I'm enjoying Chuck's articles on pNFS; I've been following the development of pNFS for sometime but I guess that is no surprise to anybody who knows about my current interest in both Clustered NAS and Clustered FileSystems.
It's great to see what was a very minority interest a few years a go become more mainstream and I suspect that there is at least one EMC blogger who must be looking at pNFS developments which a certain feeling of nostalgia (I shan't name them, he can own to up to that himself). I had the joy of trying to make SANergy actually work at a customer site and I must admit I was somewhat surprised to find that it actually still appears to be an IBM product, I'd assumed it was dead and gone. I suspect developments really rather overtook it and it'll never become a mainstream product and there are plenty who might feel that is a good thing.
It's odd watching all these concepts from the past becoming more mainstream again; from the remarkable similarities between HPC and Cloud, the attempted re-invention of the mainframe to the more esoteric areas of storage-products-past come round again. I hoping however that as we re-implement concepts that we have learnt from some of the lessons in the past.
- Make it easy to deploy
- Make it easy to manage
- Make it easy to standardise
- Make it interoperate
- Make it robust
You’re right — for just about everything being done, there’s earlier work out there to be considered and learned from.
And pNFS is no exception.
— Chuck
But it’s certainly interesting watching some of the more esoteric areas of computing in the past becoming very mainstream. And I suspect ideas which were written off as simply impossible or just too hard to do well are going to start regaining new followers. Now, there are some of us who might just mutter ‘Don’t go there….’ but I think we might be surprised what actually works this time round.
Indeed, deja vu all over again.
SANergy was an early technologu success (and my first real branding success), before we sold it to IBM. From there it got all tangled up in StorageTank and other such San Jose experiments.
Truth be told, EMC’s MPFSI took much the same approach. It is good to see EMC putting the technology back into the standards community – the innovation truly delivers performance at an incredibly low overhead.
Couple pNFS with sub-LUN FAST, and you get a three-fer: file sharing simplicity, direct I/O performance and automated tiering at a granularity smaller than at the file level.
Hmmmm….
Another protocol over IP. Like we need another one after iSCSI and FCoE who barely have enterprise attention.
Strange Quantum StoreNext, Sun SAMQFS, Lustre, GPFS and few others never got attention in the past.
StorNext does a little more than the others by doing dedup, remote replication and IP protocol support. A big plus in my book.
The beauty of pNFS will be it’s various level of support (block, file, object) and it’s price tag. Free. Can’t beat free. That will affect storage market including FCoE, iSCSI, NFS and CIFS. pCIFS is coming also.