1 |
Hi, |
2 |
|
3 |
On Tue, 26 Jan 2016 17:29:36 +0000 (UTC) James wrote: |
4 |
> > I have thoughts about caching NFS using filescached, but limited |
5 |
> > durability of the drive (400 TBW warranty for 512 GB size) restrains |
6 |
> > me here. Probably I'll use it for caching only in exceptional cases |
7 |
> > (e.g. slow remote mounts like AFS), but with 64 GB RAM I doubt I'll |
8 |
> > need additional NVMe-based caching, at least for now, with time |
9 |
> > this may change of course. |
10 |
> |
11 |
> Well, to be truthful, I was hoping your application was a speed boost |
12 |
> for clusters. Particularly a gentoo based cluster with some NVMe boards |
13 |
> on workstations that lend their excess power to a local cluster. Lots of |
14 |
> folks are building in house clusters where technical users have monster |
15 |
> workstations and use those excess workstation resources to boost the local |
16 |
> cluster. That's kinda my twist (lxqt on the desktop) for single (big |
17 |
> problem/data) on gentoo with mesos clusters. Do drop me a line, should that |
18 |
> type of usage permeate your thought_cycles..... |
19 |
|
20 |
We have clusters at work, but they have quite different hardware. |
21 |
For HA clusters we indeed use bcache on quite durable Intel SSD |
22 |
(400 GB size, 8PBW resource). For HPC cluster we planned SSD cache |
23 |
for storage, but due to funding cut-off we have only small SSDs on |
24 |
each node. |
25 |
|
26 |
Best regards, |
27 |
Andrew Savchenko |