Server crash [WAS: Re: [WinMac] Mac DHCP client <-> NT DHCP Server follow-up]

by way of Marc Bizer(expresso[at]snip.net)
Sun, 25 Apr 1999 12:30:17 -0500

        Dear Michael,

        ANY server can abend, crash, become corrupted, yada yada yada, IF it is
not managed properly. At one location I had an Apple Workgroup Server
9150/120 running System 7.5.1 and AppleShare Server 4.1 for over 3 years
nonstop - Until the disk drive failed.

        Yes, NT servers can also crash, either with a Dr. Watson when an
application running in Ring 3 blows up, or a BSOD (Blue Screen Of Death)
when a process goes haywire. And just like on the Mac, NT can be easily
brought down by an errant driver or some odd piece of 3rd party software.
Generally, I stick to "plain vanilla" installations:

        * "Lean & mean" on the Mac side, with only the minimal number of
Extensions and Conntrol Panels needed to get the job done;

        * "Vanilla" M$ HAL's on NT/x86 boxes: Often an OEM (such as Dell or
Compaq) will write a custom HAL (Hardware Abstraction Layer) for their
boxes§; but later on these will be improved upon by Service Pack releases.
I also start out with as many vanilla "on the CD" drivers as possible, and
only afterward (after applying the Service Pack) start to apply _proven_
vendor drivers for such goodies as Symbios 53c8xx SCSI and Intel 82557
ethernet cards¤.

[§ NOTE: I'm excluding from this recommendation custom HAL's for boxes with
more than 4 CPU's, which are written by the OEM from the outset since M$
only supports machines in NT4 with up to 4 CPU's "out of the box;" but if
you're running an x86 box with more than 4 CPU's then chances are you need
to look more closely at other solutions (such as clustering) as well.]

[¤ NOTE: Although I'm not a big fan of intel's CPU's, which is over 80% of
their business, I DO like their ethernet cards - They are about the best in
the business in terms of drivers, CPU overhead, and reliability.]

        As for your Dell server going down, instead of first pointing
the finger
at NT itself (which it could very well be cause), you need to find out WHAT
caused it to crash: Was it a hardware-based error, such as a soft error in
one of the hundreds of millions of gates in the semiconductors in the box?
This would bring down ANY OS, whether it is OS/2, Linux, NT, NetWare, *nix,
etc... Was it indeed a software problem with an errant driver? Dell uses
(mostly) 3Com NIC's. Was it due to a Dell-supplied HAL? Was it due to a
Dell management utility? I've not worked on a production Dell server, so
I'm not familiar with their server management software. [I mention this
because H-P's NT utilities are notorious for being flaky.]

        Cheers!
        Dan

At 06:23 PM 4/24/99 -0500, BARTOSH, MICHAEL ALLAN wrote:
>
>Well we would know, wouldn't we if ms would open up the bottom of its OS.
>
>Until then, not only no but hell no, m$'s track record regarding
>standards is too blotched.
>
>Once again last night, as it came down to a heavy-load final weekend, the
>nt server at a local universitie's Architecture dept (very well supported
>mainstream Dell server, not a serverized workstation) crashed to pieces,
>48hrs before deadline. Was it the i/o? Was it the printing? Was it the
>networking? I don't know.
>
>Tisk tisk.... when will the world tire of this.
>
>-mab
[snip]

* Windows-MacOS Cooperation List *

This archive was generated by hypermail 2.0b2 on Sun Apr 25 1999 - 10:36:07 PDT