Strange question. Its really a question about block storage in general than FC in particular, but I have a sense that FC may be the first to go down as it were.
Ok … I’ve been looking up mechanisms to help customers in a media editing environment. Their preferred file system depends, to a degree, upon IP over FC for connectivity. They need to interconnect Mac OSX machines, Linux and Windows machines to the same storage resources.
Well we can do that easily. Nothing fancy* needed.
The preferred solution for the end user is with an FC switch, and running IP over it. Which, as I discovered, was removed from the Linux kernel a while ago. And I understand why. FCoE, which has largely failed to gather anything like a significant following is a better concept for converged networks. You have an IP network and can use it to route FC packets. And you can interconnect these IP networks. FC has limits. Which render it less useful.
On the block side, SAS is a bit faster (in aggregate, 4x 6g lanes is faster than 1x 8g or 1x16g lane … with the 4x SAS 12g now available), and simpler.
So its losing its primary (block) reasons for existing. Its lost lots of its secondary reasons as well (large SAN interconnections). It never really had IP as a serious use case, but from what my research is showing, its also rapidly falling by the wayside.
More interesting to me is that the latest Mac pro desktop unit, that cylindrical thing that looks like a work of art … has 6 thunderbolt ports. And no place to plug expansion cards.
You can read into that, and you should.
Future IO for the Mac is over thunderbolt. Which begs the question as to what one should do about FC or 10/40 GbE on the mac. As I noted, I found a few thunderbolt to 10GbE converters. I found one to FC4. I’ve not yet found a thunderbolt switch (it should be basically a PCIe switch).
Linux and windows, no problem, we can serve files to these very quickly over NFS and SMB on a fast 10/40 GbE network. Even over IB. Bring Mac into the picture and you are basically constrained to go over 10GbE or FC, and the FC side has very near term sell by date.
I don’t see IP over SAS. I do see SCSI over IP, thats iSCSI. There are even high performance SCSI over RDMA fabrics with SRP and iSER.
Longer term, it looks like FC has been largely relegated to legacy tech. Nothing wrong with this, it happens to the best of tech at some point in time or the other. Anyone remember Winchester hard disks?
Better tech emerges and takes over. This is the way of advancement.
And that gets to the question on block storage as well. Object storage has been around a while and never really caught on much with the block and file crowd. But now, the biggest baddest storage bits out there are all object based. And with a little work, you can map objects to be a bag o blocks. Actually that mapping isn’t hard. So I guess the question I have is, when (not if) will we see the world migrate over to massive distributed pools of object storage with block shims for those who need them, and as API based objects for those who don’t? I can’t see FC playing in that world at all. Which is in part why I don’t think it has much of a future. As my research had suggested.
* well, as I’ve been discovering, support for Infiniband on MacOSX has all but disappeared. There are 10GbE drivers for a few cards, but no 40GbE that I am aware of. And a limited and declining number of FC card vendors and card options, with a declining number of drivers for MacOSX. There are thunderbolt to 10GbE and I found to one FC4g, but … well … why?
Viewed 72274 times by 6438 viewers