Re: Diskless boot: Anybody doing this big time?

From: Glynn Clements (glynn.clements@virgin.net)
Date: Sat Apr 28 2001 - 23:22:35 EST


James A. Crippen wrote:

> > Anybody here booting more than 200 machines from the same server?
> >
> > If yes, then I would like to know something about your experience regarding
> > boot time. There seems to be a lot of collisions, slowing things down.
> > What would be the average boot time for such a network size given ... let's
> > say 100Mb net and no other traffic but the boot?
>
> The primary consideration is how many machines are booting in parallel and
> how many are booting in serial fashion.
>
> In serial fashion it's a simple calculation based on the average time for
> a single boot, the bandwidth of the network, and the latency of the
> network from boot server to booting client.
>
> In parallel you'll have to account for the amount of bandwidth being used
> by each client host and the latency increase incurred by the increased
> bandwidth utilization.
>
> And you have to consider the amount of resources consumed on the boot
> server for each booting process and what its theoretical concurrent
> maximum would be. This is dependent entirely on the performance and
> design of the server and network.
>
> Collisions aren't necessarily bad, but a large number of them are. That
> may mean it's time to look into some network reconfiguration.

Booting diskless workstations can be a pathological case. Each client
will presumably be downloading a kernel using TFTP, which tends to
behave poorly in the presence of congestion.

I suspect that you might be better off booting blocks of clients in
several "waves". Consider conducting some experiments to measure
performance as a function of the number of hosts booting
simultaneously.

-- 
Glynn Clements <glynn.clements@virgin.net>
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org



This archive was generated by hypermail 2b29 : Mon Apr 30 2001 - 21:00:29 EST