|"At the table in the kitchen, there were three bowls of porridge. Goldilocks was hungry. She tasted the porridge from the first bowl. |
"This porridge is too hot!" she exclaimed.
So, she tasted the porridge from the second bowl.
"This porridge is too cold," she said
So, she tasted the last bowl of porridge.
"Ahhh, this porridge is just right," she said happily and she ate it all up.
-- [Goldilocks and the Three Bears]
Continuing my week in Tokyo, Japan, I was going to title this post "Chunks, Extents and Grains", but decidedinstead to use the fairy tale above.
Fellow blogger BarryB from EMC, on his The Storage Anarchist blog, once again shows off his [PhotoShop talents], in his post [the laurel and hardy of thin provisioning]. This time, BarryB depicts fellow blogger and IBM master inventor, Barry Whyte, as Stan Laurel and fellow blogger Hu Yoshida from HDS as Oliver Hardy.
At stake is the comparison in various implementations of thin provisioning among the major storage vendors.On the "thick end", Hu presents his case for 42MB chunks on his post [When is Thin Provisioning Too Thin]. On the "thin end", IBMer BarryW presents the "fine-grained" details of Space-efficient Volumes (SEV), made available with the IBM System Storage SAN Volume Controller (SVC) v4.3, in his series of posts:
- [SVC 4.3.0 - Space-Efficient (thin provisioning), Mirroring and much more]
- [SVC 4.3.0 - SEV and SEFC in detail]
- [Morcombe and Wise]
BarryB paints both implementations as "extremes" in inefficiency. Some excerpts from his post:
"... Hitachi's "chubby" provisioning is probably more performance efficient with external storage than is the SVC's "thin" approach. But it is still horribly inefficient in context of capacity utilization.
... the "thin extent" size used by Symmetrix Virtual Provisioning is both larger than the largest that SVC uses, and (significantly) smaller than what Hitachi uses."
"free" may be the most expensive solution you can buy...
Before you rush off to put a bunch of SVCs running (free) SEV in front of your storage arrays, you might want to consider the performance implications of that choice. Likewise, for Hitachi's DP, you probably want to understand the impact on capacity utilization that DP will have. DP isn't free, and it isn't very space efficient, either."
BarryB would like you to think that since EMC has chosen an "extent" size between 257KB and 41MB it must therefore be the optimal setting, not too hot, and not too cold. As I mentioned last January in my post[DoesSize Really Matter for Performance?], EMC engineers had not yet decided what that extent size should be, andBarryB is noticeably vague on the current value.According to this [VMware whitepaper],the thin extent size is currently 768 KBin size. Future versions of the EMC Enginuity operating environment may change the thin extent size. (I am sure theEMC engineers are smarter and more decisive than BarryB would lead us to believe!)
BarryB is correct that any thin provisioning implementation is not "free", even though IBM's implementation is offeredat no additional charge. Some writes may be slowed downwaiting for additional storage to be allocated to satisfy the request, and some amount of storage must be set asideto hold the metadata directory to point to all these chunks, extents or grains. For the convenience of not havingto dynamically expand LUNs manually as more space is needed, you will pay both a performance and capacity "price".
However, as they say, the [proof of the pudding is in the eating], or perhaps I should say porridge in this case.Given that the DMX4 is slower than both HDS USP-V and IBM SVC, you won't see EMC publishing industry-standard[SPC benchmarks] using their"thin extent" implementation anytime soon. IBM allows a choice of grain size, from 32KB to 256KB, in an elegantdesign that keeps the metadata directory less than 0.1 to 0.5 percent overhead. I would be surprised if EMC canmake a case to be more efficient than that! The performance tests are stillbeing run, but what I have seen so far, people will be very pleased with the minimal impact from IBM SEV, an acceptable trade-off for improved utilization and reduced out-of-space conditions.
So if you are a client waiting for your EMC equipment to be fully depreciated so you can replace it for faster equipment from IBM or HDS, you can at least improveits performance and capacity utilization today by virtualizing it with IBM SAN Volume Controller.
technorati tags: Goldilocks, Three Bears, IBM, Tokyo, Japan, EMC, BarryB, PhotoShop, Barry Whyte, HDS, Hu Yoshida, USP-V, SVC, SEV, Stan Laurel, Oliver Hardy, Symmetrix, DMX4, metadata, directory, SPC, benchmarks