I need relief for my overworked drives?!?!
I have a Gateway desktop that by necessity, I had to put in the role of a server.
I got a Promise RAID controller and maxed out the cache memory on it.
I now have four 250GB SATA drives on that controller (I'm continually impressed the power supply can handle the load). My problem is that when I monitor the %idle time in perfmon it’s usually at or around 0%. That’s 0% idle time, which means that they are ALWAYS busy. Does anyone have any suggestions on how I can maximize the efficiency of the raid controller, or anything that I can do to speed up access in general? I would love to be able to disable Symantec Corp Antivirus, but that’s a big no-no around here.
It is running Windows 2000 server. The RAID is running at RAID level 0 already, so I’m not sure how much I can do.
Thanks!
I got a Promise RAID controller and maxed out the cache memory on it.
I now have four 250GB SATA drives on that controller (I'm continually impressed the power supply can handle the load). My problem is that when I monitor the %idle time in perfmon it’s usually at or around 0%. That’s 0% idle time, which means that they are ALWAYS busy. Does anyone have any suggestions on how I can maximize the efficiency of the raid controller, or anything that I can do to speed up access in general? I would love to be able to disable Symantec Corp Antivirus, but that’s a big no-no around here.
It is running Windows 2000 server. The RAID is running at RAID level 0 already, so I’m not sure how much I can do.
Thanks!
ASKER
well... it's not as dire as it sounds. The 1TB is not all the way full. It was originally set to RAID 0 because it was part of a redundant cluster. The software that we were using (legato replistor) really sucked, so we had to remove it because it was making things worse. We are testing our new cluster with Dell PowerEdge 2600's with 3TB (ultra SCSI 3 internal and external drives) of space. I just need a band-aid until we are able to migrate to our new cluster and go live with it.
Our cluster is including 3 identical servers all with 3TB of disk space running RAID0 with our new clustering software NSI's GeoCluster. The reason that we are running RAID 0 is because there is (was) already redundancy within the configuration.
Jared
Our cluster is including 3 identical servers all with 3TB of disk space running RAID0 with our new clustering software NSI's GeoCluster. The reason that we are running RAID 0 is because there is (was) already redundancy within the configuration.
Jared
RAID 0 with 4 drives?!?!?! Very risky.
Being it is running 2000 Server, You should be able to configure a RAID 5 set in Windows.
http://www.microsoft.com/windows2000/en/advanced/help/default.asp?url=/windows2000/en/advanced/help/sag_diskconcepts_18.htm
http://www.microsoft.com/windows2000/en/advanced/help/dm_raid5.htm
Being it is running 2000 Server, You should be able to configure a RAID 5 set in Windows.
http://www.microsoft.com/windows2000/en/advanced/help/default.asp?url=/windows2000/en/advanced/help/sag_diskconcepts_18.htm
http://www.microsoft.com/windows2000/en/advanced/help/dm_raid5.htm
ASKER
The promise card is a RAID card, so it is doing hardware based raid. Putting it to HARDWARE based RAID 5 would only make things worse let alone software based RAID 5.
ASKER
I know the way it's currently set up is less than ideal... I just need to get it along for another week or so.
How much memory do you have on the desktop computer?
ASKER
The machine has 512 MB of which, task manager says there is 180 available.
Explain to me how you have any redundancy with a 3 node cluster sharing 3TB of RAID 0 storage.. Sure, you have server redundancy, but no data redundancy... Or do you have 3TB of storage on EACH server? That's kind of overkill.. You should have looked at a SAN to share the 3TB and built a RAID10 array if performance was that big of an issue...
And why, oh why, would ANYONE put a gateway desktop machine into a cluster?? For < $20k I put up a 1.2 TB IBM Cluster running Windows 2003 with ALL IBM parts. x345's, EXP400, ServeRAID 6M, U320 drives...
And why, oh why, would ANYONE put a gateway desktop machine into a cluster?? For < $20k I put up a 1.2 TB IBM Cluster running Windows 2003 with ALL IBM parts. x345's, EXP400, ServeRAID 6M, U320 drives...
512MB don't sound like enough.
What does the Commit Charge show?
What does the Commit Charge show?
ASKER
Total: 308500
Limit: 1275448
Peak: 696788
Limit: 1275448
Peak: 696788
Is the OS on a different drive or is it in the RAID array.
It sounds as if it is paging all the time.
It sounds as if it is paging all the time.
