Inside System Storage -- by Tony Pearson

Tony Pearson Tony Pearson is a Master Inventor and Senior IT Specialist for the IBM System Storage product line at the IBM Executive Briefing Center in Tucson Arizona, and featured contributor to IBM's developerWorks. In 2011, Tony celebrated his 25th year anniversary with IBM Storage on the same day as the IBM's Centennial. He is author of the Inside System Storage series of books. This blog is for the open exchange of ideas relating to storage and storage networking hardware, software and services. You can also follow him on Twitter @az990tony.
(Short URL for this blog: ibm.co/Pearson )

Comentários (5)

1 localhost comentou às Link de Retorno

Well, I still think you're mixing metaphors, comparing fruit flies with potato peelers. I explain my perspective over on my own blog (follow the link below).<div>&nbsp;</div> Thanks!

2 localhost comentou às Link de Retorno

BarryB,I enjoyed your response! And thank you for the kind words on my masterful use of English language.<div>&nbsp;</div> Perhaps cars were a bad source for analogy, as we seem to have merely carried over our differences of opinions from the storage world over to the car world. I have owned five cars (serially, of course, not all at the same time), and all of them got reasonably close to the mileage that was advertised for City MPG and Highway MPG. While I agree these will never achieve five-decimal-point accuracy to actual results, they were close enough to help with my purchase decisions in the past, and I plan to continue to use them for future car purchase decisions.<div>&nbsp;</div> And that's all I expect from SPC benchmarks, to be close enough to actual results of OLTP, database, and file transfer workloads to provide helpful guidance in purchase decisions.

3 localhost comentou às Link de Retorno

It's been fun...but you're probably the only person I've ever heard that claims to get "reasonably close" to the rated MPG. <div>&nbsp;</div> I'll go out on a limb and opine that our respective definitions of "reasonable" probably differ quite a bit (even without debating exactly how much better "uses less power" means). <div>&nbsp;</div> Me - I just assume that the MPG ratings don't really mean all that much, and if I get better than the rating I can be artificially happy, but if I get worse, I probably oughta lighten up on the gas pedal more often...but that's just me (FYI - no matter how hard I try, I can't get any better than 21.5MPG highway, even though my car is rated at 24MPG).<div>&nbsp;</div> Just curious - how are you going to rationalize the new EPA ratings when you buy your next car, adjusted downward 12% lower city, 8% lower highway (which will bring the ratings closer to my experience, but they'll probably be lower than yours)?<div>&nbsp;</div> And it's your blog, so of course you can keep making the assertion that there is some mystical relationship between SPC and actual results of OLTP, database and file transfer workloads. But so far, you've still presented no evidence of that assertion.<div>&nbsp;</div> As far as I can tell, SPC is no better an indicator of "good storage" than the LSAT is a predictor of a "good lawyer" (insert favorite lawyer joke here :&gt; )

4 localhost comentou às Link de Retorno

21.5 MPG sounds reasonably close to 24MPG. If you were deciding between a car that is 24MPG and one that is 12MPG, and you know you get only 21.5MPG with the first, you may assume you will get 10.25MPG with the second. The point is that these estimates don't have to be exact, just proportional and in the same relative order. Even if SPC benchmarks were off by 20% of what customers actually get, they would still be valuable, relevant and useful in decision making in this regard.<div>&nbsp;</div> I did point to the SPC website which explains why SPC-1 was reasonably close to certain set of OLTP workloads, and why SPC-2 was reasonably close to a different set of video transfer, file transfer and database query workloads, and tried to use the analogy of City MPG and Highway MPG to explain why they are different from each other, and why one benchmark may be closer to your workload than the other.<div>&nbsp;</div> For the new EPA ratings, I will most likely be comparing 2008 model cars with other 2008 model cars, so as long as these are measured in a standardized consistent manner, than the new ratings are valuable and relevant to my decision making process. Only if I was comparing new cars to used cars would it matter, and at that point, differences less than 8-12% would probably be over-ridden by other decision factors.<div>&nbsp;</div> And yes, as for lawyers, I am sure the ones with higher LSAT scores are better than the ones with lower LSAT scores, which is why these standardized tests are factored in the recruitment decisions of nearly every law firm.<div>&nbsp;</div> Like lawyers and cars, storage is not chosen solely on a single criteria, just that it should be factored in to the decision making process when performance-critical workloads are involved.<div>&nbsp;</div> Some day in the future, when enough people publish the results of their real-world performance, spelling out what application workloads they are running on what vendor gear they are running it on, and everyone about to make a purchase decision could find real-world performance results that appear to match what they are investigating, for their applications, for their level of workload, for the brands of storage they are interested in buying, then perhaps we wouldn't need standardized benchmarks at all. Until that day, there's SPC.<div>&nbsp;</div> <div>&nbsp;</div> <div>&nbsp;</div>

5 localhost comentou às Link permanente

Perhaps a cars 0-60mph time or 1/4 mile time are good analogies to IOPs. Adding repeated starts and stops isn't necessarily relevent. The 2 tests are standard benchmarks that measure a cars performance. Both are generally repeatable "best possible" results just like SPC-1 IOPS. Of course 0-60 is dependent on traction, and the 1/4 mile less so but it also adds drag as a factor.<div>&nbsp;</div> MB/sec could be like passengers transported/sec but that isn't a standard test. Performance per $ is occasionally reported but I don't recall the units. You would expect a Ferarri to be faster than a Civic but is it worth the price difference?

Incluir um Comentário Incluir um Comentário