The server makers of the world have been busy little benchmarketeers in the past few weeks and have released a slew of test results that show the prowess their machines have supporting SAP's ERP software. Even IT shops that don't run SAP software take a hard look at the Sales & Distribution (SD) benchmark tests because a wide …
Ho-hum, more benchmarks.
".....The only plausible explanation for Big Blue not running the SD test on a fully loaded 32-socket, 256-core Power 795...." Well, I can think of several equally plausible reasons - maybe IBM is seeig most sales activity in this size of SAP implementation; or maybe it's just a sweet spot cost-wise for the AIX but doesn't necessarily mean it can't scale; or maybe they ran out of space for the racks and racks of short-stroked storage that IBM would have need for an even larger implementation.
".....That's about twice the performance of IBM's prior generation Power 595 machine...." Seeing as my experience is the old Power6 systems weren't (usually) CPU bound but eventually ran out of steam due to the memory space available and the bandwidth to storage, I'm tempted to ask if they couldn't have made the old P595 go faster with mroe memory and the same storage setup as used with the new Power7 P795 benchmark?
"....and it costs a lot less money than either an Oracle/Fujitsu Sparc Enterprise M9000 or an IBM Power 780 or 795...." Welcome to Understatement City! However, if you want to compare traditional UNIX servers to clustered Wintel then you have to factor in the additional costs of system administration (more servers = more work), added networking and SAN requirements, and then the fun of building DR into your Wintel cluster. I have also yet to see a high-end UNIX solution from any of the vendors mentioned go in for anything near list price, so expect heavy discounting that might not be matched by thinner-margin Wintel.
I'm also curious as to how performance would compare for the Wintel solutions against Lintel ones using RedHat or SuSE, and do Oracle not want to talk about Oracle Linux now that they want to make some Slowaris licence money?
And finally, the usually caveats buried away in the vendor small print - a carefully crafted benchmark done in a controlled environemnt has just about zero rellevance to real World use, so insist the vendor PROVES that the kit can do what you want in your environment, with your OS and app stack, your SAN and LAN. If this wasn't the case then none of the vendors would have so many disclaimers in all their benchmark press releases.
Curiosity killed the Cat and Satisfaction Brought Her Back for More Bits and Creative Bytes :-)
Fujitsu might despair at the unhealthy and unwelcome reticence at Shared Processing Quantum Leaps Forwards into Fab Futures.
Such is a Fertile Virgin Space for Novel Virtual Partnerships ...... AIMagical Mystery Tours a la Turing BetaTest Master Pilot Class Course Source is an Enigma Module Program.
.... for PerlyGatesPython Purple Prose CodeXSSXXXX Revelation of Sublime Control Universal Devices. ..... with all Requisite Provision Streams, Systems Supplied. ....... for a NEUKlearer HyperRadioProActive Source Intel Share.
Slow And Painful
Wouldn't describe myself as an expert, but have been working with the product for a few years now.
SAP has a lot of different configuration options - and these will depend upon a wide variety of factors (modules used, quantity of items being processed, database set-up, application stack etc). Even if you throw massive amounts of hardware at it, the system can run like a three legged dog if the configuration is wrong.
As Matt identifies, benchmarks have little relevance to the real world - I would suggest that they are the IT vendors equivalent of a comfort blanket. Benchmarks have their value, but need to be absorbed with a large amount of seasoning of your choice.
"As Matt identifies, benchmarks have little relevance to the real world - I would suggest that they are the IT vendors equivalent of a comfort blanket. "
Tony S. If you work with SAP you should also know that the SAPS sizing numbers you get from the vendors are actually based upon the SAP STD 2 Tier benchmarks that are run.
Sure in the 'we wanna sell you something' they'll use benchmark values, and in the 'real sizing' phase they'll use benchmark values - some%. But it still builds upon these benchmark numbers.
You're missing the point
Benchmarking isn't about business, it's the world's most expensive sport.
RE: You're missing the point
More expensive than Larry's yatchs?