8 messages in com.mysql.lists.clusterRe: show tables query time| From | Sent On | Attachments |
|---|---|---|
| Tanner Postert | 29 Jun 2006 18:15 | |
| Stewart Smith | 29 Jun 2006 22:27 | |
| Tanner Postert | 30 Jun 2006 00:32 | |
| Stewart Smith | 30 Jun 2006 00:54 | |
| Tanner Postert | 30 Jun 2006 11:53 | |
| Tanner Postert | 30 Jun 2006 13:01 | |
| Ross McFarland | 30 Jun 2006 13:37 | |
| Stewart Smith | 30 Jun 2006 23:25 |
| Subject: | Re: show tables query time |
|---|---|
| From: | Stewart Smith (stew...@mysql.com) |
| Date: | 06/30/2006 12:54:23 AM |
| List: | com.mysql.lists.cluster |
On Fri, 2006-06-30 at 00:32 -0700, Tanner Postert wrote:
from MyISAM 3.23 to NDBCLUSTER 5.0.22
i ran some further tests. a like %...% query on a varchar field with an index returns in 0.53 sec on the original server while there is a full 300-500 user load on it. the ndb cluster runs the same query in 11.6seconds.
isn't the cluster supposed to improve performance?
not start-to-end query performance. Improved performance is through parallelism of queries (many queries at once). i.e. with cluster you should be able to support a lot more than the 300-500 users, but each query probably isn't going to be quicker (network hops do have a cost).
improving the start-to-end query time can be done by using a faster interconnect such as SCI.
the like %...% query won't be using an index. I recommend you make sure that engine_condition_pushdown is enabled to make this evaluated in the data nodes instead of the sql nodes. It's still a table scan, but at least it's not a table scan over the wire.
let us know how you go.
-- Stewart Smith, Software Engineer MySQL AB, www.mysql.com Office: +14082136540 Ext: 6616 VoIP: 66...@sip.us.mysql.com Mobile: +61 4 3 8844 332
Jumpstart your cluster: http://www.mysql.com/consulting/packaged/cluster.html