Exadata TPC Benchmarks

Posted in: Technical Track

One of the questions in my webinar this week related to TPC benchmark results for Exadata. Kevin Closson blogged about a TPC-H test run against an Exadata V1 machine back in June 2009. It was the fastest 1,000GB TPC-H result at the time, but has since been overtaken by a ParAccel result that makes a mockery of the benchmark according to Curt Monash by running almost entirely in RAM.

I’m not aware of any TPC results for Exadata V2. However a TPC-C test was run in late 2009 using a Sun F5100 flash array on Sun T5540 servers, setting a record for the benchmark that still stands.

The Oracle wiki has a great overview of the various benchmarks and what they actually measure.

email

Author

Want to talk with an expert? Schedule a call with our team to get the conversation started.

About the Author

Marc is a passionate and creative problem solver, drawing on deep understanding of the full enterprise application stack to identify the root cause of problems and to deploy sustainable solutions. Marc has a strong background in performance tuning and high availability, developing many of the tools and processes used to monitor and manage critical production databases at Pythian. He is proud to be the very first DataStax Platinum Certified Administrator for Apache Cassandra.

6 Comments. Leave new

Couple things:

First, the TPC-H benchmark you site did use Exadata V1 Storage (but was not a HP Database Machine) but was also an in-memory benchmark using Oracle’s In-Memory Parallel Execution.

Second, the TPC-C benchmark you site used the Sun Storage F5100 Flash Array, not the Sun Flash Accelerator F20 PCIe Card which is based on Oracle’s Sun FlashFire technology.

Reply
Marc Fielding
August 13, 2010 4:53 pm

Ahh, I was wondering if it the benchmark had been run in-memory. And that probably helps explain why we haven’t seen a TPC benchmark for Exadata V2 either, since the balanced configuration, storage intelligence, and fast I/O mean little when the whole database runs in memory.

I’ve also corrected the mention of FlashFire disks on the Sun benchmark. It was indeed done with F5100 storage.

Reply

Hi Marc,

I think you may have confused a few things, and I’d hate for your readers to think you intentionally misrepresented the truth.

Allow me to clear them up:

* The Exadata 1000GB V1 benchmark was in-memory (as Greg stated) in June 2009

* The ParAccel 1000GB benchmark which beat Exadata’s performance (a little, but not hugely) is dated April 2010. Note, this was a price/performance win objective at 7.8x better price/performance and not as much an all-out performance win objective

* The Curt post you mention is:
1) Dated Jun 2009…so almost a year BEFORE the 1TB ParAccel benchmark that beat the Exadata benchmark
2) Is criticizing ParAccel’s 30TB benchmark (not the 1TB) which ParAccel had to pull following Oracle’s objection that specifying compression syntax in the DDL was a technical violation.
3) And Curt has proven he has a cynical bias for his non-clients…so you need to take his posts with a grain of salt anyways: https://www.dbms2.com/2010/07/30/advice-for-some-non-clients/

Exadata is a great product, and it can stand on its own. It would be a shame to taint its success with simple FUD about its competitors.

Reply
Marc Fielding
August 13, 2010 4:53 pm

Thanks for the clarifications Bob!

Reply
Kevin Closson
August 16, 2010 12:22 pm

Hi Marc,

A simple read of my blog post (to which you are referring) would have made it clear that Exadata was used in that old TPC-H in-memory PQ result but the test was *not* a showcase of Exadata since it ran in memory (database grid memory). There was no offload processing involved.

Reply
Marc Fielding
August 16, 2010 7:24 pm

Thanks for the note!

Reply

Leave a Reply

Your email address will not be published. Required fields are marked *