Path: utzoo!attcan!uunet!decwrl!ucbvax!SYSTEM-M.PHX.BULL.COM!Beattie
From: Beattie@SYSTEM-M.PHX.BULL.COM (Art Beattie)
Newsgroups: comp.protocols.appletalk
Subject: Re: KFatPath: Reliability? (atalk [0007])
Message-ID: <900712072314.754849@SYSTEM-M.PHX.BULL.COM>
Date: 12 Jul 90 07:23:00 GMT
Sender: usenet@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 65

There are a number of questions I could ask about your problems with the
FastPaths (now supported by Shiva); Using FastPath-4s?  Phase1 or phase
2?  What were the symptoms of your problems?

We have 10 FastPath-4 units that are online.  We only use the PROM code,
ie, AppleTalk Phase 1.  We have no need to concern ourselves (yet!) with
supporting a lot of Macs/PCs on our Ethernet backbone.

All of our approximately 500 users except the 10 Ethernet connected Macs
are connected to our AppleTalk network with Localtalk.  Each FastPath
creates a separate zone on the network so on average we have about 50
connections per zone.  In each zone we first connect the users using
backbone telco wiring (no more than 10-12 RJ11 jacks on each port) to a
Farallon StarController.  The StarControllers are then connected
together and then to the FastPath.

The only problems we have seen with them is keeping their internal
battery charged and heavy network activity.

1) The initial FastPath-4s had a design problem in the internal cabling
   of the unit which didn't keep the internal battery charged which kept
   power to the RAMs that hold the configuration data.  They ended up
   coming out with 2 changes to fix this problem.  When the unit failed
   and was accessed by "FastPath Manager", it looked like the unit had
   just come out of the box, no configuration data, no net numbers, no
   zone names, etc.

   The sticker on the new unit I got (last week) is at Rev J, with
   ECN2506.  This sticker is on the bottom and displays the current
   status of the unit along with the enthernet address.

2) The heavy network activity problem has not been resolved.  I had
   talked to someone at Kinetics about 2 months before Shiva took over
   the product (June 1 of this year, I think) and they said that they
   where working on a PROM change for the problem.  I asked Shiva about
   it last week and they had not heard about the problem or the PROM
   change.  All our units are running with revision 4.1 PROMs.  This
   problem only occurs with whatever unit that defines one of our zones.
   We know we are overloading this particular zone as it has 5 Sitka
   (alias TOPS) InBox mail servers, 2 AppleShare servers and several
   print spoolers for printers in other zones (this is not
   recommended!).  However, the problem is that the FastPath slows way
   down to a crawl which seems unnecessary.  It should just plateau at
   whatever rate it can handle.

I had noticed after we had our units for a while that they ran hot.
Some of our units are operating in wiring closets that are not well
ventilated but seem to be working OK.  I have heard that Kinetics
claimed that the components were operating within their specified
operating temperature limits.  Shiva claims that some are operating 15
degrees overtemp.  The committee is still out on this problem.

One big reason for staying with the PROM code and AppleTalk Phase1 is
that when power failures occur (our monsoon season, about 2 months long,
involves thunder storms that our power utilities cannot deal with), the
units have to reloaded with the KStar software.  I don't like the idea
of getting into the plant just to load up 10 units in the middle of the
night.  They supposedly have remote boot capability but we don't have
the resources to support that functionality.  If the KStar software were
put in PROMs, then maybe I would consider it.  I hear the Shiva is also
considering it.

Art Beattie
Bull HN Information Systems
Phoenix, AZ