Windows TCP Settings - Follow up - Help with server communication

Message boards : Number crunching : Windows TCP Settings - Follow up - Help with server communication
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 6 · 7 · 8 · 9 · 10 · 11 · 12 . . . 14 · Next

AuthorMessage
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1345579 - Posted: 11 Mar 2013, 22:04:10 UTC - in response to Message 1345539.  

Yes, TCPOptimizer wants to set my MTU to 1,500. Problem here is that I have Jumbo Frames on on my LAN, meaning that I run with an MTU of 8KB on the 1Gbit LAN. The MTU onto the internet is already 1,500, as set by my router, it can't be any greater anyway.

It has cost me a great deal to get the Jumbo Frames to work correctly on the network. It means that I can throw large files to the TV-server and not have to wait an hour for them to get there. Transfer speeds are in excess of 100MB/sec.

I am not going to throw all that away so that BOINC can download tasks at a sustained speed. It's weird though, as everyone now quoting this as if it's the New Holy Grail was before telling everyone complaining that the project never said they'd have work available 24/7, or that since the bandwidth was saturated to calm down, as everyone felt it. Apparently those notions have changed and it's now a class of important people, and the rest of the rabble. Which is sad.

But as I said, I don't mind waiting. Or my BOINC does. When it runs out of Seti work, there'll be more to do for the backup project.
ID: 1345579 · Report as offensive
Profile Bernie Vine
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 26 May 99
Posts: 9958
Credit: 103,452,613
RAC: 328
United Kingdom
Message 1345583 - Posted: 11 Mar 2013, 22:13:45 UTC - in response to Message 1345579.  

Yes, TCPOptimizer wants to set my MTU to 1,500. Problem here is that I have Jumbo Frames on on my LAN, meaning that I run with an MTU of 8KB on the 1Gbit LAN. The MTU onto the internet is already 1,500, as set by my router, it can't be any greater anyway.

It has cost me a great deal to get the Jumbo Frames to work correctly on the network. It means that I can throw large files to the TV-server and not have to wait an hour for them to get there. Transfer speeds are in excess of 100MB/sec.

I am not going to throw all that away so that BOINC can download tasks at a sustained speed. It's weird though, as everyone now quoting this as if it's the New Holy Grail was before telling everyone complaining that the project never said they'd have work available 24/7, or that since the bandwidth was saturated to calm down, as everyone felt it. Apparently those notions have changed and it's now a class of important people, and the rest of the rabble. Which is sad.

But as I said, I don't mind waiting. Or my BOINC does. When it runs out of Seti work, there'll be more to do for the backup project.


I never mentioned TCP optmiser, you can set timestamps without altering any other settings.

I have no idea what "this class of important people means" personally I want everyone to set "timestamps" why would I not? It is the way TCP is supposed to work. All Windows users who can should set their machines this way not just for S@H but for the internet in general to deal with what will become more common, congested links.

ID: 1345583 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13855
Credit: 208,696,464
RAC: 304
Australia
Message 1345740 - Posted: 12 Mar 2013, 7:01:54 UTC - in response to Message 1345579.  

Yes, TCPOptimizer wants to set my MTU to 1,500.

So don't use it, use the command line commands as i did. If you don't like the result, just delete the registry entries they make.

Grant
Darwin NT
ID: 1345740 · Report as offensive
tbret
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 3380
Credit: 296,162,071
RAC: 40
United States
Message 1345749 - Posted: 12 Mar 2013, 7:31:09 UTC - in response to Message 1345579.  



Apparently those notions have changed and it's now a class of important people, and the rest of the rabble. Which is sad.




Oh, I don't know. Do you really think there's that much ego involved?

I think what separates "us" from "them" is the amount of interest we / they show.

I'm not saying you're wrong. I have no clue what might be in the mind of anyone else unless they tell me.

I just can't imagine anyone thinking they are "important" and someone else really isn't. Well...I don't know how many politicians or people suffering from the Narcissism personality disorder there are here, but it can't be that many.

