Is Tiering Dying?
by Michael Hay on Feb 23, 2010
According to NetApp’s CEO it is. Well that is at least what he is cited as saying within NetApp’s most recent earnings conference call. (Here is an article at the Register, and here are the Google search results for this topic.) The rationale that NetApp’s CEO uses is that with the advent of SSD/Flash combined with S-ATA and controlled by the NetApp PAM (Performance Acceleration Module) tiering is obsolete. Effectively SSD/Flash becomes the cache of the system and PAM is really a Flash aware set of caching algorithms. Wait just a second here, so PAM moves file system blocks between the Flash and S-ATA media just like a Compellent does with their chunklets, just like EMC is claiming to do with FAST v2, just like modern HSM systems do with files between disk, MAID and tape/optical, just like Hitachi does with files from HNAS to HCP/NetApp, etc. Frankly, I think that this message seems well, strange. To me what I understand after reading the transcript — and I read the raw unedited transcript — is that NetApp thinks that tiering is dead, replaced by caching algorithms but which is basically a form of tiering of file system blocks from Flash to S-ATA and back again. Huh? Does that say “Tiering is dead, long live tiering,” or am I misreading something?
As to what I think of tiering, well I certainly don’t think it is dead. I do believe that we will be seeing some consolidation and combination effect between file and block tiering. Hitachi is already able to combine file and block tiering at the LUN level today orchestrated by HDDS and HTSM on the USP-V. As with all things Hitachi you should expect this to evolve over time, but I will not speculate on what else is in store for our future in my blog. However, I will say is that EMC gets that tiering needs to be expanded and not set out to pasture. They have debuted FAST as a concept that includes LUN tiering today, just like HTSM and USP-V, file tiering from Celerra to another Celerra, a Centera or an Atmos(t). This is very similar to how we do it with HNAS to HCP or HNAS to NetApp or HNAS to HNAS or or or, with the major difference being that the data migration engine is intrinsic to the NAS device and not an external appliance. (Note that I cover the differences in the RAINFinity + Celerra approach versus the HNAS and HCP approach here.)
So while we are staunch competitors and have different approaches I will say, again, that EMC is getting tiering even though they are new to the game. However I wonder if perhaps it is not tiering that will be put out to pasture, but NetApp instead. (While I know that their most recent financial results came out pretty solid, I do have to think that they just don’t get that their architecture is slowly going to strangle them. I point out the basic flaw in the NetApp architecture in this post.)
Comments (2 )
[...] Computing, Featured — By Chris Evans on February 23, 2010 at 3:44 PM I’m not the first to post on the subject of Netapp’s President and CEO Tom Georgens commenting during their [...]
[...] I also like the post by my good friend Michael Hay. [...]