Hi Dan, I wanted to comment on the comparison between the scsi cards - plase see below.
2003年 6月 10日 火曜日 16:55、Jeremy Zawodny さんは書きました: > On Mon, Jun 09, 2003 at 05:27:00PM -0700, Dan Edwards wrote: > > > Can you grab the output of 'vmstat' or iostat or something that'll > > > measure disk I/O next time this happens? > > > > I'm going to try real hard not to do that again until it is upgraded or > > another solution is found. > > Yeah, I know what you mean. > > > > What'd help more is adding more disks not a single faster one--unless > > > the SCSI disk is an order of magnitude faster. > > > > What is usually the better way to take advantage of multiple disks, raid > > stripping or seperating the dbs to different disks? I've been playing > > with scsi raid on the slave machine. It didn't help other performance > > problems as much as I was hoping it would. The issue there is takeing > > these large tables and generating reports. I have one report that takes > > about 6 hours. > > That depends on your access patterns and where the bottlenecks are. > I've seen some pretty good speed boosts from putting the most active > databases on sepraate disks. On the other hand, if all your activity > ins in a single database (or even a single table) then RAID-0 might be > better. > > Which type of RAID did you try, anyway? > > > Also while I'm asking, the scsi setup I have on the slave is a p4 2.4, > > lsi160 card with 4 fujitsu Ultra160 10,000 rpm 9 gig drives, useing > > linux software raid. Definatly the cheapo route. My question is, how > > does this setup with the $45 lsi card compare to say a $300+ adaptec > > card with hardware raid. Especially considering that the cpu is so > > underutilized? Is there any good sites that compare mysql performance on > > different hardware? All I could find is on mysql.com with 4 year old > > hardware. I strongly believe that there is a substantial difference between the scsi cards available on the marcet. Depending on the I/O request you are sending from the OS the data will either be transported f.e a) from the harddisc A to harddisc B - for data copying b) from hardisc to CPU - for data processing c) from harddisc to another memory (USB-stick, Flash etc) - for data copying etc... I understood that adaptec has good implemented algorythms f.e for a) This means that f.e. while f.e an LSI card has a similar speed and feature for b) and c) it might be completely outperformed by adaptec on a) requests. (or any other maker of course) So depending on your sql request type and way how you use your database (raid level, etc.) I believe that the performance will vary. While f.e. the adaptec scsi card might perform such requests as in a) almost independent from your main CPU - using their own CPU provided on the SCSI card - it allows you Main CPU to be idle for other tasks to perform. The LSI card would need to pass the request to your main CPU and use the Main CPU Power for this. While this is a complex topic I believe that figures regarding several I/O request types are available from different companies. You would then need to go and compare these "benchmarks" to be able to decide if the Adaptec, LSI, Tekram etc. is the one you believe gives you the best value. > > I suspect that you'll see little difference with a more expensive > card. Linux software RAID usually performs quite well. > > Jeremy I disagree with Jeremy, I think depending on your request type you will definitely see performance differences, the question is if they apply to your needs ;-). > -- > Jeremy D. Zawodny | Perl, Web, MySQL, Linux Magazine, Yahoo! > <[EMAIL PROTECTED]> | http://jeremy.zawodny.com/ > > MySQL 4.0.13: up 7 days, processed 208,195,890 queries (342/sec. avg) best regards --- Valentin Nils Internet Technology E-Mail: [EMAIL PROTECTED] URL: http://www.knowd.co.jp Personal URL: http://www.knowd.co.jp/staff/nils -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
