Skip to Content

It looks like I missed an interesting platform announcement by not going to ASUG/Sapphire this year.  SAP/IBM announced testing has opened on using Power CPUs to run HANA, rather than the x86 Intel chip sets that have been the sole source of HANA platforms.  This is not going to be a rant, as my Fast Is Not A Number post was in 2013, though I’ve got more questions than answers.  I’ll explore what this means to me.  Disclaimers: I am not representing my company, which uses both IBM and Intel (and AMD and HP and a few others I am sure) hardware architectures (and the cloud, yeah, we use that too). I’m also not going to disclose any conversations I had with either IBM or SAP on this topic in the past 2 years, other than to say I know stuff.  Some stuff. Not much, but some.

The initial “leak” I got was on Twitter; here are some posts I found with the tip:

@IBMcloud – Announcing more options for #cloud and big data: @IBMPowerSystems and SAP #HANA text [sic] program: http://bit.ly/1mTzlyo4:09 PM – 4 Jun 2014

More Power for big data: SAP begins testing HANA on Power Systems | Smarter Computing Blog

@IBMAIX – More Power for big data: #SAP begins testing HANA on #PowerSystems –  bit.ly/1kMqQCi (via @KristinBryson) –
SAP begins testing HANA on IBM Power Systems | ZDNet

Then the blowback, as analysts heard the news and pondered.

@sufw  [ @Sascha_Wenninger  | @ Sascha Wenninger ]  The news of HANA being available on POWER feels like a crushing victory of SAP’s past over its future. Zero value for net-new customers. – 8:52 AM – 9 Jun 2014

The upshot seemed to be negative on this announcement, though a few folks jumped in with the thought that more competition, or more choices, is better.

@PaulTom – @sufw You’re reaffirming my point. I’m not saying Power to replace x86 – I’m saying there should be more HW options available for HANA – 9:11 AM – 9 Jun 2014

@steverumsby [ Steve Rumsby] –

@PaulTom @sufw There’s really no reason for hardware restrictions – if it runs more slowly, that’s the customers tradeoff choice, right? – 9:14 AM – 9 Jun 2014

(and more via the above links)

Since I have somewhat of a professional interest here, I piped in:

@jspath55 – @PaulTom @steverumsby @sufw “would you like HANA to exist on…?” Hardware platform is important. I’m still looking for x86 SD benchmarks. – 9:25 AM – 9 Jun 2014

.

No answer posted to that question about actual SAP benchmarks on the HANA system (comments to that statement should go to the above blog, not here!)

Upside?

The commentary about tradeoffs leads to the idea that more choices in the market should drive down costs.  Seems reasonable, unless you think about how much of the HANA pricetag goes to SAP, and how much goes to the hardware vendor.  I wish I could be the fly on the wall for those negotiations (like I’m going to have with 2 vendors that could supply internet to my house – coax or fiber?).

Would an IBM Power CPU outperform an Intel CPU?  What would be the cost per transaction (or per analytic, or per whatever)?  On the first question, I’ve run my own benchmarks over the years, to get a sense of the performance of each available hardware platform.  Challenges with single versus multi-threading, and with applicability of those tests mean I would not have a definitive answer should the question of “SAP HANA on IBM Power – yes or no?” be posed in my day job.  Could I run tests on sample hardware?  That would be great.  Will we join the test program?  First, we have to apply these patches, see…

Downside?

Sascha’s tweet hits the nail on the head. I could argue that current x86 hardware also has long ties to an 1980s IBM PC architecture, and there do seem to be regular advances in the IBM Power CPU series.  As to cost, normal consumers would pay much more for IBM machines than for “PC clones” as I think of them.  Scaling up, I have a bias toward big iron.

The larger risk I see for customers who go that route is a later disengagement between IBM and SAP on this product line. Surely SAP’s backoffice maintenance overhead goes up if they’d need to triage an extra layer of complexity (“which chip set, with which firmware, on which family of CPU?”).  Before I started there, the company where I work had deployed SAP R/3 on Digital Equipment Corporation’s VMS/VAX OS/hardware.  The quote I heard was we were one of 3 in the world.  SAP stopped porting to that platform, as they have later abandoned some database ports (Informix I believe) for R/3 use. No wait, Netweaver. No, I mean ECC.

Sideways?

The dilemma here isn’t one of shopping for a system to buy today, as the announcement was not “on sale now”, it was “test drive now”.  Whether this platform goes to market will depend on the testing results, and, no doubt, bean counters locked up in a room figuring out the margins.

When I first took a database administration class, back in the 1980s, there were folks from the military (or related government agencies) in the room.  As the instructor talked about raw disk versus standard filesystems, the Feds indicated that they wanted the maximum throughput, so bypassing UNIX file system I/O and letting the database manage the disks was worth the risk of someone seeing a spare disk volume, formatting it for their own use, and blowing away the data contents without blinking.  Yeah, this is like that.

For why “I can Haz”, see: slang – Origin of "I can haz"? – English Language & Usage Stack Exchange

To report this post you need to login first.

2 Comments

You must be Logged on to comment or reply to a post.

  1. Gregory Misiorek

    Hi Jim,

    excellent post as usual. i cought your comment in my twitterstream, but the ongoing discussion around hanaonpower has temporarily died out, so thank you for bringing it back.

    as a former IBMer, i’m obviously biased, but i can’t simply forget where the origins of SAP, MSFT, and to some extent ORCL are. granted, the success is by no means guaranteed for the Big Blue if they are really serious about funding HANA on their flagship hardware, but they do need a success story and SAP can definitely help them with that.

    there are many things that can go wrong as many alliances in the past have shown, but i’m really optimistic that this initiative will get some serious tracking. it offers excellent alternative to Intel platform, which serves very well smaller organizations, but the large installations really need something like Power that can scale up to allow for millions of transactions per day.

    Big Blue may be late to the cloud, but once it’s starts moving, it’s the only hardware company that can provide comfort to the largest customers and if there’s anything that HANA really needs it’s the hardware on which it has to run.

    please keep us updated on what you are going to find out about this as it may require more than average sleuthing.

    rgds,

    greg

    (0) 
  2. Andy Silvey

    Hi Jim,

    +1 regarding great article. Almost gonzo journalism ala Hunter S Thompson.

    The more platforms which are supported by Hana the better, finally, at the O/S and platform layer we are seeing the democratisation of the Hana platform, only last week we saw RHEL becoming supported.

    I’ve heard it said by people at a company we all know, that for the largest customers, there are only two choices, HP or IBM, and the IBM customer base has been waiting patiently for Hana on IBM Power.

    Despite what the doubters say, this is good news.

    Best regards,

    Andy.

    (0) 

Leave a Reply