Elisabeth Stahl on Benchmarking and IT Optimization

Moonshot vs. Metrics

with 2 comments

Some of you may know that last week I was on some college visits. I love this time of info sessions, tours, cafeterias, classrooms, and dorms. I even got to stay in a real dorm one night. Funny, it was not any dorm I had ever been in, not at a college that I had ever been at, and not even a section of a city I had ever been to – but the cooking smells brought me right back to senior year.

Anyway, one thing I’ve noticed with these visits is how important statistics are. At first I was so tired of asking/hearing the same questions on metrics. I mean, isn’t it the feel of the campus that matters? But I found that the really key questions are ones like these: What percent of freshmen live in dorms, what is the student to professor ratio, what is the placement rate 6 months after graduation in the field of study? The answers to these questions which are backed by real data really matter because in the end that’s what the big bucks are paying for.

I’ve been feeling that way this week as I look at some of the latest IT news. I see Moonshot claims, backed by “internal HP engineering.” I see Fujitsu M10 and Oracle faster performance claims that don’t even go that far. If I don’t have any data maybe if I don’t include a footnote nobody will notice.

Meanwhile, IBM this week published a new #1 SAP Sales and Distribution 3-tier benchmark result. 266K users, over 1.4M SAPS, over 29M line items per hour, over 88M dialog steps per hour, on the POWER7+ IBM Power 780 with DB2 10.5.(1) With more metrics available than you probably even want to know about.

I hope HP’s alternative thinking with Moonshot was referring to the United States Apollo program — and not the abortive Soviet moonshot or the defunct beer with caffeine.

(1)IBM Power 780 three-tier SAP SD standard application benchmark on SAP enhancement package 5 for SAP ERP 6.0 achieved 266,000 SAP SD benchmark users. Configuration: 8 processors / 64 cores / 256 threads, POWER7+ 3.72 GHz, 512 GB memory, running AIX 7.1, DB2® 10.5; dialog resp.: 0.84s, line items/hour: 29,433,670, dialog steps/hour: 88,301,000, SAPS: 1,471,680, DB time (dialog/ update): .036s/.061s, DB CPU utilization: 97%, average application server CPU utilization: 88%. Certification #2013010.
Source: Results current as of 4/10/13.

SAP, mySAP and other SAP product and service names mentioned herein as well as their respective
logos are trademarks or registered trademarks of SAP AG in Germany and in several other countries all
over the world.

The postings on this site solely reflect the personal views of the author and do not necessarily represent the views, positions, strategies or opinions of IBM or IBM management.

technorati tags: , , ,,,,,,,,,,,,,,,


Written by benchmarkingblog

April 10, 2013 at 12:03 pm

Posted in POWER7, SAP

Tagged with , ,

2 Responses

Subscribe to comments with RSS.

  1. Big deal….
    Only 2 other 3-tier SAP SD benchmarks have been published for the last 5 years!!!
    Good job beating the benchmarks of some really out-dated Intel servers running on
    Windows/Linux… ;P

    Tan Ah Beng

    April 15, 2013 at 4:44 am

  2. Dear Blog Owner: There is only one neutral source for benchmarks which is

    Comparing the numbers there, one finds that Power7 4GHz systems indeed deliver the highest throughput performance with 64,128 and 256 core systems for CINT2006rate and CFP2006rate, with impressive numbers.

    In this same category the Fujitsu SPARC64 X (M10-4S) delivers currently the highest available throughput at all with 1024 cores (proofing the nice scalability and SMP features of the SPARC architecture), while being on position #2 up to 256 cores, behind Power7.

    The highest performance per core and up to 16 cores is still available with Intel Xeon E5-2690 2.90GHz systems (here top 5 leaders are Fujitsu in CINT2006 and Huawei and Oracle in CFP2006).

    More important, with regard to the surprisingly strong new SPARC architectures, seems to be that SPARC S3 and Sparc 64X systems have now surpassed even the fastest Intel midrange and high-end systems with 8 CPUs, 128 cores or more and are getting closer and closer to the Power7 in terms of pure integer and fp benchmark performance.
    These integer and fp benchmarks by the way do not at all take advantage of advanced chip features (like crypto engine on SPARC S3 and statements on chip with SPARC64 X) while real world apps do. They basically take only advantage of ALU speed.

    The importance of these advanced chip features is shown in benchmarks which simulate real world applications. Here Oracle is leading with far distance in terms of Java Enterprise System performance. (with Oracle Weblogic Server Standard Edition Release 12.1.1 on Oracle SPARC T5-8 and 57422 EjOPS. Oracle occupies the first 6 positions in this benchmark.
    The fastest IBM Power7 system is on position 9: WebSphere Application Server V7 on IBM Power 780 and DB2 9.7 on IBM Power 750 Express with 16646 EjOPS.)


    May 12, 2013 at 1:42 pm

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: