benchmarkingblog

Benchmarking and Systems Performance

New Oracle M5 and T5 SAP Benchmark Results: No SPARC at all

with 14 comments

If you were hoping for some Last Friday Night excitement from Oracle’s new SPARC servers announcement this week, we haven’t seen it yet. Oracle just this morning published two SAP SD 2-tier benchmark results, one on the M5-32 and one on the T5-8.

The IBM POWER7+ with DB2 10 result from back in September was 1.3x greater per core than the M5 and 1.9x greater than the T5 result.(1) The IBM average database request time was also much better and the CPU utilization of the IBM system was also more effective.

Will the sun come out tomorrow for Oracle?

************************************************

(1)IBM Power 780 (3.72 GHz) two-tier SAP SD Standard Application Benchmark result (SAP enhancement package 5 for the SAP ERP 6.0 application: 12 processors / 96 cores / 384 threads, POWER7+, 1536 GB memory, 57,024 SD benchmark users, running AIX® 7.1 and DB2® 10, dialog resp.: 0.98s, line items/hour: 6,234,330, Dialog steps/hour: 18,703,000, SAPS: 311,720, DB time (dialog/ update): 0.009s / 0.014s, CPU utilization: 99%, Certification #2012033

Oracle SPARC Server M5-32 SAP SD 2-tier result of 85,050 users, Average dialog response time: 0.80 seconds, Fully processed order line items per hour: 9,452,000,Dialog steps per hour: 28,356,000,SAPS: 472,600,Average database request time (dialog/update): 0.018 sec / 0.044 sec,CPU utilization of central server: 82%,Operating system, central server: Solaris 11,RDBMS: Oracle 11g,SAP Business Suite software: SAP enhancement package 5 for SAP ERP 6.0,32 processors / 192 cores / 1536 threads,SPARC M5, 3.60 GHz, 16 KB (D) and 16 KB (I) L1 cache and128 KB L2 cache per core, 48 MB L3 cache per processor,4096 GB main memory,Certification #2013009

Oracle SPARC Server T5-8 SAP SD 2-tier result of 40,000 users,Average dialog response time: 0.86 seconds,Fully processed order line items per hour: 4,419,000,Dialog steps per hour: 13,257,000,SAPS: 220,950,Average database request time (dialog/update): 0.049 sec / 0.131 sec,CPU utilization of central server: 88%, Operating system, central server: Solaris 11,RDBMS: Oracle 11g,SAP Business Suite software: SAP enhancement package 5 for SAP ERP 6.0, 8 processors / 128 cores / 1024 threads,SPARC T5, 3.60 GHz, 16 KB (D) and 16 KB (I) L1 cache and 128 KB L2 cache per core, 8 MB L3 cache per processor,2048 GB main memory,Certification #2013008.

Source: http://www.sap.com; Results current as of 03/25/12.

SAP and all SAP logos are trademarks or registered trademarks of SAP AG in Germany and in several other countries. Other names may be trademarks of their respective owners.

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: , , ,,,,,,,,,,,

About these ads

Written by benchmarkingblog

March 25, 2013 at 12:28 pm

Posted in Oracle, SAP, SPARC T5

Tagged with , , , , ,

14 Responses

Subscribe to comments with RSS.

  1. performance/$ cost is also important..why are you not mentioning the total cost of SQ + HW . performance/$ is the factor.

    martin francis k

    March 27, 2013 at 8:54 am

  2. A few callouts:

    – Price/Performance matters, core performance doesn’t
    – How’s that 10GHz Power processor projection from late 2000s working out for you?
    – We’ve Power7 and I can tell you it is around 30-40% better than Power6, we don’t care about rPerf numbers.
    – We’ve T4-4s and T5-4s, the performance boost is measured in ‘x’ not ‘%’.

    Ignore the T4, T5, & M5 at your own risk. Does it mean Power won’t leapfrog SPARC in the future, NO. Right now SPARC is better.

    IndianRepublican

    April 3, 2013 at 11:22 pm

  3. MA

    April 15, 2013 at 10:40 am

  4. Core performance absolutely matters when software is licensed per core. With Oracles “John Deere Green” software pricing for products such as Enterprise Edition database & RAC, customers get duped into buying lower cost servers (TCA) then rushed past the 3 – 5 year Total Cost of conversation. I’m sure you would like to dismiss this as well. Per core performance for Power allows customers to use far fewer (5 – 20X, even more) software licenses due to this higher performance. Even with Oracle manipulating their licensing factors to favor their technology (Oh Larry say it ain’t so). Furthermore, IBM’s Power7+ servers provide consistent results based on processor speed and number of cores. Unlike Fuji – I mean Sun – I mean Oracles servers which have different results whether it is their CMT based or SPARC64 based servers.

    Customers should not only ignore T4/T5 servers but they should run from them otherwise they will overpay for subpar RISC servers (T5 is more x86 like than RISC in features, particularly RAS).