But just from an objective point of view, if the project really wants to accomplish something / anything, the project personnel should have a bias toward those who contribute the most in helping them accomplish their goals.

I've never understood anyone who (and I'm not saying YOU are "anyone who," I'm just showing a blank spot in my conscience) thinks that as long as we all share nothing equally, we're better-off than if someone else is allowed to have something.
ID: 1345749 · Report as offensive
juan BFP Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 9786
Credit: 572,710,851
RAC: 3,799
Panama
Message 1345760 - Posted: 12 Mar 2013, 8:19:57 UTC

I realy don´t belive if they change from 100 GPU WU per host to 100 GPU WU per GPU will "crash" the DB... and that will give the fastest crunchers enought WU to pass the scheduled outages.

If the DB size is the problem, then why not simply decrease the 100 CPU WU limit to 50WU? That will give a big diference (100k users with less 50WU each = 5MM WU!) against probabily few 100´s who have 2 or 3 GPU hosts (lets say 1K x 100 =100K WU). Don´t mention, very few hosts could do 100 CPU WU in a day...

Totaly out of focus... Hi Bret, how is your kittie?



ID: 1345760 · Report as offensive
Profile Bernie Vine
Volunteer moderator
Volunteer tester
Avatar

Send message
Joined: 26 May 99
Posts: 9958
Credit: 103,452,613
RAC: 328
United Kingdom
Message 1345920 - Posted: 12 Mar 2013, 22:08:44 UTC

Ok as one of the "important people" who has set "timestamps on" I am reporting that after the outage all three current crunchers downloaded over 80 wu's without a pause.

Any other "important people" have any comments.
ID: 1345920 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 36828
Credit: 261,360,520
RAC: 489
Australia
Message 1345925 - Posted: 12 Mar 2013, 22:15:46 UTC - in response to Message 1345920.  

Ok as one of the "important people" who has set "timestamps on" I am reporting that after the outage all three current crunchers downloaded over 80 wu's without a pause.

Any other "important people" have any comments.

I must be even more "important" as I have never had this problem but upon checking my registries my rigs already had it set (don't ask as I don't know why it was when many others don't).

Cheers.
ID: 1345925 · Report as offensive
bill

Send message
Joined: 16 Jun 99
Posts: 861
Credit: 29,352,955
RAC: 0
United States
Message 1345971 - Posted: 13 Mar 2013, 1:42:15 UTC - in response to Message 1345925.  

You must be Uber Important. It's a good thing
there are important people to get things done.
Otherwise we'd all be just sitting around watching TV.
ID: 1345971 · Report as offensive
Profile rebest Project Donor
Volunteer tester
Avatar

Send message
Joined: 16 Apr 00
Posts: 1296
Credit: 45,357,093
RAC: 0
United States
Message 1345983 - Posted: 13 Mar 2013, 2:20:37 UTC

Many thanks to whomever found the TCP fix and shared it. A few days ago I had 90 downloads hung up on one machine and 147 timeout errors.

With the TCP fix, my transfer rates are still painfully slow, but the WUs are getting through. That's what matters.

Join the PACK!
ID: 1345983 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13855
Credit: 208,696,464
RAC: 304
Australia
Message 1346027 - Posted: 13 Mar 2013, 6:41:39 UTC - in response to Message 1345983.  


Just got home from work.

For the first time i can remember in a very long time, neither of my machines had WUs queued up to download in various stages of backoff or project backoff.
Both have as much as the limits will presently allow. All without several hours of pounding the retry button repeatedly.
Wonderfull.
Grant
Darwin NT
ID: 1346027 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22535
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1346029 - Posted: 13 Mar 2013, 6:50:24 UTC

I wish to propose a vote of thanks:

"To those who uncovered, and shared, the Windows TCP settings that work without the need for a proxy"
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1346029 · Report as offensive
tbret
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 3380
Credit: 296,162,071
RAC: 40
United States
Message 1346051 - Posted: 13 Mar 2013, 7:48:46 UTC - in response to Message 1345760.  



Totaly out of focus... Hi Bret, how is your kittie?




