[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Symbolics Service Issues
I have never understood the voodoo behind Symbolics hardware maintenance.
I have used many different machines (no Ivorys, mainly 3640s, 45s, and
G-machines), at several sites, with different Symbolics hardware technicians.
Still, with the exception of one site, hardware problems have the same pattern:
a machine will work fine for six to nine months and then encounter some
hardware problem. When the problem is reported to Symbolics, Symbolics will
promptly dispatch someone to examine the machine. Usually, however, it becomes
apparent that the tech has no clear idea of what is causing the problem or what
will fix the problem. So a long (several week to several month) process of
experimentation begins. Boards are swapped again and again and each time the
system is placed under observation for a few days. Eventually one of two
results occur: either the process converges and the problem is eliminated or the
user finds he can live the system. Nb: there is no guarantee the process
converges. Often during the course, one problem will be swapped for another.
Now this pattern should be eliminated! It is expensive and frustrating for
Symbolics and even more so for the customers.
How Symbolics can reduce hardware maintenance costs:
If you must build hardware that is so mysterious, you can reduce costs by
eliminating the pattern by admitting that it exists and when it is recognized
out in the field (by a problem machine which persists for a week), you can
immediately swap the board set for some sort of tested, verified board set.
It sounds drastic, but if it would keep problem resolutions down to a week,
I bet Symbolics would find it is cheaper to swap entire board sets at the
customer site and inspect and fix them back in the Symbolics labs then it is
to play and play and play with the boards out in the field.
And of course the side affect is that for many people, Symbolics hardware has
a very poor reputation in terms of quality and reliability. It doesn't matter
how fast it is, or what the software can do, it's not knowing when the machine
is going to die, or for how long it will be dead.
And the exceptional site: it wasn't so exceptional, it was a site which had
bought many machines under a contract which specified 99% uptime, and this
(and some other factors) had the effect that Symbolics supports this site with
a) two spare machines, and b) its own hardware technician.