinelli
BTW we are a financial firm who basically tries to predict the stock market using different methods....all of em cpu intensive..thus the dualies on workstations as well as servers
Click to expand...

Well as soon as you guys get a foolproof method, could you PM me? My 401k could use a shot of adrenaline.
 
Originally posted by spinelli
the second one would be correct djnes.

My boss ordered the parts....I put em together /shrug

as far as suicide raid...these servers are used just for tests, not production. Each test takes about 4 hours and we run em all day. Speed is more important then the data which is duplicated on other boxes. I did offer to my boss to switch all the new servers to raid 5 instead of 0 and got shot down quickly. so raid 0 it is

Nope network speed is more then adequate...anyway all the data for the tests is sent to one of the arrays prior to starting a test run.

BTW we are a financial firm who basically tries to predict the stock market using different methods....all of em cpu intensive..thus the dualies on workstations as well as servers
Click to expand...
OK, we absolve you of blame if you didn't order the parts;)

so basically it sounds like the tests rely on raw processing power, which is why you'll barely notice a difference between the wkstations and server. In fact, I'm willing to guess you'd probably notice about the same speed if you didn't run RAID.

Consider a Beowulf cluster maybe? I'm not sure if what you're running lends itself to distributed computing.

(if it does, I'm willing to bet that putting together a bunch of much cheaper dually AMDs and running it distibuted, would be faster...don't tell your boss:p)
 
haha...

actually I did try getting him to buy the 64bit athlons....it is like talking to a wall though.

Also tried to get the company to go to win2k3 enterprise so I can use more damn ram.....blocked by the wall.

In our office only M$ products are used so dont think that cluster would fly. ..

Just ordered a 2k3 enterprise trial which I will install on one of the new servers and drop more ram in the beast.......never went over 4gb though, are there any issues I should be aware of?

Thx for the input guys.

Oh and also..I dont know sql from a can of paint so the queries could be written in Phoenecian for all I know. But we have programmers here out the yazoo so I assume they know what they are doing.
 
LOL damn! I went to that site and configured that server with all the best options. hahahah, $57065.00!!
 
the place i work for just spent $20,000 on two servers from DELL... i can't believe they cost that much... one is single xeon, and one is dual xeon, the good one has 3 60 gigs in raid and only 1.5 gb memory.. i think the damn rack was a real huge chunk of the cost...

i need to convince the network guy that we need to save money and buy AMD setups instead of p4... but he says intel is robust.. and he only uses microsoft too.. (he did show some interest in playing around with a red hat server though, so will we do that sometime) .. but i guess he is going with the school of thought that no one ever got fired for using microsoft and intel.. but if he get's linux and amd people might try to blame problems on his choices.
 
To tell you the truth there is not a whole lot of difference between the workstation setup and the server, sans the raid card, this is where you are seeing the extra performance.


~N
 
I did price out a 4 proc server....I was looking at about 40k w/o HD or raid cards so total wouldda been around 46k.

I was afraid to recommend that though to my boss. This was my first week on the job and well....if we spent 46k on one server based on my recommendation and it wasnt fast enough to make jeebus cry....I would be back in the bread line.



BTW I didnt think you could run regular Xeons in a quad setup...thought you had to go to MP...at 5-8k a pop
 
i dont think with that set up you would really see that much speed increase with 200mhz, if you set up the raid you might see a vast improvement
 
You must log in or register to reply here.
Back
Top