Thanks for asking. Mitch seems to be back to normal (not real bright, but loveable) and he's even getting hair back on his belly following his surgery.
ID: 1346051 · Report as offensive
Patrick

Send message
Joined: 1 Aug 12
Posts: 2
Credit: 28,790
RAC: 0
Germany
Message 1346096 - Posted: 13 Mar 2013, 9:56:41 UTC

Timestamp is a good choice for Systems that are in different timezones.
Lamport clocks and Vector clocks allow to determine the events of the casual order without a synchronization of the realtimeclocks at different positions.
Such clocks are used from Network Protocols and Transaction Systems.

A Vectorclock is a software component (or protocol) to assign unique timestamps on messages. It is therefore a logical clock, which allows the events in a distributed system, based on a time stamp, to assign a causal order to determine (sequencing) and especially the concurrency of events. It represents an extension of Lamport clock, which also satisfies the strong condition Watches.

One Sentence to the concurrency

On a massive parallel computer, these instructions are executed in parallel. Some modern CPUs are even able to detect such situations and individual commands the same time in different parts of the same processor.

I have this Information translated from 3 sites on wikipedia perhaps that´s the thing why timestamp works for many people much better and the servers and or routers are running more parallelized so you get more performance over the network and the connection does´nt break or held so often.

It´s just something i´ve read and thinking about that.
If someone knows more over this let us know or donate to seti if you want perhaps that can help too.

ID: 1346096 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14679
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1346110 - Posted: 13 Mar 2013, 10:42:59 UTC - in response to Message 1346096.  

The timestamps in RFC1323 are nothing to do with real time clocks or time zones. They are simply a second set of 32-bit counters which help to distinguish duplicate TCP packet sequence numbers. A 100 megabit network connection (such as Berkeley is using) can cycle round the entire sequence number space in as little as three minutes. See the section on PAWS in the RFC.
ID: 1346110 · Report as offensive
Profile Siran d'Vel'nahr
Volunteer tester
Avatar

Send message
Joined: 23 May 99
Posts: 7379
Credit: 44,181,323
RAC: 238
United States
Message 1346136 - Posted: 13 Mar 2013, 12:20:06 UTC

Greetings all,

We're having a problem and I need some advice on what to do about it. I pertains to what this thread is all about.

When I discovered this topic a few days ago, I decided to give it a go. I followed the instructions for using TCPOptimizer explicitly. I am running an older generation Intel i7 860 box with 8 GB of memory, blah, blah... I am not having the problem.

I use Speedtest.net to check our connection speed on the Internet on occasion. I used to have 9 to 10 Mbps DL and .7 to .9 Mbps UL. Now I have seen DLs as high as 20 Mbps and ULs as high as 2.x Mbps. What a major change! I'm loving it! :)

Our other PC is a newer generation Intel i7 3770 with 8GB of memory, blah, blah... It was having about the same speeds as my i7 was before running TCPOptimizer on it. We downloaded and ran TCPOptimizer on it and things went downhill BIG time. It is getting just over 1 Mbps DLs and .5 Mbps ULs. That's not good! :(

First of all: Can I use TCPOptimizer to go back to having the Tcp1323Opts disabled? I would much rather do it that way than manually working in the registry.

Second: Can anyone come up with something we can do with Tcp1323Opts enabled to get the speeds I'm getting on my i7 on our other one?

Any ideas / help would be GREATLY appreciated! :)

Keep on BOINCing...! :)

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath
ID: 1346136 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1346147 - Posted: 13 Mar 2013, 12:50:37 UTC - in response to Message 1345925.  

I must be even more "important" as I have never had this problem but upon checking my registries my rigs already had it set (don't ask as I don't know why it was when many others don't).

Cheers.


Maybe because you are using a version of Windows (if Windows it is) that was obtained from a bittorrent site? Some of those have performance tweaks built in, and the timestamp may have been one of them.
ID: 1346147 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1346171 - Posted: 13 Mar 2013, 14:20:43 UTC - in response to Message 1346136.  

Greetings all,

We're having a problem and I need some advice on what to do about it. I pertains to what this thread is all about.