ASKER
it's on a differnet drive (ATA drive on the MB's IDE Controller)
Add counters to your monitor for Memory Pages/sec, Page/reads and writes and see how much activity you have there.
I would boost your memory up to 1GB at least.
I would boost your memory up to 1GB at least.
Also add counters for Paging files.
ASKER
exx1976:
I hear you! Trust me.. it wasn't supposed to come to this! : )
Your right... there are three TB of space on EACH server. One server is the primary that will remain online at all times. The other two are secondary servers in the cluster. One will be OFFLINE at all times. When the offline system is done replicating with the online master, then that one goes offline and the one that WAS offline then comes online.
Because of the way that the power structure (politically) and the network is layed out around here, we have been burned by other peoples mistakes, so this is the environment that we have come up with to combat that. If a virus breaks out, then we have a COMPLETE server with data that will most likely only be one day old that we can bring online and have up and running while everyone else is down taking care of the virus.
It's also a disaster recovery plan, because one of the servers is in a completely differnet building. If this one burns down, then we have a WHOLE server waiting to come online with (almost) full data.
This type of setup can not be done with a SAN unless you are talking about hundres of thousands of dollars. My boss had a very specific goal in mind, and this is what we came up with.
It's just a matter of getting the bugs ironed out of the new setup and getting user data moved over to the much more powerful and appropriate server/storage systems.
Jared
I hear you! Trust me.. it wasn't supposed to come to this! : )
Your right... there are three TB of space on EACH server. One server is the primary that will remain online at all times. The other two are secondary servers in the cluster. One will be OFFLINE at all times. When the offline system is done replicating with the online master, then that one goes offline and the one that WAS offline then comes online.
Because of the way that the power structure (politically) and the network is layed out around here, we have been burned by other peoples mistakes, so this is the environment that we have come up with to combat that. If a virus breaks out, then we have a COMPLETE server with data that will most likely only be one day old that we can bring online and have up and running while everyone else is down taking care of the virus.
It's also a disaster recovery plan, because one of the servers is in a completely differnet building. If this one burns down, then we have a WHOLE server waiting to come online with (almost) full data.
This type of setup can not be done with a SAN unless you are talking about hundres of thousands of dollars. My boss had a very specific goal in mind, and this is what we came up with.
It's just a matter of getting the bugs ironed out of the new setup and getting user data moved over to the much more powerful and appropriate server/storage systems.
Jared
ASKER
crazijoe:
Page read/sec jumps from around 0 to about 30 mostly.. it has spiked to about 110
page writes/sec hardly moves and is usually 0
pages/sec is usually (on average) under 100.
pageing file usage is under 5%
Page read/sec jumps from around 0 to about 30 mostly.. it has spiked to about 110
page writes/sec hardly moves and is usually 0
pages/sec is usually (on average) under 100.
pageing file usage is under 5%
I would watch your Commit Charge. When it gets around 522000 it will start paging to the hard drive. It showed it peaked at 696788 so it has been paging. But I couldn't guarranty that's where all your HD activity is coming from. Unless alot of people are hitting that machine all the time. Is this just a file server, terminal server of the cluster?
ASKER
This at the current time is just a file server. The OS drive really isn't doing much. It's the D: drive that's getting hammered (the RAID array with all the user folders).
As soon as we get some 30 amp plugs installed and get our new ups's installed, this machine is going to get migrated to MUCH more powerful servers. I've get my fingers crossed for next week.
As soon as we get some 30 amp plugs installed and get our new ups's installed, this machine is going to get migrated to MUCH more powerful servers. I've get my fingers crossed for next week.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
yea... as I suspected. I was just hoping someone would have some secret tweak I could try to make things work a little bit better...
Thanks for your time.
Thanks for your time.
My first suggestion is to buy a different machine (SERVER comes to mind), but, if that's not possible...
I'd get a nice Adaptec hardware RAID card and some 15K SCSI drives. Higher spindle speed, higher transfer rate, larger disk cache.. You should fare quite well with that.
Oh, and do yourself a favor.. Switch to hardware RAID 5 at your earliest convenience.. Otherwise, I hope you have REAL good backups and a LOT of time to restore.... You're just asking for trouble with that config. If performance is THAT big of an issue that you're willing to risk losing it all, it sounds like the compay you built this thing for is doing well enough (they're very busy, no?) that they can afford to buy a real server and put at least 0+1 on it, if not RAID 10...