Re: random-access I/O
Posted by:
Rick James
Date: February 03, 2012 11:08PM
Vadim's blog covers the speed of loading, with InnoDB doing the compressing. That does not cover the speed of reading, which is presumably more frequent. Also, it seemed like he changed the block size; I don't know how (if) this makes any performance difference during loading.
If the client compresses the blob(s), then the CPU effort is kept out of MySQL. This is likely to make better use of multi-core boxes (or separate server). I would _guess_ this might eliminate the speed degredation that he found.
Was your I/O test that was running at the same time you were loading the table? Was that test using _sequential_ I/O? What OS are you using? If linux, what elevator strategy is it set to? Do you have RAID?
No, I don't know exactly where I am headed with this discussion. I don't have a model that explains what you are seeing.
Subject
Views
Written By
Posted
3907
January 31, 2012 03:50AM
1181
January 31, 2012 08:25PM
1143
February 01, 2012 05:19AM
1140
February 03, 2012 01:41AM
1360
February 03, 2012 02:39AM
Re: random-access I/O
1194
February 03, 2012 11:08PM
1114
February 04, 2012 08:54AM
1651
February 05, 2012 12:56PM
1155
February 08, 2012 07:22AM
1611
February 08, 2012 11:17PM
1033
February 09, 2012 05:30AM
1122
February 09, 2012 01:48AM
996
February 09, 2012 01:58AM
1205
February 09, 2012 10:52AM
Sorry, you can't reply to this topic. It has been closed.
Content reproduced on this site is the property of the respective copyright holders.
It is not reviewed in advance by Oracle and does not necessarily represent the opinion
of Oracle or any other party.