So the background to this question is I’ve recently worn through the write endurance of all of my SSDs as swap/scratch disks and I’m now looking for an alternative solution.
Is there a way to have multiple blade nodes, whereby the amount of RAM per blade node is maxed out, and then I can either create a ramfs or tmpfs on it (RAM drive), and then export said RAM drive using pNFS or some other distributed/parallel file system/format so that to the cluster, it would appear as though it was one giant tmpfs/ramfs RAM drive?
If I understand question correctly… You can mount a RAMFS/TMPFS and it becomes a block device, once it is a block device it’s like having a hard disk or ssd mounted. You can even set it up to mount at boot if need be.
So jumping out on a limb here, i say yes.
I’m wondering if @wendall has ever used it before – either pNFS or GlusterFS/Ganeasha whereby I can take 64 GB out of 128 GB of RAM, create a tmpfs RAM drive, and then add it to the pNFS/GlusterFS-Ganeasha network storage pool so that it will show up as a 256 GB “drive” (a.k.a. mount point) comprised of four 64 GB RAM drives from each of my four nodes in my micro cluster (tied together using Mellanox ConnectX-4 dual port 100 Gbps 4x EDR Infiniband network adapters via an externally managed Mellanox 4x EDR IB switch), so bandwidth shouldn’t be much of an issue.
And then the idea with it being pNFS is that hopefully, it would be pNFSoRDMA in order to take advantage of the RAM drive speeds being that they’re RAM drives.
This is the idea.
So I’m looking at perhaps Linux experts here if anybody ever done something like this before.
Me either. In fact, the only context I’ve seen pnfs in is ESXi datastores, but I’m not even sure if it’s supported by repo versions of nfs, or at least I haven’t seen it being used.
This would be my suggestion, although I don’t know if there are any gotchas with the RAM disk (other than volatility).
