[Hcoop-discuss] Server pricing

[Adam Chlipala]

Justin S. Leitgeb wrote:

>I guess I'm having trouble answering the question because I think it 
>depends on the type of file that is shared.  Surely we are going to 
>approach resolv.conf differently than html pages, right?  What files do 
>you think should have a global namespace, and how are these going to 
>play into the rest of our system architecture?
>  
>
In my current vision, there are two kinds of files in the shared filesystem:

1) Files whose primary logical location is in the shared filesystem.  
This includes HTML pages.
2) Files whose primary logical location is on a local filesystem, but 
that we periodically synchronize to the shared filesystem for ease of 
centralized back-ups.  This includes resolv.conf.

>We could buy a scaleable system like you're suggesting above, but:
>
>1) It would be expensive.
>2) It's not necessary right now, in terms of space.
>3) We would outgrow it at some point and need to buy a completely new server
>  
>
I think it's entirely clear that we have current and future members who 
would love to be able to use arbitrarily much disk space.  Keep in mind 
that many of us now are purposely avoiding using HCoop for purposes that 
would incur significant disk usage; it's not that we're all only 
interested in low disk usage services, but rather that we impose low 
disk quotas ATM.  Can you elaborate on your point 2 in light of this?

Can it really be that expensive to have a slow fileserver for which it's 
relatively easy to add new disks?  I'm fine with bad performance as long 
as we can get commercial-level reliability and protection against data loss.

>If we are really just talking about a shared namespace for the 
>user-level files, perhaps we can do the following:  make the public web 
>server be both an AFS server and client.  Then, just plan on adding a 
>separate fileserver later to join the AFS cell.
>
>Since you know AFS better than I do, do you think this would be possible?
>  
>
I don't know very much about AFS.  I only used it a lot and heard in 
operating systems class that it has significant technical advantage over 
NFS.

>Certainly, it's not for something like what we're doing with apache 
>right now.  But it would work for sendmail, DNS, hosts files, ntp...  
>Basically, while we have specialized needs for some services, there is a 
>lot that is "centralized", or at least able to be controlled by a small 
>group of admins in our network.  It seems to me that in those places of 
>our environment we could benefit from something like cfengine, eventually.
>  
>
DNS and Exim (no sendmail for us!) are also member-controlled for us, 
but at semantically shallower levels than Apache configuration.  I agree 
that admin tools like you're suggesting could be very helpful, though.