ESIX Networks and X

Bob Palowoda palowoda at megatest.UUCP
Mon Dec 4 23:30:42 AEST 1989



>From article <1989Dec3.044931.5780 at NCoast.ORG>, by mikes at NCoast.ORG (Mike Squires):
> In article <13366 at s.ms.uky.edu> jpenny at ms.uky.edu (Jim Penny) writes:
>>
>>Does anyone have experience with this (relatively recent) implementation?
>>In particular, is anyone using it with RFS?  What hardware is supported
>>wrt network cards?  Does it seem solid?
>>					  Thanks
>>					  Jim Penny
> 
		     It's 415-623-8809 and 415-623-8806.
		     The '8806' number has a TB+ on it.
 
                       |||||||||||||
		       VVVVVVVVVVVVV
> The "fiver" BBS at 415 562 8809/8806 runs ESIX.  The Sysop reports problems
> with X (slow) but otherwise he likes it.  I don't know if he's tested the
> networking.

  Well actually I have been running the later Rev ESIX and the X-windows 
is faster. However I beleive my system configuration is incorrect to
test the performance of X.
First let me give you an idea what I'm running on:
----------------------------------------------------------------
CT Chip set 25Mhz (2 meg interleave) (6meg 16bit expansion mem)
                  8 Mhz io bus.
No coprocessor
Two Toshiba ESDI drives with Adaptec Controller.
Vega VGA card  (One of the first slow bit chip sets to come out)
Two dumb IO serial ports.
------------------------------------------------------------------

 A bunch of other stuff but I don't think it's relevant to the performance
of the X-windows. First I don't have a good way of testing the performance
of X and compareing it with other systems. I have heard of some X benchmark
test that I would like to get a hold of. In the ESIX Rev C they have changed
the scrolling method within the screens and it is faster. What's interesting
is that the disk activeity from the Cnews program slows the scrolling down.
What I would like to know how screens are drawn in X with VGA cards and
what factors will change this.  The ESIX X now supports the Vega Vram cards,
Paridise 16 bit VGA cards, Orchid's 16bit chards, and Tatung 16 bit VGA 
and Everex's VGA cards. It also supports 800x600 and 1024x764 resoulutions
on the fast 16bit cards. Now I've seen it on the Everex Step machines with
their VGA cards and it's fast. But there machines do have an advantage over
mine. If your going to do this don't configure your machine the way I have.
Get a fast cache machine with a fast video card etc etc. 

 As for the networking I'm useing NFS. It's Wollogangs and it's sold as 
an addition to the ESIX package. It dosn't have support for yellow pages,
but it does work with PC-NFS. I'm running it connected to a 386SX machine
and run QEEMS to get a little aditional memory when running large programs
on the dos machine. This combination works with Ventura Publisher. As long
as the unix filename where something that the vp could understand. I also
tried running MS-windows along with PC-NFS. You have to remove the QEMMS
for MS-windows as they will not co-exsist. Windows workds ok accross the
net. You have to convert the files to a dos format and make sure the filename
is something that dos can understand. I found it's easier to open a window
in one write session and just transfer a cut over to another write window.

 I'm useing Western Digitals WD8003E eithernet cards. And it also works 
with Everex's eithernet cards along with 3Comms. In my opinion the ESIX
package with NFS and PC-NFS is cost effective and would be productive in
an office enviorment. I think ISC or SCO is going to have a hard time
to justify the cost in this arena.

 As for bugs I haven't found any show stoppers. Couple of things I noticed
no 'newaliases' in the sendmail package. It could be hidden somewhere but
I haven't found it yet. I'm not useing sendmail anyways. I've heard that
there is a problem running sendmail and TB's. I have a TB and get news
every night and it dosn't seem to affect my performance. The only time
I had a problem is when I was running some io benchmarks on the disk drives,
when was doing 5 20meg benchmarks on the disks it slowed the system down 
and respawned a new sendmail. In the manuals it says there is support 
for Imake, but I haven't found it. Would be nice for compiling X stuff that
comes with Imake files. 

  I like ESIX manuals better than the Prentice Hall books. Three ring type
binders. Looks like they where done with Jetroff. They have done a better
job on the installation notes than the previous version. 

 There is also support for SCSI drives. I beleive with the adaptec controller.
I will probably stick with  ESDI and go with a cacheing controller when they
become more reasonable in price. Rumor has it rev D will have the fast file
system.  About benchmarks, I ran it on a Xenix 386 machine running 33mhz
with ESDI drives (and the latest 386 Xenix OS) and my ESIX system with
a similiar drive/controller combination. The Xenix machine turned out
to be only 20k/bytes faster on the writes and slower on the reads (by 30k). 
This was done without tuneing on either system.  

All in all I'm quite happy with what Everex did with ESIX. These are
my opinions. I'm sure there are other out there that will disagree with me.
I'm sure I'll have more to say about X after I make some changes in my 
system. (And afford it, expensive little hobby here.)

 Has anyone tried the Cyrix math coprocessor on unix? If so have you run
benchmarks on it. Speaking of math coprocessor's, I took my 80287 out
of my 386 board because after running some benchmarks I found out
it slows auto types, function calls, multiply's ????, and reg int's.
The only thing it really helped was double's as far as I can see.
Amazeing, X felt faster after I removed the 287 coprocessor. 

 I would also like to here any comments on how ddx interacts with the 
VGA card and what factors improves the screen updates.

---Bob

-- 
 Bob Palowoda    *Home of Fiver BBS*                   login: bbs               
 Work: {sun,decwrl,pyramid}!megatest!palowoda                           
 Home: {sun}ys2!fiver!palowoda   (A XBBS System)       2-lines   
 BBS:  (415)623-8809 2400/1200 (415)623-8806 1200/2400/9600/19200



More information about the Comp.unix.questions mailing list