[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: High Availability..

On Fri, 6 Oct 2000, Gregory Wood wrote:

> Several years ago, I sudied the HP 9000 family. It had this ability.
> Their system had different programs running on different servers which
> minimized the file locking issues. When one computer failed, the other
> would start up the tasks from the failed machine. The switch over time
>was under a few seconds.
> There is a software package (I don't remember the name) that runs on
>  Novell and NT, that keeps a full duplicate of the primary server. If
> the primary fails for any reason, the backup comes on-line in under a
> second.

Well, the thing I had thought to do was having one central file server for
a whole bunch of other servers. This server has to be able to do keep
running even during hardware failures, and also scheduled uppgrades..
The packages you talk about are when you want a failsafe between say two
webbservers, but I'm talking about a pure network storage thing here.. 
There should be about 10 machines accessing this, and they must under no
circumstance have the disk mounted locally, as these are 'open' systems
and might be compromised.

As NFS *is* supposedly stateless, you should be able to switch the server
you are talking to between two nfs packets. But I'm not sure how this NFS
locking works? It really has to be implemented in the server, but how does
this comply with the "statelessness" ??

I know there are solutions which might fix my problem.. Symmetrix is one..
But we havent reached there quite yet, as they can be a bit costly.
As for Novell & NT, I wouldnt touch those without a very good reason (gun
pointing at my head would be one :)).

/Roger A

Roger Abrahamsson, Sys/Net Admin, Obbit AB
Radhusespl.17D, S-90328 Umea, Sweden

Reply to: