Ok, I officially give up, and believe it's problems with the SDAC software.
To answer some inquiries:
- with or without uttsc makes no difference, the standard login is also
blocky and laggy
- kiosk mode on/off
- We just spun up a dedicated Solaris 10 u8 box ... no improvement in
SDAC, dtu's work perfect.
- We put only 1 user on the above
- We've tried different client os's - winxp pro 32 bit, win 7 pro 64 bit
- 10 / 100 1000 Mbps networking
- on solaris, set highres_tick = 1, no difference
- Different architectures of processor and video cards
- different types of network switching
- we have security enabled to the maximum in our SRS environments
- basically we've replaced every single component we can to make this
work.
Basically, SDAC is the only common item which has not performed to the same
level of a sun ray DTU or rdp.
If/when we re-examine this software, I'll advise this group.
Thanks,
Devin Nate
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Craig Bender
Sent: Monday, March 01, 2010 9:05 PM
To: SunRay-Users mailing list
Subject: Re: [SunRay-Users] EXTERNAL:Re: EXTERNAL: Sun Desktop AccessClient
(SDAC / soft client) poor performance
How is performance not using uttsc?
Devin Nate wrote:
> Hi Folks, William;
>
> I've just tried a totally different manchine. My home WinXP Pro 32-bit
> machine. Yes we'd tried some, but they were IBM/Lenovo, whereas mine has a
> totally different CPU /Networking /Video /etc. Ping rtt's are small ...
> performance of SDAC is just as bad.
>
> Ping times attached. Sample utcapture output attached, showing poor perf (the
> long tokenID is that of SDAC). Second utcapture showing a sun ray DTU, with
> good times. Finally, another ping capture to show that MTU's up to and
> including 1300 are usable (and the client is set to 1133 for this test).
>
> This is going to kill the project for a few hundred users, and more
> importantly it's driving me nuts. I suspect either a bug in SDAC or an
> incompat for SDAC with RHEL. Every other app works great across this network,
> even the native Sun DTU's work great across this network.
>
> Very frustrating.
>
> Dare I ask, any more suggestions?
>
> Is anyone here from the Sun SDAC devel team? If so, maybe we could arrange
> something? [email protected].
>
> Thanks,
> Devin
>
>
> 1. Ping time RTT.
> C:\Documents and Settings\nated>ping 204.12.xx.yy
> Pinging 204.12.xx.yy with 32 bytes of data:
> Reply from 204.12.xx.yy: bytes=32 time=25ms TTL=57
> Reply from 204.12.xx.yy: bytes=32 time=15ms TTL=57
> Reply from 204.12.xx.yy: bytes=32 time=14ms TTL=57
> Reply from 204.12.xx.yy: bytes=32 time=14ms TTL=57
> Ping statistics for 204.12.xx.yy:
> Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
> Approximate round trip times in milli-seconds:
> Minimum = 14ms, Maximum = 25ms, Average = 17ms
> C:\Documents and Settings\nated>
>
> 2. utcapture of BAD performance using SDAC.
> r...@srs1(/root)
> # /opt/SUNWut/sbin/utcapture -r -s srs2 b976bba2338c4fcccf61d1c162ff1697
> # TERMINALID TIMESTAMP TOTAL PACKET TOTAL LOSS BYTES SENT PERCENT
> LOSS LATENCY
> b976bba2338c4fcccf61d1c162ff1697 20100301201048 2963 0
> 1557314 0.000 501.000
> b976bba2338c4fcccf61d1c162ff1697 20100301201103 2963 0
> 1557314 0.000 501.000
>
> 3. utcapture of GOOD performance using Sun Ray DTU (model 270).
> r...@srs1(/root)
> # /opt/SUNWut/sbin/utcapture -r -s srs2 00144fd354cc
> # TERMINALID TIMESTAMP TOTAL PACKET TOTAL LOSS BYTES SENT PERCENT
> LOSS LATENCY
> 00144fd354cc 20100301201152 4744 21 483326
> 0.000 15.823
>
> 4. ping showing that mtu up to and including 1300 is usable without
> fragmentation.
> C:\Documents and Settings\nated>ping -l 1300 -f 204.12.152.133
> Pinging 204.12.152.133 with 1300 bytes of data:
> Reply from 204.12.152.133: bytes=1300 time=56ms TTL=57
> Reply from 204.12.152.133: bytes=1300 time=16ms TTL=57
> Reply from 204.12.152.133: bytes=1300 time=22ms TTL=57
> Reply from 204.12.152.133: bytes=1300 time=29ms TTL=57
> Ping statistics for 204.12.152.133:
> Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
> Approximate round trip times in milli-seconds:
> Minimum = 16ms, Maximum = 56ms, Average = 30ms
>
>
>
>
>
> ________________________________________
> From: [email protected] [[email protected]]
> On Behalf Of William Yang [[email protected]]
> Sent: Monday, March 01, 2010 5:40 PM
> To: 'SunRay-Users mailing list'
> Subject: Re: [SunRay-Users] EXTERNAL:Re: EXTERNAL: Sun Desktop AccessClient
> (SDAC / soft client) poor performance
>
> Maybe try Windows XP if you haven't already (sorry if I missed that you
> already tried). I'm on XP and performance is great.
>
> William Yang
>
>> -----Original Message-----
>> From: [email protected] [mailto:sunray-users-
>> [email protected]] On Behalf Of Devin Nate
>> Sent: Monday, March 01, 2010 7:25 PM
>> To: SunRay-Users mailing list
>> Subject: Re: [SunRay-Users] EXTERNAL:Re: EXTERNAL: Sun Desktop
>> AccessClient (SDAC / soft client) poor performance
>>
>> Thanks Bob for the followup.
>>
>> I've been up and down over mtu ... that was my guess at first too. I can
>> confirm by packet capture that SDAC is using a mtu no larger than the
>> slider in the client is set to for udp packets. Tcp packets appear to be
>> using pmtud and getting an fine mtu also.
>>
>> I'm spinning up a host in our data center to test... but so far, no good
>> solutions have come out of this.
>>
>> I suspect I'm going to need to get Oracle/Sun to open a case.. this is
>> frustrating enough that I'd like to get folks with access to the source
>> code involved. Where every other app works...?
>>
>> I may get a Solaris SRS running here as well for comparative testing.
>>
>> Anyhow, if anyone has any more ideas, let me know.
>>
>> Thanks,
>> Devin
>>
>>
>>
>>
>> -----Original Message-----
>> From: [email protected] [mailto:sunray-users-
>> [email protected]] On Behalf Of Bob Doolittle
>> Sent: Monday, March 01, 2010 3:43 PM
>> To: SunRay-Users mailing list
>> Subject: Re: [SunRay-Users] EXTERNAL:Re: EXTERNAL: Sun Desktop
>> AccessClient (SDAC / soft client) poor performance
>>
>> MTU settings sound more plausible than uttsc to me.
>> uttsc should not affect latency as measured by utcapture - that's purely
>> a metric measured between the X server and the client.
>>
>> -Bob
>>
>> Nishimura, Scott L (IT Solutions) wrote:
>>> Devin,
>>>
>>> I'll have to defer to more knowledgeable members of this list. The
>>> performance you describe, however, is suspiciously similar to what we
>>> experienced a few years back before we had tuned our MTU settings [we
>>> ultimately settled on 1470]: when going through a PowerPoint stack,
>>> we'd get 5-second delays in screen redraw.
>>>
>>> The only other time I've seen something like this was when I was using
>>> PCSC [PC Smart Card] 1.0 and when I got a normal workload on the SRS,
>>> pcscd started taking up 6% of the CPU and users would see many 5-second
>>> delays during their session.
>>>
>>> Scott
>>>
>>> -----Original Message-----
>>> From: [email protected]
>>> [mailto:[email protected]] On Behalf Of Devin Nate
>>> Sent: Monday, March 01, 2010 2:27 PM
>>> To: SunRay-Users mailing list
>>> Subject: Re: [SunRay-Users] EXTERNAL:Re: EXTERNAL: Sun Desktop
>>> AccessClient (SDAC / soft client) poor performance
>>>
>>> Ahh, now that's something we haven't done yet... packet capture of the
>>> uttsc data. So far we've done a packet capture of the dtu<-->srss and
>>> compared with sdac<-->srss, and we couldn't see any difference there.
>>>
>>> I'll report observations asap.
>>>
>>> That said, would a difference in uttsc traffic to the terminal servers
>>> cause utcapture to indicate a higher latency?
>>>
>>> Thanks,
>>> Devin
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: [email protected]
>>> [mailto:[email protected]] On Behalf Of Nishimura, Scott
>>> L (IT Solutions)
>>> Sent: Monday, March 01, 2010 3:16 PM
>>> To: SunRay-Users mailing list
>>> Subject: Re: [SunRay-Users] EXTERNAL:Re: EXTERNAL: Sun Desktop Access
>>> Client (SDAC / soft client) poor performance
>>>
>>> Devin,
>>>
>>> How about if you run a truss or pstack or similar troubleshooting tool
>>> on uttsc and compare the results between the actual DTU and SDAC?
>>>
>>> Does your sniffer test show excessive retransmissions?
>>>
>>> Also, what MTU size is being used? Too large [> 1470?] could lead to
>>> fragmentation; too small [< 500] could lead to "laggy" performance
>>> [we've run into that before on a physical DTU before].
>>>
>>>
>>> Scott
>>>
>>> -----Original Message-----
>>> From: [email protected]
>>> [mailto:[email protected]] On Behalf Of Devin Nate
>>> Sent: Monday, March 01, 2010 2:06 PM
>>> To: [email protected]; SunRay-Usersmailing list
>>> Subject: EXTERNAL:Re: [SunRay-Users] EXTERNAL: Sun Desktop Access Client
>>> (SDAC / soft client) poor performance
>>>
>>> Hi Lars;
>>>
>>> I completely agree... the problem is, the DTU and SDAC clients are on
>>> identical networks... in fact, I've taken the cable out of a perfect DTU
>>> and put it into a PC. I was thinking perhaps QoS or something tagging
>>> the udp packages.
>>>
>>> The PCs have Intel nic's. I'm tracking down totally different
>>> architectures. Our core is all Cisco networking gear. We're looking to
>>> setup a host on the same LAN directly connected to the core, that said
>>> we had some hosts in the datacenter getting latency of 500.000, gig
>>> connected but on a different subnet.
>>>
>>> All that aside, the Sun DTU's work perfect on the same network. We're
>>> going through everything with a fine tooth comb, but I'm not optimistic.
>>>
>>> More thoughts?
>>>
>>> Thanks,
>>> Devin
>>>
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: [email protected]
>>> [mailto:[email protected]] On Behalf Of Lars Tunkrans
>>> Sent: Monday, March 01, 2010 2:36 PM
>>> To: SunRay-Users mailing list
>>> Subject: Re: [SunRay-Users] EXTERNAL: Sun Desktop Access Client (SDAC /
>>> soft client) poor performance
>>>
>>> 2010-03-01 20:31, Devin Nate skrev:
>>>
>>>> I'm really appreciating the ideas, please keep them coming?
>>>>
>>>> Thanks,
>>>> Devin
>>>>
>>>>
>>> Hi,
>>>
>>> When I experience laggy screen updates on a DTU ist is usually
>>> becase there is
>>> a problem with the network .
>>>
>>> As the list can testify:
>>>
>>> One returning problem is Low end flaky L2 switches with a
>>> gigabit input from the CORE switch
>>> and 100Mbit link out to the DTU. These L2 switches drops UDP packets
>>>
>>> and forces the DTU
>>> to ask for them again . Hence the Laggy screen updates.
>>>
>>>
>>>
>>> Now heres a long shot.....
>>>
>>> if you are using mainly the same brand pc , do they all have the
>>> same LAN Card / CHipset ?
>>> There are many kinds of problems with Chinese wierdo combines of
>>> PHY / MAC chipsets.
>>> They come up with new combinations every week just to be able to
>>> produce the motherboard 5 cents
>>> cheaper.
>>>
>>> Should you try an Intel Pro1000 GT Desktop adaptor ? I think
>>>
>>> so.
>>>
>>>
>>> http://www.intel.com/products/desktop/adapters/pro1000gt/pro1000gt-overv
>>> iew.htm
>>>
>>> This is probably the the most stable ethernet card today, since
>>> 3COM stopped making
>>> ethernet cards.
>>>
>>>
>>> Did you try to attach the PC directly to core switch yet ?
>>>
>>> //Lars
>>> _______________________________________________
>>> SunRay-Users mailing list
>>> [email protected]
>>> http://www.filibeto.org/mailman/listinfo/sunray-users
>>> _______________________________________________
>>> SunRay-Users mailing list
>>> [email protected]
>>> http://www.filibeto.org/mailman/listinfo/sunray-users
>>> _______________________________________________
>>> SunRay-Users mailing list
>>> [email protected]
>>> http://www.filibeto.org/mailman/listinfo/sunray-users
>>> _______________________________________________
>>> SunRay-Users mailing list
>>> [email protected]
>>> http://www.filibeto.org/mailman/listinfo/sunray-users
>>> _______________________________________________
>>> SunRay-Users mailing list
>>> [email protected]
>>> http://www.filibeto.org/mailman/listinfo/sunray-users
>>>
>> _______________________________________________
>> SunRay-Users mailing list
>> [email protected]
>> http://www.filibeto.org/mailman/listinfo/sunray-users
>> _______________________________________________
>> SunRay-Users mailing list
>> [email protected]
>> http://www.filibeto.org/mailman/listinfo/sunray-users
>
> _______________________________________________
> SunRay-Users mailing list
> [email protected]
> http://www.filibeto.org/mailman/listinfo/sunray-users
> _______________________________________________
> SunRay-Users mailing list
> [email protected]
> http://www.filibeto.org/mailman/listinfo/sunray-users
--
Craig Bender
1-877-255-1537
Sun Ray Engineering
Sun Microsystems
A Wholly Owned Subsidiary of Oracle Corporation
_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users
_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users
