🚀 go-pugleaf

I have heard the same story so many times now that I think I can recite it in
my sleep.  To get a file server running at maximum speed, you need to eliminate
bottlenecks.  The way I understand it follows:

1)  First of all, you need a quick machine.  A Mac Plus creates a bottleneck
    because of slow processor speed.  It can't process requests as fast as they
    can come in.  A Mac SE improves on this, and the Mac II and above basically
    eliminate this request queue (with some excpetions: see point 2).
2)  The second bottleneck is the hard disk.  You need fast disk access.  I
    think that 20ms or faster (with a 1:1 interleave running on some machine
    that can handle this) would do the job.  You do not want the machine to
    have to sit there and wait for an I/O request to be completed, while
    letting other service requests just pile up in the queue.
3)  The third bottleneck is the network:  a 16MHz Mac (for the most part) can
    send out packets faster than AppleTalk will accept them, causing the
    server to sit and wait until it can send out another packet.  EtherTalk
    reduces this bottleneck.
4)  Assuming that you have gone to EtherTalk, you have to look back to the
    CPU and hard disk as bottlenecks.  You would probably want a 030 machine
    as a server.  And that machine should have at least 2MB of memory for
    disk caching, cutting down response time, since most of the time there
    will be multiple requests for the same data.
5)  Now the EtherTalk interface card becomes a bottleneck.  You would want a
    card that has sufficient buffering (without the processor having to worry
    about it).  And you would probably not want a card that is limited by the
    NuBus; the SE/30 with an EtherTalk card that can directly access the CPU
    should help out here.  (The SE/30 also eliminates the NuBus interrupts,
    causing the machine to move faster as well--in theory.)

So that is how I see it.  I don't consider myself an expert at increasing
performance of networks, but from various lectures that I have listened to
from Apple system engineers, this in the information that I have gathered.

I welcome any responses to this posting.  Please send them directly to me
and I will summarize and post the responses (unless you think your comment
is too important to wait).

-Michael



-- 
Michael Niehaus        UUCP: <backbones>!{iuvax,pur-ee}!bsu-cs!mithomas
Apple Student Rep      ARPA:  mithomas@bsu-cs.bsu.edu
Ball State University  AppleLink: ST0374 (from UUCP: st0374@applelink.apple.com)