Hi Brent,

Wow, it seems like you are going to extremes. To jump from myisam to heap is a big step. Did you try using InnoDB? It would handle locking issues much better since it doesn't lock the table. Heap tables can be pretty dangerous since it's all in memory. If the machine crashes, you'll lose the data.

I know that, but I do regular (cornjobs) backups to myisam-tables and I can reconstruct the whole table from the machines that insert/update the data.

Based on your information, you want to get the best disk I/O you can. You won't get that out of a single "IDE" drive, even if it is one of the latest SATA based with command queuing. I don't think you'll get anything faster than heap tables and tons of RAM. But there is certainly finite scalability because of the use of RAM. Clusters may be the way to go for scalability, but I would work on getting your data disk based for maximum scalability.

For my case, scalability means more updates/second and more selects/second. Not larger tables. At least not much larger, and this table is using less than 300 MB of memory right now. So I see no point in using anything disk-based. 2 years ago we started with myisam, then changed to innodb, found out it wouldn't give any better performance in our case and switched back to myisam since that makes the setup of replications much easier. Then we changed it again 6 months ago and now use memory-tables.

I would try InnoDB and maximize you disk setup. I don't know how many disks you have in your RAID and if it's hardware or software based. More disks will add speed by splitting the load across more disks. Just keep in mind the limits of your SCSI card too. You may need to add a card to split the load.

These two systems have Hardware-RAID (SCSI storage controller: LSI Logic / Symbios Logic (formerly NCR) 53c1030 (rev 07)) width 2 disks in raid1 and the DB is myaybe 40 Gb of size. I have no performance trouble on any other table.

Jan

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

Reply via email to