| Pvwave on Vax Hangs on Network [message #6386] |
Wed, 19 June 1996 00:00 |
James Knopp
Messages: 1
Registered: June 1996
|
Junior Member |
|
|
I have a case of pvWave running on a Vax 6620 with VMS and DecWindows that
hangs randomly on output. Output is to a "X" display server over a local
network. Please bear with me folks -- this takes a bit of explaining.
The application collects around 200+ values for each of 512 parameters.
The output is arranged into x-y plots with 32 display "pages" each having 4
linear grids of 4 parameters with titling and some annotation on each grid.
Data points are plotted as symbols. Controls allow the engineer to page from
one display to another in succession. The Vax will be running 4 sessions
like this concurrently. There's plenty of memory and cpu load is low -- about
25% and less.
A hang can occur when changing pages. Perhaps the third grid and
annotation on the page is only partially done. Each page represents about 130
KBytes over the network. The transport and protocol can be either Tcp/Ip or
Decnet. For the display server I have used Macintosh with MacX/MacTcp, a
100 MHz pc with pc-TCP or pc-NFS, an HDS hardware X-Terminal, a Dec
VaxStation 3100 with DecWindows, and a Sun Sparcstation 5 with Solaris. The
faster platforms are more difficult to hang. A platform on a local network
segment (no routers) hangs more easily than a server more remote on the
network.
In each case we have a fast Vax with a slower display platform.
For debugging, to achieve a degree of synchronization, and to slow things
down a bit, I placed an EMPTY command after each output statement. There
is a mixture of PLOT, PLOTS, OPLOT, and XYOUTS commands all using /normal
coordinates. The hang will occur in the EMPTY statement -- you just never
return from it. Visual Numerics reports they use all Xlib calls -- no Xm
motif nor any Xt tookkit calls. In this case all poly-line and flush calls.
A network "sniffer" capturing Ethernet/Ip packet traffic will report
that a packet from the Vax was never acknowledged by the server. Aha !! the
problem must be at the server!! Right?? Does anyone think seven platforms,
different software vendors with two different protocols and a mix of
operating systems are all going to fail in the identical way? The problem
is at the Vax.
Has anyone run into problems like this with separate client and
server applications?
Jim Knopp
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
