Simpler Navigation for Servers and Operating Systems - Please Update Your Bookmarks
Completed: a much simpler Servers and Operating Systems section of the Community. We combined many of the older boards, so you won't have to click through so many levels to get at the information you need. Check the consolidated boards here as many sub-forums are now single boards.
If you have bookmarked forums or discussion boards in Servers and Operating Systems, we suggest you check and update them as needed.
Operating System - Tru64 Unix
cancel
Showing results for 
Search instead for 
Did you mean: 

Xwin performance of 5.1B-4 vs. 4.0F

Bengt Nilsson_2
Regular Advisor

Xwin performance of 5.1B-4 vs. 4.0F

Hi!

I have noticed that for two comparable machines (PWS500au), same "software content" except that one runs 4.0F and the other 5.1B-4, there is a dramatic difference in the remote Xwin performance.
To open a window showing some graphics, the 5.1B-4 takes 6 seconds to show up, while the same takes 1 second on the 4.0F system.
The two machines are sitting next to each other, connected to the same switch, and the display machine (a DS10 running 4.0F) also connected to the same switch.

Anyone have some experience of this?

BN
21 REPLIES
Rick Retterer
Respected Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

BN,

If you are displaying the graphics remotely, there could be a difference in how the network controllers are configured?

Is one only using 10mb -vs- 100mb on the other?

Just a thought.

Rick


- Rick Retterer



Hein van den Heuvel
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

I like the suggestion to look at the network.

try to time some FTP of NFS transfers and see whether the difference exists just for the display tasks, or also for other transfers.

ping -f (flood) might be interesting.

Coudl be a duplex failure, or TCP/IPC sysconfigtab setting difference.
For example tcpnodelack.

Check with

sysconfig -q tcp ( ipc, inet, vfs, ...)

Good luck,
Hein.




Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

Thanks for the suggestion, but no, they are both at 100MBit.
Actually, a similar difference can be seen locally also, but not as pronounced. One application is showing me a table with numbers and text, and it starts as a larger array of cells and then adapting to the current context. Locally, this process is hardly noticed on the 4.0F machine, but is clearly seen on the 5.1B-4 system. Remotely, the difference is dramatic, of course.
I have another 5.1B-4 system running on a DS10, here my 'testapp' window shows up after 3 seconds, still slower than the 1 second I have from the 4.0F system. All are connected to the same switch.

BN



Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

I have a feeling the reason is a much more intense communication between the application/X-call layer and the X server.
If I connect to the v4.0F and 5.1B-4 systems from home, where I have a 2MBit ADSL connection the difference becomes just ridiculous. When connecting to the 4.0F system there is some normal latency time before the windows comes up, while on the 5.1B-4 I can go and make myself a cup of coffee and come back before something happens. Instant coffee, but anyway.


Pieter 't Hart
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

suggestion: take a look at name-resolution
what names and namespaces are involved and how are nameservers or hosts file configured.
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

You mean DNS stuff? Not likely, everything else works, I see no such errors in the logs.
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

As a response to the suggestion on the network speed settings, here it is:

# hwmgr -get attr -cat network -a name -a media_speed -a full_duplex
20:
name = tu0
media_speed = 100
full_duplex = 1
48:
name = tu1
media_speed = 100
full_duplex = 0
#

tu0 is external network, tu1 is used for internal subsystem communication.

BN
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

Oops, forgot o mention that the 5.1B-4 PWS500au machine is controlling some lab hardware using VXI boxes, thereof the tu1 network interface and subsystems communication.

BN

Pieter 't Hart
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

can you specify the network config (ifconfig -a) from all systems involved?

could be the answer from the v5.1b machine is primarely returned on the wrong tu and only in second instance sent to the right interface.
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

The 5.1B-4 PWS500au machine:
(mc2-p009)/ ifconfig -a
lo0: flags=100c89
inet 127.0.0.1 netmask ff000000 ipmtu 4096

sl0: flags=10

tu0: flags=c63
inet 129.16.138.202 netmask ffff0000 broadcast 129.16.255.255 ipmtu 1500

tu1: flags=c63
inet 192.168.27.1 netmask ffffff00 broadcast 192.168.27.255 ipmtu 1500

The 4.0F PWS500au 'clone':
ifconfig -a
tu0: flags=c63
inet 129.16.138.198 netmask ffff0000 broadcast 129.16.255.255 ipmtu 1500

tu1: flags=c63
inet 192.168.27.1 netmask ffffff00 broadcast 192.168.27.255 ipmtu 1500

sl0: flags=10

lo0: flags=100c89
inet 127.0.0.1 netmask ff000000 ipmtu 4096

The 'remote display' workstation, a DS10:
ifconfig -a
tu0: flags=c63
inet 129.16.138.199 netmask ffff0000 broadcast 129.16.255.255 ipmtu 1500

tu1: flags=c63

sl0: flags=10

lo0: flags=100c89
inet 127.0.0.1 netmask ff000000 ipmtu 4096


BN
Pieter 't Hart
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

thanks for quick respons
but see no clue there.

please post output from "netstat -rn" and "grep hosts /etc/nsswitch.conf" on these systems
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

netstat -rn on the three systems:

PWS500au system with 5.2B-4
Routing tables
Destination Gateway Flags Refs Use Interface

Route Tree for Protocol Family 2:
default 129.16.1.4 UGS 1 29656 tu0
127.0.0.1 127.0.0.1 UHL 4 107346 lo0
129.16/16 129.16.138.202 U 5 695751 tu0
129.16.138.202 129.16.138.202 UHL 1 896452 tu0
192.168.27/24 192.168.27.1 U 12 3103365 tu1
192.168.27.1 192.168.27.1 UHL 0 0 tu1


