Obviously critical to Sun Grid's success is the ability to support transactional applications on the compute farm. Though the core economics of a Solaris 10 cpu at $1/cpu-hr are best in class (versus $7-$14/cpu-hr for many self hosted enterprise environments), who would have thought that the economics could get even better so quickly. Today, Sun announced couple that with some recent performance benchmark figures from our Java Application Platform:

“The benchmark configuration was comprised of 13 Sun Fire V20z servers, each equipped with two AMD Opteron processors Model 248 and running the Sun Java System Application Server 8.1 Standard Edition. In addition, a Sun Fire E2900 server, equipped with 12 UltraSPARC IV 1.35 GHz processors, ran the latest version of the Oracle 10g Enterprise Edition Database.

The benchmark delivered a class leading price / performance of $164/JOPS in the application tier (combined software / hardware and support cost) versus IBM's best submission of $504/JOPS(3). This translates to 90% of the performance for 33% of the total application tier cost.

The benchmark also demonstrated Java System Application Server's ability to handle complex workloads under very heavy load. During the benchmark, the Application Server supported over 9400 concurrent clients, and processed over 4.3 million complex transactions per hour.”

Wow! combine this with the reduced cost and complexity associated with a unified development, test to deployment model in which the integrity of the “digital runbook” is maintained, and people should start to ask not IF but WHEN refactoring for Sun Grid should occur.

Permalink
Trackback: http://blogs.sun.com/dhushon/entry/application_containers_cost_per_transaction