You random reads are in the 50'000 range and writes almost 20'000, so I still think you have a good SAN
If that's enough for that many clients, hmm, that's beyond me... You might want to reach out to Brian Mason (he is on this list), he is one of the few I know with really large sites. They are running SQL locally though. / Johan From: [email protected] [mailto:[email protected]] On Behalf Of Roland Janus Sent: den 23 oktober 2013 01:02 To: [email protected] Subject: RE: [mssms] SQLIO tests values? That is actually a surprise. It isn't a really good SAN, it's never over 50'000/20'000? Dividing the database data then? Oh, before I forget: This would be for two primaries, each managing up-to 100'000 clients... Remote SQL (due MS support) For a CAS with 200'000 clients? Shouldn't that mean (much) higher values then? It is really painful that there isn't anything officially. Yeah, please add some notes :) From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Johan Arwidmark Sent: Mittwoch, 23. Oktober 2013 00:28 To: [email protected]<mailto:[email protected]> Subject: RE: [mssms] SQLIO tests values? Actually no, the small ones (read 8 kb) yes, are a bit slow... But the other block sizes 64 kb and 512 kb are good. And your write operations are very good. SQL uses a lot of 64 kb blocks in its read/write operations. I guess I need to add a few notes in the post on how to interpret the data returned... Your SAN looks good to me, I have seen a lot worse :) / Johan From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Roland Janus Sent: den 22 oktober 2013 23:24 To: [email protected]<mailto:[email protected]> Subject: RE: [mssms] SQLIO tests values? I'm pretty sure you consider those numbers to be bad then as I do. 50GB file on a VM (ESX 5) While most of them are close to the once mentioned 5000 some are way below and not close to what you expect from a good SAN. Maybe it is more related to the VM though. .\SQLIO.EXE -s120 -kR -fsequential -b8 -t4 -o2 -LS -BN G:\Benchmarkfile.dat IOs/sec: 13634.10 MBs/sec: 106.51 .\SQLIO.EXE -s120 -kR -fsequential -b64 -t4 -o2 -LS -BN G:\Benchmarkfile.dat IOs/sec: 6453.21 MBs/sec: 403.32 .\SQLIO.EXE -s120 -kR -fsequential -b512 -t4 -o2 -LS -BN G:\Benchmarkfile.dat IOs/sec: 1888.51 MBs/sec: 944.25 # Read Random, various blocksizes (8, 64 and 512 kb) .\SQLIO.EXE -s120 -kR -frandom -b8 -t4 -o16 -LS -BN G:\Benchmarkfile.dat IOs/sec: 44944.89 MBs/sec: 351.13 .\SQLIO.EXE -s120 -kR -frandom -b64 -t4 -o16 -LS -BN G:\Benchmarkfile.dat IOs/sec: 19339.86 MBs/sec: 1208.74 .\SQLIO.EXE -s120 -kR -frandom -b512 -t4 -o16 -LS -BN G:\Benchmarkfile.dat IOs/sec: 2868.92 MBs/sec: 1434.46 # Write Random, various blocksizes (8, 64 and 512 kb) .\SQLIO.EXE -s120 -kW -frandom -b8 -t4 -o16 -LS -BN G:\Benchmarkfile.dat IOs/sec: 19250.84 MBs/sec: 150.39 .\SQLIO.EXE -s120 -kW -frandom -b64 -t4 -o16 -LS -BN G:\Benchmarkfile.dat IOs/sec: 7891.82 MBs/sec: 493.23 .\SQLIO.EXE -s120 -kW -frandom -b512 -t4 -o2 -LS -BN G:\Benchmarkfile.dat IOs/sec: 965.55 MBs/sec: 482.77 From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Johan Arwidmark Sent: Dienstag, 22. Oktober 2013 21:52 To: [email protected]<mailto:[email protected]> Subject: RE: [mssms] SQLIO tests values? There are no official numbers, we (the ECM MVPs) have been asking for years, but so far nothing... These numbers simply comes from my own testing at various customers... and even though they are far from exact, they should give a rough estimate. Shorthand, when you see a site server starting to perform sluggish, to take several minutes to generate a policy, or long time to process incoming data, long time to evaluate collections, die because of summarizations kicking in etc... I immediately suspect disk IO (even though it can be of other reasons too). / Johan From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Roland Janus Sent: den 22 oktober 2013 16:32 To: [email protected]<mailto:[email protected]> Subject: RE: [mssms] SQLIO tests values? Still a couple of questions while the file is created :) Where you get the IOPS numbers from? They may indicate good and bad SAN, but what CM likely needs is just good enough performance (for either disk layout and I was going with the same approach idea depending on performance). So far the only number ever mentioned (afaik) was 5000 IOPS at lasts year MMS and that is certainly not official and way below what you have. -R From: Roland Janus [mailto:[email protected]] Sent: Dienstag, 22. Oktober 2013 16:15 To: '[email protected]' Subject: RE: [mssms] SQLIO tests values? Perfect timing, thanks Johan. And yes, I've used fsutil :( All over again I guess. -R From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Johan Arwidmark Sent: Dienstag, 22. Oktober 2013 14:45 To: [email protected]<mailto:[email protected]> Subject: RE: [mssms] SQLIO tests values? Check this post: Sizing your ConfigMgr 2012 R2 Primary Site Server http://www.deploymentresearch.com/Research/tabid/62/EntryId/115/Sizing-your-ConfigMgr-2012-R2-Primary-Site-Server.aspx / Johan From: [email protected]<mailto:[email protected]> [mailto:[email protected]] On Behalf Of Roland Janus Sent: den 22 oktober 2013 13:10 To: [email protected]<mailto:[email protected]> Subject: [mssms] SQLIO tests values? I'm testing our SAN currently and while I get like ridiculous good numbers for reading (6 number figures IOPS and 1GB/s) it is a different story for writing. But I have no idea if those tests are valid :) Anyone having some advice on good parameters to test with? -R
