> >It is possible that IBM / Transarc AFS may die out before
> >Arla servers arrive. E.g. Intel's computing support people
> >tell me that the only servers we have that support AFS are
> >being EOL'ed. As it is now, it costs me 2,500$/GB to buy disk
> >space on our AFS capable servers, as compared to, what, 230$ US
> >for 60GB on a PC.
I am requested to emphasize that I am not making statements for
Intel overall, that Intel is not EOL'ing AFS overall, and that
Intel has many computing groups with different plans, and that
the computing support people I have talked to may not accurately
know even their own groups plans.
My expression of frustration as a customer who would like to
see continued use and expansion of wide area secure filesystems
like AFS remains accurate.
---
As for disk price, I have been told that 318$/GB is possible on
some other AFS servers using industry leading performance disk
systems. That's a lot better, but still 100X worse than
PC disk prices, and unfortunately may not be available to me.
The best price reported to this group, 859$/50GB,
is only 5X worse than PC disk prices. I could live with 5X.
Do I blame AFS? Well, yes I do: if AFS servers were available at reasonable
cost for LINUX, I could consider running them on "PCs stuffed full of
disks",
at 1/5th the disk cost of the best prices reported to this list.
Because of AFS port status and pricing, it's restrictedto running
on relatively expensive server disks (and/or the AFS licence cost
overwhelms the disk price cost).
Yes, yes, I know that there are lots of other factors in disks:
performance, reliability, power consumption, noise, cost of ownership.
Wrt performance, my group of users needs lots of capacity, less
disk performance, less reliability, and not so much network performance.
We only want occasional access across the net from other sites,
but prefer AFS and ACLs to having to arrange FTP transfers, accounts, etc.
Because of the disparity between server disk prices and PC disk prices,
we have begun deemphasizing servers, and emphasizing local PC disks
more and more. Backups??? IT doesn't backup PC disks across the network,
so more and more we run without a net, with only occasional copies to
server disks for IT toput stuff on tape. Scares me shitless, but it's
the way the world is going - there's even an editorial about it in Storage
magazine. Because server disks are so expensive, more and more data
lives on unbacked up PC disks. IMHO this reflects a fundamanental failure
on the part of IT.
Anyway, given this trend to local disks rather than server disks,
I would like to make all the local disks globally accessible on AFS.
But AFS servers for large numbers of low end machines don't seem
to be a recommended way.
Or, we are considering creating low end servers - PCs stuffed
as full as possible with PC disks, running AFS. It's not a nice as
a high end IT managed fileserver, but it's a hell of a lot more affordable.
Again, need AFS servers running on relatively low end systems.
Running LINUX, because we aren't going to bother with anything else.
But, let me emphasize, this is dreaming. Realistically, AFS seems not to
be able to play in our plans. So, we look for these "PCs stuffed full of
disks"
servers to run NTFS/Samba/NFS, and we give up on all of the advantages
of AFS. Too bad.
The worst part of talking like this in this forum is that I know that
corporate IT members read it, and they may go out and kill the plans
I describe. Turf wars. It's happened before: IT people killing LINUX in
favour of NT, only to rediscover LINUX 2 years later. Probably I shouldn't
post this.
But, the truth remains: I'm a user. I want to use AFS for all my data.
Costs prevent me. And IT isn't helping.
And now that it looks like AFS is dying, the non-AFS alternatives begin to
look more and more attractive. Which will create a vicious circle.