PWS500au system with 4.0F
Routing tables
Destination Gateway Flags Refs Use Interface
Netmasks:
Inet 0.0.0.0
Inet 255.255.0.0
Inet 255.255.255.0

Route Tree for Protocol Family 2:
default 129.16.1.4 UGS 0 0 tu0
127.0.0.1 127.0.0.1 UH 7 1380331 lo0
129.16/16 129.16.138.198 U 2 62575 tu0
192.168.27/24 192.168.27.1 U 9 571789 tu1


DS10 system with X-display
Routing tables
Destination Gateway Flags Refs Use Interface
Netmasks:
Inet 0.0.0.0
Inet 255.255.0.0

Route Tree for Protocol Family 2:
default 129.16.1.4 UGS 0 58077 tu0
127.0.0.1 127.0.0.1 UH 7 1089921 lo0
129.16/16 129.16.138.199 U 62 6084660 tu0

BN
Rob Leadbeater
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

Hi,

I think this output suggests a speed/duplex mismatch, as has previously been suggested, especially if everything is connected to the same switch.

20:
name = tu0
media_speed = 100
full_duplex = 1
48:
name = tu1
media_speed = 100
full_duplex = 0
#

What sort of switch are you connected to, and what are its speed/duplex settings ?

If it's an unmanaged switch, then you'll probably be best setting the NICs to auto-negotiate... If it's a managed switch, then the switch end MUST match the NIC... The output of netstat -i might indicate errors to confirm this.

Cheers,

Rob
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

tu0 is set to auto-negotiate and it is connected to a HP2424 switch.
tu1 is set to 100Mb fixed half-duplex and is connected to a small hub which distributes the communication to three VXI sub-computers. This setting is prescribed by the equipment vendor.

Same setup vas used on the 4.0F machine.

The reason for this discussion is that we acquired the equipment with the 4.0F system about ten years ago, and we recently got the vendor support for an upgrade to 5.1B-4.
The transition went fine, except for the Xwin performance issue. There are a few other installations of this kind of equipment, and I know at least one where the upgrade to 5.1B-4 also caused noticeable degradation in Xwin performance.
We also have two DS10's which are NOT connected to any hardware and where we have the same software installed and when we do the same primitive test we see the same thing. Opening the one of the simplest applications in the vendor suite takes one second or less for the DS10 with 4.0F, and about 3 seconds for the DS10 with 5.1B.
Having said that, it should be mentioned that the test is performed using one of the the vendor's software products which probably is written for 4.0F. Could it be that the programming different for 4.0F and 5.1B to get the best performance?
However, "xterm" is coming up instantly from 4.0F, and maybe after 0.5-1 seconds from 5.1B, so I doubt this idea.

BN

Pieter 't Hart
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

I'm curious about the two lines on the PWS500au with 5.1b
129.16/16 129.16.138.202 U 5 695751 tu0
129.16.138.202 129.16.138.202 UHL 1 896452 tu0
192.168.27/24 192.168.27.1 U 12 3103365 tu1
192.168.27.1 192.168.27.1 UHL 0 0 tu1

where the 40.f system has one per interface
PWS500au system with 4.0F
129.16/16 129.16.138.198 U 2 62575 tu0
192.168.27/24 192.168.27.1 U 9 571789 tu1

do you have IPv6 enabled on the PWS500au with 5.1b?
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

Interesting question.
How do I check that?


Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

I was doing "ip6_setup", answering "no" on all important questions, and I got the following:

"IPV6 option is not configured in the kernel.
Can not configure IPv6 interfaces and routes."

Is then the ipv6 issue out of the discussion?

BN

Pieter 't Hart
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

I guess so.
I'm not familiar with ipv6 but found some reference for the flags to ipv6.

have you tried "grep hosts /etc/nsswitch.conf" ?
this parameter determines the order of nameresolution (hostfile, dns ...).

and if present can you post the entries in the host files for these nodes?
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

I am not sure I understand, but here is the response.

PWS500au 5.1B-4:
grep hosts /etc/nsswitch.conf
# hosts: files dns nis
# hosts
hosts: files dns nis

DS10 5.1B-4
grep hosts /etc/nsswitch.conf
# hosts: files dns nis
# hosts
hosts: files dns

The 4.0F systems does not have this file.
Pieter 't Hart
Honored Contributor

Re: Xwin performance of 5.1B-4 vs. 4.0F

as far as i know, in the background X does a reverse namelookup when a connection is requested.
so it tries the different possibilities (DNS, /etc/hosts ...) to resolve the ip-adress back to a hostname.
this takes some extra time escpecially if the name cannot be resolved.
so the order of lookup is important (nsswitch config) and entries in /etc/hosts for your display machine is important.

some other thing came up to me.
can you try this with the 4.0F machine shut down (or disconnected)?
there may be something in the application configuration that references the 4.0F system!
if so, when you connect from the displaystation to the 5.1b server, the application here makes in the background a connection to the 4.0F system, wich causes the extra delay. not the performance of the 5.1b system itself!

X produces logfiles in the "$HOME/.dt" directory.
can you check these for clues?
Bengt Nilsson_2
Regular Advisor

Re: Xwin performance of 5.1B-4 vs. 4.0F

I could find nothing remarkable in the log files in .dt.

I am now about to sign up for a support contract on this machine, so maybe I can get some more info after that.

Let's rest this case for the moment.

BN