Triton Consulting Delivers on DB2 pureScale
March 28th, 2012

For organisations who are looking to their IT function to bring 100% uptime for growing workloads, then DB2 pureScale needs to be at the top of the shopping list.


Business hours are never over in a globalised world meaning that even a small gap in availability can cost millions in lost revenue and customer confidence. For those companies experiencing workload peaks DB2 pureScale allows the IT department to manage those increases in demand whilst maintaining the essential aspect of excellence in customer service.


IBM has calculated that on average server utilisation is around the 15% figure which leaves 85% of unused capacity. DB2 pureScale enables cost cutting by reducing wastage, and with ‘pay by the day' pricing you are able to pay for the extra capacity you need when it's required.


IBM DB2 pureScale offers clustering technology that delivers continuous availability, unlimited scalability and application transparency. DB2 pureScale is based on industry leading DB2 for z/OS data sharing technology and is now available as an option on IBM DB2.


Triton Consulting are pleased to provide DB2 pureScale consultancy and services to help organisations plan and implement the groundbreaking technology. James Gill, Principal Consultant at Triton Consulting explains in more detail the benefits of pureScale: ‘Geographically Dispersed pureScale Cluster (GDPS), also known as ‘stretched cluster', has been introduced by IBM to overcome the potential for service outage in the event of a whole data centre failure. By extending the range of the Infiniband interconnects, and utilising SAN-attached GPFS clustered storage, the solution is able to provide an active-active configuration spread over two data centres.'
James continues: ‘The stretch cluster solution relies on two key technologies: Infiniband extenders, to increase the range of the signalling network, and GPFS mirroring. The range that the cluster can be extended to will depend on the workload being performed - the higher the volume of update work, the shorter the separation without adverse impact on application throughput.


The performance cost of stretched cluster is largely due to two aspects. Firstly, regarding additional lock timing, a lock request will take somewhere in the region of 10 - 15 microseconds for the RDMA request in a standard pureScale cluster. When using the stretch cluster technology, it's necessary to account for an additional 5 microseconds per kilometre of fibre optic.


Secondly, we have to look at disk write lag caused by GPFS mirroring. The greater the separation of the GPFS copies in the SAN, the greater the write I/O latency in the solution. This typically has the most impact on database logging. Note that absolute logging rates can be reduced (per member) in a balanced pureScale cluster by introducing additional members to take up the load.'


Importantly, with DB2 pureScale there is no need to change your application code and it's possible to grow your application to meet the most demanding of business requirements. With critical systems available all the time and access to data being uninterrupted throughout planned and unplanned outages it's not surprising that industry is demanding DB2 pureScale for optimum scalability and availability.