| Bug #44203 | Throughput of MySQL Cluster does not improve | ||
|---|---|---|---|
| Submitted: | 10 Apr 2009 8:37 | Modified: | 14 May 2009 11:53 |
| Reporter: | Kazuki Kudo | Email Updates: | |
| Status: | No Feedback | Impact on me: | |
| Category: | MySQL Cluster: Cluster (NDB) storage engine | Severity: | S5 (Performance) |
| Version: | 5.1.30-ndb-6.3.20 | OS: | Linux (RHEL5.3) |
| Assigned to: | Assigned Account | CPU Architecture: | Any |
[14 Apr 2009 11:49]
Gustaf Thorslund
Kazuki Kudo, Can you please provide your config.ini and my.cnf files. Best regards, Gustaf
[16 Apr 2009 2:09]
Kazuki Kudo
config.ini
Attachment: config.ini (application/octet-stream, text), 2.05 KiB.
[16 Apr 2009 2:21]
Kazuki Kudo
my.cnf
Attachment: my.cnf.txt (text/plain), 2.97 KiB.
[14 May 2009 23:00]
Bugs System
No feedback was provided for this bug for over a month, so it is being suspended automatically. If you are able to provide the information that was originally requested, please do so and change the status of the bug back to "Open".
Description: Hello, Throughput of MySQL Cluster does not improve. 4 threads improve. It goes down rapidly in 5 threads. storage engine :ndbcluster memory clienttool :mysqlslap mysqlslap -u rduser -ppassword -h sqlnode --auto-generate-sql \ --auto-generate-sql-guid-primary --engine=ndbcluster --number-int-cols=3 \ --number-char-cols=5 --auto-generate-sql-load-type=[xx] \ --auto-generate-sql-write-number=1000 --auto-generate-sql-execute-number=1000 \ --concurrency=[yy] ndb-cluster-connection-pool option of my.cnf was also changed. Result. (1) auto-generate-sql-load-type=mix ndb-cluster-connection-pool=1 thread throughput ------------------- 1 1733 2 3118 3 3755 4 4221 5 112 10 111 (2) auto-generate-sql-load-type=mix ndb-cluster-connection-pool=2 thread throughput ------------------- 1 1757 2 2994 3 3926 4 4854 5 102 10 102 (3) auto-generate-sql-load-type=key (pkey select) ndb-cluster-connection-pool=1 thread throughput ------------------- 1 2547 2 4970 3 5323 4 5953 5 49 5 48 (4) auto-generate-sql-load-type=write (insert) ndb-cluster-connection-pool=1 thread throughput ------------------- 1 1357 2 2439 3 3178 4 3789 5 1124 10 959 In all the cases, the activity ratio of a resource (CPU, HARD DISK, NETWORK) is low. SELECT is a bottleneck. Best Regards, Kazuki Kudo. How to repeat: mysqlslap -u rduser -ppassword -h sqlnode --auto-generate-sql \ --auto-generate-sql-guid-primary --engine=ndbcluster --number-int-cols=3 \ --number-char-cols=5 --auto-generate-sql-load-type=[xx] \ --auto-generate-sql-write-number=1000 --auto-generate-sql-execute-number=1000 \ --concurrency=[yy]