So in recent weeks I have had a massive response, over 300 hits both internally and externally, to the blog about our spectacular UCP for Oracle customer results. It has generated a huge amount of discussion and I am finding a lot more out about the detail. Even Oracle responded (see previous post). In the detail it transpires that on average we reduce the number of Oracle licenses by 50%, not my previously posted 20%. Since Oracle is prone to significant discounting (I've heard from one customer up to 85% off list price), it is not possible (today) to say how much that means in hard cash in each of the cases I have presented.
Notwithstanding this, I have been contacted by another colleague who thinks we can add more kudos to our customer PoC results.
We have have had some independent testing done by a fascinating organisation in Switzerland. Benchware is a small SW house that builds and sells benchmark software that looks at the reality of running an (Oracle) application in all environments. This means, unlike SPC-1 or TPCC bench-marking, where you can tailor the array to meet the outcomes, you run the test to reflect a realistic working environment and the outcomes are far more accurate as they look at end-user response time, speeds across the environment and access times according to varied workloads. So, what benchware did was to take a UCP from HDS and run a battery of tests against it. They have also done the same against several different Exadata configurations and one of HP servers running with a Violin memory All Flash Array (AFA).
The bottom line is that the UCP stands head and shoulders above the other two as it scales massively well and has far greater throughput, using far less cores, drives and therefore Oracle Licenses. This can be seen in the graphs that demonstrate varied workloads of sequential reads and writes and random reads and writes. It also graphs the scalability by adding nodes to each environment, which, when you compare the graphs from each test, illustrates just how good the UCP is. Also, remember that the UCP can do 9i, 10g and 11g applications too, which the Exadata cannot. It can also do varied workloads in the same chassi, which the Exadata cannot. Now you can understand the real value of the UCP, it is a consolidation machine as well as a cost-efficient performance beast!
you can find all the results here. Benchware Benchmark