MySQL Forums
Forum List  »  NDB clusters

Poor read throughput - 240 read operations per second (+)
Posted by: Anton Mironenko
Date: September 30, 2005 12:59AM

Hi,
at the moment we are thinking about buying MySQL cluster, so it can be a part of our telekom solution.

So we are doing tests.

Right now we have a poor read throughput - 240 read operations per second.
Here is a configuration:
Number of hosts - 4
hosts' hardware configuration:
1 - SunFire Blade B200x, 2 x Intel(R) Xeon(TM) LV CPU 2.0GHz HyperThreading, RAM - 4GB, OS - Solaris x86 9
2 - SunFire Blade B200x, 2 x Intel(R) Xeon(TM) LV CPU 2.0GHz HyperThreading, RAM - 4GB, OS - Solaris x86 9
3 - 2 x Intel(R) Xeon(TM) CPU 2.40GHz HyperThreading, RAM - 4GB, OS - Linux RHEL 3
4 - 2 x Intel(R) Xeon(TM) CPU 2.40GHz HyperThreading, RAM - 4GB, OS - Linux RHEL 3

2 NDB nodes are installed on each host, so there are 8 NDB nodes in the cluster.

Number of replicas = 2.

Hosts are connected to each other with Gigabit network.

Read operation query is performed from JDBC client and looks like:

SELECT * FROM ACC_USER WHERE ACC_USER_ID = [specified-acc-user-id];

Table structure is the following:

CREATE TABLE `ACC_USER` (
`ACC_USER_ID` decimal(18,0) NOT NULL default '0',
`PERSON_ID` decimal(18,0) default NULL,
`ACCOUNT_ID` decimal(18,0) default NULL,
`CURRENCY_ID` decimal(18,0) default NULL,
`LANGUAGE_ID` decimal(18,0) default NULL,
`IS_LOCKED` decimal(1,0) NOT NULL default '0',
`M_TYPE` decimal(3,0) NOT NULL default '1',
`ACCOUNT_LEVEL_ID` decimal(18,0) default NULL,
`COMPANY_ID` decimal(18,0) default NULL,
`OWNER_ACC_USER_ID` decimal(18,0) default NULL,
`T_STATE_ID` decimal(18,0) default NULL,
`T_COS_ID` decimal(18,0) default NULL,
`T_CHRG_PROFILE_ID` decimal(18,0) default NULL,
`BEGIN_DATE` timestamp NOT NULL default '0000-00-00 00:00:00',
`END_DATE` timestamp NOT NULL default '0000-00-00 00:00:00',
`M_VPN_EXT` decimal(32,0) default NULL,
`IS_RING_TIME_INCLUDED` decimal(1,0) default NULL,
`T_LAST_TRANSITION_ID` decimal(18,0) default NULL,
`T_ASP_ZONE_CHP_ID` decimal(18,0) default NULL,
PRIMARY KEY (`ACC_USER_ID`),
KEY `ACCOUNT_ID` (`ACCOUNT_ID`),
KEY `OWNER_ACC_USER_ID` (`OWNER_ACC_USER_ID`)
) ENGINE=ndbcluster DEFAULT CHARSET=latin1;

The number of objects in this table is about 500K.


The question is: why do we have such a small throughput? is it a limit for MySQL cluster with the configuration above,
or we need to dig into our client - it might be not optimized?

If 240 read operations per second is not a limit for MySQL cluster, what is its limit for the configuration above?

Thanks in advance.

Options: ReplyQuote


Subject
Views
Written By
Posted
Poor read throughput - 240 read operations per second (+)
2324
September 30, 2005 12:59AM


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.