Db2 purescale and Oracle RAC

IBM announced a new technology called Purescale recently.It is basically a technology that excels in horizontal scalability.

So what is horizontal scalability??How is it different from vertical scalability??

Horizontal scalability is the ability to add capacity by adding nodes to the cluster whereas vertical stability is the ability to increase capacity by adding extra resources to the existing entity/server.

Purescale is aimed at achieving 3 important goals:
1)Application transparency: No coding changes are required when you add extra nodes

2)Unlimited capacity:This is achieved by adding as many nodes as needed.But there are limitations on the platforms to begin with

3)Data availability:This system aims at zero downtime and is completely available if one or many nodes fail

Is Purescale a replacement for HADR??

No. Purescale is not a replacement and it has only one shared database copy.So, if HADR can be applied for a purescale system it is the optimal availability scenario.IBM thinks that they can do this in the coming days


Oracle RAC Vs DB2 Purescale:Which is better??

When you use Oracle RAC lot of application changes have to be made when we add new nodes.But with Purescale the application is completely transparent to the changes in nodes.Also Purescale has centralized resource management system that manages the lock and other resources.

Basic db2 federation setup

Db2 federation can be used to retrieve information from non-db2 sources like oracle,sql server.It also can be used with db2 datasources.Setting up federation can be confusing at times.Below is the basic procedure that can be used to setup federated system:

1)Enable federation: Federation can be enabled by running the following command

update dbm cfg using federated yes immediate;

2)Create wrapper:Below is the command you would issue for a db2 datasource.Wrappers differs for each RDBMS.Please refer to IBM documentation to get more information on wrappers

create wrapper DRDA;

3)Catalog the datasource information: Datasource should be cataloged properly and the federated server uses the access method depending on the datasource type.

4)Create a server definition.Refer to the below example for db2/udb datasource

db2 "create server SAMPLEHOST

type DB2/UDB

version 9.5 wrapper drda

authorization 'user1'

password '*****'

options(node 'SAMPLENODE', dbname 'SAMPLTST')"

5)Create user mapping:

db2 "CREATE USER MAPPING FOR federuser server SAMPLEHOST options(remote_authid 'user1',REMOTE_PASSWORD '*****')"

6)Create nickname:

CREATE NICKNAME EMP FOR SAMPLEHOST.SAMPLTST.EMP;

If you follow all the above steps, basic federation setup is done.Nicknames can be on tables or views that reside in the datasource.Once the federation is setup, nicknames can be used to refer to the actual datasource objects.If some part of the code can not be processed by the data source, it is not passed to the data source.The datasource in this case will be using alternate functionality that is close or the data set will be sent to the federated server for additional processing

HADR performance - part 1

HADR stands for high availability disaster recovery .This solution is used for disaster recovery purpose on db2 udb databases.HADR should be configured properly inorder to have optimal performance:

1)HADR synchronization mode:HADR can be run in 3 different modes SYNC,NEARSYNC,ASYNC.SYNC mode gives the best protection to data.In this mode primary has to wait until the changes are committed and written on the standby.Primary waits for the acknowledgement from the standby server.In NEARSYNC mode, standby sends acknowledgement as soon as the logs are in memory of standby server.And in ASYNC mode , primary does not wait for any kind of acknowledgement from the standby.Proper synchronization mode has to be chosen for optimal performance

2)DB2_HADR_BUF_SIZE:This registry variable controls the size of the receive buffer .Receive buffer is the area of memory where the logs are received before they are replayed.You can use the db2pd -db dbname -hadr on standby to monitor the usage of receive buffer.If you see it reaching 100 during the workload, you need to increase the value of DB2_HADR_BUF_SIZE

3)DB2_HADR_SOSNDBUF and DB2_HADR_SORCVBUF:There are the socket send buffer size and socket receive buffer size respectively.If the size for these parameters is too small then the full bandwidth can not be utilized.Generally increasing this to a bigger value would not impact performance negatively.


4)Logfilsz:Size of the logfile plays an important role in the performance,Generally this size should be few hundred MB.