Marty Scholes wrote:
Bob Doolittle wrote:
It might be a bit before we can get to any deep-dives regarding
performance on OSol. We will get there and appreciate your patience.
Please test with 4.2!
I am going to revive this thread a little. Since I last wrote, I replace my feeble server with a more robust 4 way v40z last week.
I have CPU cycles, network bandwidth and memory to burn. We have noticed some
odd behavior with the DTUs.
Initially, things were not much better, the managed switches showed very low
network utilization on the DTU ports and server CPU utilization was nigh-zero.
After a few days of uptime, one or two of the heads on the triple head displays
started refreshing very quickly, and those ports on the switches showed much
higher, ~50%, link utilization.
Sometimes the heads would revert to slow refreshes. I haven't figured out the
pattern of when they will behave a certain way.
There appears to be two refresh modes:
1. Low bandwidth refresh which is slow and updates blocks across the screen top
to bottom and left to right within each row, taking more than a second. This
makes scrolling in a browser/PDF/OO Document/gnome-terminal/etc. a painful task.
2. High bandwidth refresh which is fast and consumes the ethernet link,
smoothly refreshing the whole screen in less than half a second.
I seem to recall that 4.1 ran in "high bandwidth" mode quite often but 4.2
prefers the slower mode. Changing the DTU to the 4.1 firmware had no effect, so I can
only assume it is SRSS which is making these decisions.
Assuming (naive, I know) that this is how SRSS operates, is there a way to coax it into
the "high bandwidth" mode permanently?
SRSS is certainly not designed to operate this way :)
I presume that the difference is due to changing loads on your server
and/or network.
The only low/high bandwidth issue that comes to mind is related to using
GB NICs on servers. Sometimes the server can overrun certain models of
switches which can cause behavior similar to what you describe.
It might be worth running utcapture and looking for packet loss.
-Bob
_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users