    Btw, IBM Power7+ offers 4.2 GHz servers with x86 pricing at the low end and servers up to 4.42 GHz for enterprise customers at the high end. Oracle’s answer is buy a Exa this, a T5 that and maybe a M5 what?

    Let me pay the “T” one compliment – nice to see you finally broke the 2.0 GHz barrier with T4 – only took you 6 years! “T” must stand for terrible.

    Brett Murphy

    April 29, 2013 at 6:33 pm

  5. Sorry, I didnt understand, why we are talking of bit and Bytes. The customer always want to know the price/performance for a 2, 4, 8, 16, and so on Processors Machine or Cluster. And we always decide on this Ratio. Everybody talk about T5 & M5 but the Natural sucessor of the Sparc M Series is the M10 Series from Fujitsu, and with the M10 Series fujitsu follow a Very Good Idea from IBM, The build Block Machines. The Idea from IBM was very Good. The Sparc64-X from fujitsu is a better Processor compared with T5 M5 and Power7+ with the advanced of the License by core base. The benchMarks from Spec.org and SAP show it. About the others features, all Processors are very similar, the customers decided by the Operating System Facilities and the software they will install on. They don´t care about bit and bytes, and rpi per core, but only the final price for the solution that full fill the requirements.

    Miguel Lopes

    May 31, 2013 at 4:23 am

  6. Intel Xeon is faster and cheaper than Power 7+ and T5.

    Mr Bean

    June 20, 2013 at 11:53 pm

    • Neither of those statements are true. IBM has 2U & 4U servers that run at 4.2 GHz for 16 cores with 512 GB & 1 TB of memory respectively. The 2U servers offer both 1 & 2 sockets and have price parity with x86. So, build your 1 or 2 socket with 4, 6 8, 12 or 16 cores and whatever OS and virtualization you want and the Power7+ server will be virtually identical. Now that IBM has given us products that commodity sellers can’t ballyhoo as being “so expensive” we can get back to the real discussion – how much is the overall TCA and overall TCO. Power servers with the PowerVM virtualization suite deliver the greatest efficiency, reliability, performance and savings. x86 is trying to get better to move up to the enterprise space while IBM is taking what is already better and moving it down to the entry level space – watch out x86!

      Brett Murphy

      June 22, 2013 at 3:38 pm

      • IBM sales force worked very hard to create this myth. I was involved into many cross-platform migrations for Oracle databases (this probably mean you do integer operations only). If you take same technology level Intel will always be faster (unless you need a monster server). You will never buy SPARC or POWER If you compare real price (and price of upgrades). Take big organization with POWER/SPARC/Intel and compare internal charges to get virtual environment. Intel in hundreds times cheaper in my very big organization.
        Different thing is that all vendors have Intel base servers. Why IBM/Oracle selling Intel servers if POWER/SPARC products are better? Do not sell different vendor if you think your product is better. Educated customers choose better products – not a better sales forces.

        Serge

        August 8, 2013 at 4:14 pm

      • You are missing the point and value prop of Power servers. On any given day one manufacturers technology may be ‘faster’ than another…that is the leapfrog effect. Even if I accept that a x86 server is less expensive than a Power server. That is the mis-direction IT shops have been trained to look at – the acquisition price of the servers. The actual cost of the solution is what impacts the business. Servers, software, power, cooling, rack space, switch ports, # of OSes to maintain, other licensing costs, etc, etc. Even if a Power server was $100k and a x86 server is $10k to buy the fact that the customer may have to buy 10 of those $10k servers to do the same work that the $100k Power server can do with 1 server. Next, factor in the software licensing such as Oracle. 16 cores of Oracle Enterprise Edition on Power is $47,500 x 1.0 (licensing factor) x 16 (number of cores) which is $760,000 list price for the Oracle licensing. Nobody pays list but that is a constant. For the x86 solution lets say they are using 10 x 2 socket servers. Let’s make them 2 x 4 cores on each server just so the numbers don’t seem outrageous.. The Oracle licensing factor is .5. So, 10 x 8 x .5 x $47,500 equals $1,900,000. (This doesn’t even include Oracle RAC which would be a requirement for x86 and adds another $920,000) You can see the price of the servers gets lost in these kinds of costs. Factor in the 22% annual Oracle maintenance costs and it gets more and more expensive. The 22% Larry tax alone pays for the $100k Power server. I apologize if you think I’m calling x86 ugly or bad. I’m not. I work with clients daily demonstrating and actually doing the above with Power.

        You will need to provide more details to claim that Intel is “hundreds times cheaper in my very big organization”. I’ll defer to you in saying that you know your organization but unless you are using Xen or KVM and not VMware or HyperV then you are unaware of the cost of virtualization on Power servers.

        I do not know how we can connect but I would be happy to discuss this with you, fairly, openly and politely to show you the value prop of Power. Given what you have said I think you will be pleasantly surprised.

