RAID1 performance issue
Hi guys,
I have recently built a server for me to practice on at home. I have two Western digital 500GB Caviar Blue hdds running in software RAID1, performance for opening files etc is absolutely fine but when I am uploading files from my client machines to
my server over a gigabit network after around 700mb has been transfered the transfer suddenly halts and stops responding. When I look at my performance section of the task manager on the server, my CPU usage suddenly climbs to 100% as the transfer begins,
and then will drop back to idle when the transfer halts on my client's screen. One thing that I am noticing is that when the transfer is under way, only my first drive of the two is working "hard", the second one is only gently blinking. Surely they
should both be working just as hard seeming as the data is being mirrored?
My OS is Windows Server 2003 Enterprise, running on a dual core Celeron E3300 with 2gb of RAM. Maybe it is my hardware which is holding it back? Maybe the CPU can't cope with such fast data transfers between hard disks?
Your opinions on the matter are most appreciated.
Thanks in advance,
Alec
May 28th, 2010 7:58pm
Do you have anti-virus running on the server, and does it make a difference if you exclude the upload directory from scanning (just as a test offcourse)
Free Windows Admin Tool Kit Click here and download it now
May 29th, 2010 1:21pm
Hi, nope no Anti Virus running on the server as of yet. I have done some more tests, I have a third 500gb drive which is not part of the RAID1 configuration, I can send files to it over the network absolutely fine- no stopping and starting and no high CPU
usage. I am now moving the files over from that hard disk to my RAID1 drives, the entire system is acting borderline unresponsive to user input, the primary RAID1 drive is constantly active and the second drive in the RAID array is blinking not as much, the
same applies to the drive which is sending the data over to the RAID drives. The CPU usage is very very high and it would appear that is the reason why the system is so sluggish in response to my inputs. I am transferring around 20gb worth of data and
have had it running for a good 20 or so minutes now.
Maybe I need to just reconstruct my RAID1 array? Maybe something is malfunctioning to cause such high CPU usage? I am a bit confused to be honest.
Cheers for the reply though, most appreciated.
Alec
May 31st, 2010 7:09pm
Hello,
If the problem occurs after 700MB, my guess is that it's a store and forward buffer issue with the RAID, the disk drives or with some network device, either router, switch and/or LAN card. Maybe the transfer buffer becomes full at that point and
the data transfer starts to crawl, which might appear to become 'frozen'.
Only way to find out is to run some tests. Have you tried to wait it out?
It is funny that CPU increases to 100% when you start to upload. Obviously, performance is degraded when writing to the disk, so the s/w RAID is likely the culprit. FTP transfer is not such a hog so as to take up 100% CPU. Another problem may be is that
it's a software based RAID (which I don't recommend), the s/w raid is sloooowwww compared to hardware RAID.
Try writing a large file (like from CD or USB key NOT from a network location) to the hard drive and see if the same problem occurs. If it does, then your software RAID or the computer's I/O is the likely culprit. If it does not, look into your networ
equipment.
I'm glad this is your home practice server and not a production server. In a production environment, I recommend a GOOD hardware RAID, SAS or SCSI and definitely NOT Celeron CPU.
Miguel
Miguel Fra /
Falcon ITS
Computer & Network Support, Miami, FL
Visit our Knowledgebase Sharepoint Site
Free Windows Admin Tool Kit Click here and download it now
May 31st, 2010 8:46pm
Yes at work I use dual Xeon processors usually, much more powerful machines but since I am just using this at home for practicing I don't need anything really powerful, I am more leniant towards low electricity consumption and heat output since it is in
my house (which isn't like a fancy server room with air conditioning).
I have tried sending files over the network to my other hard disk which isn't in the RAID and they transfered flawlessly, no stopping and starting and no high CPU usage. I shall try sending a file over from a USB pen in a moment and report back how that
goes.
Cheers,
Alec
May 31st, 2010 9:00pm
Ok I have just completed a transfer from a USB stick to the RAID hard drives, the CPU usage was steady maxing out between 7 and 10 percent usage, however as soon as the transfer had presumably completed, the system became unresponsive (could not click on
windows and one window had the image of a since minimized window still displayed in parts of it), and the secondary hard disk in the RAID was still flashing, around 20 seconds later the system jumped back into life but task manager reported no high CPU usage.
RAM isn't an issue as it is only using 330MB at the moment, leaving approximately 1670MB available for use.
I have also just done another test transferring a 1gig file from one drive to the RAID and the CPU usage didn't appear to go high, however task manager froze during the transfer. Something is definitely hogging performance.
It feels like the software RAID controller isn't coping very well in managing both disks- one disk always seems to be active a lot more than the other; the primary drive seems to be very busy while the secondary drive is not so busy, and maybe that is bottlenecking
performance as the system has to catch up with it's self.
Maybe I should invest in a hardware RAID controller rather than use the software RAID controller? Maybe with my processor the software RAID controller simply cannot work to its fullest and thus is bottlenecking performance.
Reading from the RAID is no problem at all, however.
Thanks,
Alec
Free Windows Admin Tool Kit Click here and download it now
June 1st, 2010 12:41am
What if you break the RAID on the two drives, and test to the individual drives
if this is fast, then it is the software RAID which may be limiting
if this is slow, then it may be the physical drives themselves.
rgds,
Edwin.
June 1st, 2010 3:32pm
Hey guys jsut to let you know after it sitting still for half a day it suddenly "clicked into place", now my CPU usage is very low and transfer speeds are fast. I no longer have the issue of a file transferring x amount of data and then bottlenecking
itself.
Really not sure what the cause was, but hey- it works now. Cheers guys for your feedback!
Kind Regards,
Alec
Free Windows Admin Tool Kit Click here and download it now
June 8th, 2010 6:25pm