If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
|
Thread Tools | Display Modes |
#11
|
|||
|
|||
Smart Array 642 + HP DL380
"Mikael Antonsen" wrote But still, 5x Write speed with that module. I just dont think ppl that bought the 642 without would think.."ohh thats a okai speed". I know the acceleration will help in some ways yes, but still. There really must be something wrong here. Actually, most ppl who buy a 642 never bother testing disk performance, so would never notice either way. For an objective performance, try IOMETER http://www.iometer.org/ That's the 'standard' for testing disk performance. I have a customer right now (Danish, ironically) who's complaining of a similar problem. We have just put a LOT of resource into proving that the performance is actually about right for the system, and I have to ring him tomorrow to explain to him why he's a complete ****wit... |
#12
|
|||
|
|||
Smart Array 642 + HP DL380
"NuTCrAcKeR" wrote call HP and complain. - LC Please do, and let us know how you get on - I'm quite curious to see how long it takes you to get a decent answer from the DK Proliant team... I'd almost go as far as to putting money on it that they offer to sell you a new SA642, withoug asking you for an IOmeter report, or probably even an ADU report. |
#13
|
|||
|
|||
Smart Array 642 + HP DL380
Actually ... another tool that is VERY useful to testing/stressing disk
systems is JetStress, from our buddies at MS. We use it all the time to benchmark new cluster implimentations, and confirm that our configurations will deliver what the customer's target IOPs are. They dont want to know exactly how many IOPs the system is capable of, just weather or not the system can sustain thier usages models. We went for broke once and cranked up all the manual settings to rediculous values. We exceeded thier requirements by a mere 1700+ %, and our storage management guys called us and wanted to know what we were doing because we were "melting the symmetrix" .... heheheh - LC "Jez T" wrote in message ... "Mikael Antonsen" wrote But still, 5x Write speed with that module. I just dont think ppl that bought the 642 without would think.."ohh thats a okai speed". I know the acceleration will help in some ways yes, but still. There really must be something wrong here. Actually, most ppl who buy a 642 never bother testing disk performance, so would never notice either way. For an objective performance, try IOMETER http://www.iometer.org/ That's the 'standard' for testing disk performance. I have a customer right now (Danish, ironically) who's complaining of a similar problem. We have just put a LOT of resource into proving that the performance is actually about right for the system, and I have to ring him tomorrow to explain to him why he's a complete ****wit... |
#14
|
|||
|
|||
Smart Array 642 + HP DL380
"Jez T" wrote in message ... Well anyways cant be right that even if not all that is done, the write speed is 10MB/s, that is just DAMN slow, and even 12 HDD's? I tried with the same bay on a Smart Array 5300 controller, and then i got like 50-70MB/s in write, that is "okai", SA5300 has a write cache. but still not what i would call good, but its 100MB/s read. Do you think that a Battery module would help THAT much? Oh yes!!! See ? |
#15
|
|||
|
|||
Smart Array 642 + HP DL380
Jez T skrev:
"Mikael Antonsen" wrote But still, 5x Write speed with that module. I just dont think ppl that bought the 642 without would think.."ohh thats a okai speed". I know the acceleration will help in some ways yes, but still. There really must be something wrong here. Actually, most ppl who buy a 642 never bother testing disk performance, so would never notice either way. For an objective performance, try IOMETER http://www.iometer.org/ That's the 'standard' for testing disk performance. what cluster size do you normally test with in IOMETER ? And yes it seems like you guys really mean that 10MB/s is okai for a raidcontroller without Battery stuff. I really dont think they would make a raidcontroller that would perform so slow? |
#16
|
|||
|
|||
Smart Array 642 + HP DL380
"Mikael Antonsen" wrote in message . .. Jez T skrev: "Mikael Antonsen" wrote But still, 5x Write speed with that module. I just dont think ppl that bought the 642 without would think.."ohh thats a okai speed". I know the acceleration will help in some ways yes, but still. There really must be something wrong here. Actually, most ppl who buy a 642 never bother testing disk performance, so would never notice either way. For an objective performance, try IOMETER http://www.iometer.org/ That's the 'standard' for testing disk performance. what cluster size do you normally test with in IOMETER ? And yes it seems like you guys really mean that 10MB/s is okai for a raidcontroller without Battery stuff. I really dont think they would make a raidcontroller that would perform so slow? there are many, many factors that determine array controller performance. None of us are saying that 10MB/s is "okai", just that it is not suprising given your setup, and lack of other important details in your post. Determining your cluster size can be a tricky thing. What it comes down to is what type of application(s) you are running on your server, and what is optimal for those apps. If its MSSQL ... 64K. If its Exchange, 4K. If its a file/print server, it then depneds on your typical data sets ... hundreds of large files, hundreds of thousands of small files ???? There is no one-size-fits-all that will get the best performance in all cases. Beware that you can adversely affect performance, AND waste disk space if you do not use the optimal cluster size. Trial and error is usually a fair approach to figuring this out. Also, with windows (i dont do opensource), the cluster size is determined at the time you format your partions. The only volume you do not get this choice with is your boot partition, as that is created for you during setup. Also take into consideration the Stripe Size (how much data is read-from/written-to the disk at a time. a 16K stripe will out perform a 4K stripe in most cases. I usually use the highest value available (128K or 256K), depending on the controller. How did you do your testing to determine what your performance was ? - LC |
#17
|
|||
|
|||
Smart Array 642 + HP DL380
NuTCrAcKeR skrev:
"Mikael Antonsen" wrote in message . .. Jez T skrev: "Mikael Antonsen" wrote But still, 5x Write speed with that module. I just dont think ppl that bought the 642 without would think.."ohh thats a okai speed". I know the acceleration will help in some ways yes, but still. There really must be something wrong here. Actually, most ppl who buy a 642 never bother testing disk performance, so would never notice either way. For an objective performance, try IOMETER http://www.iometer.org/ That's the 'standard' for testing disk performance. what cluster size do you normally test with in IOMETER ? And yes it seems like you guys really mean that 10MB/s is okai for a raidcontroller without Battery stuff. I really dont think they would make a raidcontroller that would perform so slow? there are many, many factors that determine array controller performance. None of us are saying that 10MB/s is "okai", just that it is not suprising given your setup, and lack of other important details in your post. Determining your cluster size can be a tricky thing. What it comes down to is what type of application(s) you are running on your server, and what is optimal for those apps. If its MSSQL ... 64K. If its Exchange, 4K. If its a file/print server, it then depneds on your typical data sets ... hundreds of large files, hundreds of thousands of small files ???? There is no one-size-fits-all that will get the best performance in all cases. Beware that you can adversely affect performance, AND waste disk space if you do not use the optimal cluster size. Trial and error is usually a fair approach to figuring this out. Also, with windows (i dont do opensource), the cluster size is determined at the time you format your partions. The only volume you do not get this choice with is your boot partition, as that is created for you during setup. Also take into consideration the Stripe Size (how much data is read-from/written-to the disk at a time. a 16K stripe will out perform a 4K stripe in most cases. I usually use the highest value available (128K or 256K), depending on the controller. How did you do your testing to determine what your performance was ? - LC I just ran IOMETER on the array. Both at formatted drive and without, no really effect. Well maybe the cluster size can change the speed, but are we talking from 10MB/s here to like 50MB/s or? or is there really a problem with the controller in the first place. As i said i tried with a Smart Array 5300 controller, all i did was swap the bay cable to the other controller. And then iometer with a "10000" sector its like 100MB/s in write, and same in write. I would love to do some more tests, and uses other programs. I talked to a guy who have had same problem with 642, but only when it was in another machine than compaq. But the 642 is in DL380 G3. -Mikael |
|
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Recover array after replacing failed Smart Array 3200 controller on Proliant 1600 | [email protected] | Compaq Servers | 7 | November 30th 06 07:00 PM |
Compaq Proliant 5500R- Smart Array 3200 Controller to slow ? | Sebastian Gödecke | Compaq Servers | 11 | September 19th 06 06:31 PM |
Compaq Smart Array 532 Controller - Raid 1 Query | Richard Denny | Compaq Servers | 3 | November 24th 04 01:15 AM |
RAID Array "Off Line" on P4C800-E Deluxe | macleme | Asus Motherboards | 4 | September 1st 04 07:22 PM |
Critical errors? ATA Error Count | Al Bogner | Storage (alternative) | 0 | June 13th 04 12:14 PM |