        Brett Murphy

        August 9, 2013 at 10:45 am

      • When comparing workloads details matters. Saying you need 10 x86 servers to match 1 Power 7 is just as flawed as saying x86 is always cheaper and fast then RISC. So gets add in a few more details to the x86 Power 7 comparison.

        A few inputs guiding the process
        RAC is not used for both. I have seen Power 6 frame dieing taking dozens of LPARS with it. So if RAC is required for x86 its required for Power 7 and SPARC. However HA is different for each app, some need RAC, some need RAC One Node others DataGuard and some require no HA.

        On average for Intel E5 xeons Power 7 does hold a 10% advantage in integer performance when comparing of the line Power vs Xeon.

        Servers used
        IBM 750 Power 3.5 Ghz 256 RAM – List $88,000 – SPECint 1060
        Dell R720xd 2.9 Ghz 128 RAM – List $9,200 – SPECint 697

        So looking at integer performance which matters more for DB performance we only need 2 x86 boxes to get more performance not 10! 697 x 2 = 1394. Side note Redhat on Power actually gets higher performance than AIX.

        HW cost
        IBM $88,000 Does not include expensive AIX OS cost and maintenance
        Dell – $18,400 Free to $2,600 for two servers

        Oracle cost
        Dell is 16 licenses (32 cores/.5) 16 x 47500 = 760,000
        IBM is 32 licenses 32 X 47500 = 1,520,000

        IBM + Oracle license (no support fees for OS/Oracle = 1,608,000
        Dell + Oracle license (no support fees for OS/Oracle = 778,400

        So IBM Power 7 is more than 2x the cost then Dell plus with Dell you get about 30% more CPU capacity.

        Details matter. And I will admit I didn’t all the details and each company will have a different list of details that matter to them.

        DetailsMatter

        September 17, 2013 at 1:10 pm

  7. I hope you detect that as I write I do with with the intent of being open, honest and not misleading. I enjoy technology and evaluate it based on what vendors say, customers say but also what the data says. Absolutely, Power servers have failed – it’s called bugs, operator error, power spikes, etc. Every vendor has them. As a general rule any RISC server is far more reliable than any off the shelf x86 vendor such as Dell, HP, Cisco, Oracle Exa, even IBM x. Within the RISC family I would argue at the bottom is SPARC, then Itanium with Power having the greatest reliability features.

    It’s interesting that your handle is “DetailsMatter” because you have failed to gather them all….only what suits you. Yes, there are lots of situations where what is written here may not be what customer A has but customer B does. Would hope the readers would understand this.

    It isn’t uncommon for those who work with x86 technologies to not understand Power technology. You can break up a core into 20 segments as small as .05. Each has up to 4 SMT threads or up to 80 threads per core. The way the server dispatches workload onto the cores is very, very efficient. As a result we can aggregate the workloads together delivering a QoS or Quality of Service to each.

    You license the Power server like the x86 server because you are either unaware or want to mislead the reader. Using your example you have your 16 licensable x86 Dell cores but I only have to license the cores I need. Now, any number I use here you will argue with saying it isn’t fair. I can’t help that. You have 2 x 16 core servers because that is the way you package them and thus have to license them that way. I could use a single socket 4, 6 or 8 core server and still only license the cores needed on that server. I can also turn off cores on that server so a 8 core server is only a 1 core….turn on another to make 2 and so on as the workload grows. OS, virtualization and application licenses always have to be considered when doing this.

    The last customer that “Dell” recommended 2 x 16 core server to I was able to do the workload on a single 6 core 740 server. Oracle was limited to a 4 core pool so I licensed 4 x $47,500 = $190,000. That particular 6 core server is about $35,000 making my solution cost about $225,000.

    I do this day in and day out with customers helping them get off inefficient x86 technology that only contributes to server sprawl, increased infrastructure costs, management nightmares and out of control software licensing.

    If you work for Dell, Cisco (includes VCE), Oracle or HP I wouldn’t expect you to tell your customers how they can save money with IBM Power. Just like buying a car, customers have a responsibility to do their homework. We work to educate them on the technology, inform them about our abilities, strengths but also weaknesses.

    Brett Murphy

    September 28, 2013 at 3:54 pm

  8. Hi,
    I really like your post.
    Thanks for sharing such great information.
    It is very informative and provides knowledge ofsap sd certification
    .
    Thanks again

    ramdin

    December 4, 2013 at 7:26 am

  9. […] “competition” was certainly IBM, which promptly took aim at Oracle’s benchmarks. That clearly wasn’t going to be the end of it; and while IBM’s investment was almost certainly […]

  10. […] M5 and 1.9x greater than the T5 result,” Elisabeth Stahl, chief technical strategist for IBM, wrote in a post taking aim at Oracle’s SAP benchmarks. “The IBM average database request time […]


Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s

%d bloggers like this: