You should read up on the "show status" and "show variables" output. That will give you a start as to perhaps some obvious issues. For instances, your opened_tables and threads_created should not be large. Using the right table types is also a very big issue, although you may have gone down that path already. You can try doing a search on remember.yahoo.com and mysql, they ran into some interesting performance bottlenecks they had to solve. Installing mytop may also be helpful to determine what's going on. Finally, you should also consider what is going on in the OS, I like vmstat for a quick overview. I don't know what you definition of "performance is not good". Following a "strict formula" may be part of your problem. Nobody normalizes their database to fifth normal form, it would be too slow and complicated.

Check this link out:
http://dev.mysql.com/books/hpmysql-excerpts/ch06.html
The last paragraph has a bit on yahoo.


On Sep 11, 2007, at 2:10 AM, Hailiang Ji wrote:

Folks,

A help needed. My manager's pushed me to optimize the tables that I
created in distributed in several DBs. I have tried best to explain to him that I have followed the strict formula design to do the Join, Search and
so on. However, the system performance is not good enough yet when our
system get thousands' users visiting in the same time.

Could anyone point me something so that I can tune the system performance? I don't think my db tables have problems. I guess I should investigate on the deployment, for example, cluster, load-balancing and so on? Our system
hs big traffic at daytime, usually at noon time or at evening.

Thanks in advance,

------------------------
Hailiang Ji, Developer
Email: [EMAIL PROTECTED]
Web: www.mydanwei.com
myDanwei, Inc.
-----------------------

--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe: http://lists.mysql.com/mysql? [EMAIL PROTECTED]



--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:    http://lists.mysql.com/[EMAIL PROTECTED]

Reply via email to