When I discovered this topic a few days ago, I decided to give it a go. I followed the instructions for using TCPOptimizer explicitly. I am running an older generation Intel i7 860 box with 8 GB of memory, blah, blah... I am not having the problem.

I use Speedtest.net to check our connection speed on the Internet on occasion. I used to have 9 to 10 Mbps DL and .7 to .9 Mbps UL. Now I have seen DLs as high as 20 Mbps and ULs as high as 2.x Mbps. What a major change! I'm loving it! :)

Our other PC is a newer generation Intel i7 3770 with 8GB of memory, blah, blah... It was having about the same speeds as my i7 was before running TCPOptimizer on it. We downloaded and ran TCPOptimizer on it and things went downhill BIG time. It is getting just over 1 Mbps DLs and .5 Mbps ULs. That's not good! :(

First of all: Can I use TCPOptimizer to go back to having the Tcp1323Opts disabled? I would much rather do it that way than manually working in the registry.

Second: Can anyone come up with something we can do with Tcp1323Opts enabled to get the speeds I'm getting on my i7 on our other one?

Any ideas / help would be GREATLY appreciated! :)

Keep on BOINCing...! :)

The Tcp1323Opts will only give you a higher throughput benchmark if you were previously having a high rate of packet loss & then have a reduced rate of packet loss afterward.
However to change the settings back you can use the "Windows Default" button or do it manually.
http://www.hal6000.com/seti/images/Tcp1323Opts_disabled.png
http://www.hal6000.com/seti/images/Tcp1323Opts_enabled.png
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1346171 · Report as offensive
Profile Siran d'Vel'nahr
Volunteer tester
Avatar

Send message
Joined: 23 May 99
Posts: 7379
Credit: 44,181,323
RAC: 238
United States
Message 1346205 - Posted: 13 Mar 2013, 16:03:29 UTC - in response to Message 1346171.  

Greetings all,

We're having a problem and I need some advice on what to do about it. It pertains to what this thread is all about.

-[ snip ]-

The Tcp1323Opts will only give you a higher throughput benchmark if you were previously having a high rate of packet loss & then have a reduced rate of packet loss afterward.
However to change the settings back you can use the "Windows Default" button or do it manually.
http://www.hal6000.com/seti/images/Tcp1323Opts_disabled.png
http://www.hal6000.com/seti/images/Tcp1323Opts_enabled.png

Greetings Hal,

Ok. This is an area of computing that I have not spent any great amount of time in in trying to learn about. I know just enough to get me by (and probably to be of danger to myself as well). ;)

Let me see if I understand what you said: My PC was having major packet loss so the fix fixed the problem. Our other computer had little if any packet loss so the fix unfixed it and broke it. Is that essentially correct? So what I need to do is unfix the fix, or fix the unfix. Right? ;) lol

Thanks Hal! :)

Keep on BOINCing...! :)

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath
ID: 1346205 · Report as offensive
Profile Siran d'Vel'nahr
Volunteer tester
Avatar

Send message
Joined: 23 May 99
Posts: 7379
Credit: 44,181,323
RAC: 238
United States
Message 1346211 - Posted: 13 Mar 2013, 16:32:58 UTC

Greetings,

@ Hal9000: When doing the TCPOptimizer fix, I did notice that on the newer PC the check box for "Window scaling" was checked and "Timestamps" not before doing anything with TCPOptimizer. I'm thinking all I need do is un-check the "Timestamps" check box and the PC should be back to 'normal'.

Looks to me like the 'version' of Win7 64 bit we got for the other PC was slightly different then the one I got for mine a year before. I don't and won't buy pre-built PCs by the way, I build them as needed.

Keep on BOINCing...! :)

CAPT Siran d'Vel'nahr - L L & P _\\//
Winders 11 OS? "What a piece of junk!" - L. Skywalker
"Logic is the cement of our civilization with which we ascend from chaos using reason as our guide." - T'Plana-hath
ID: 1346211 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1346246 - Posted: 13 Mar 2013, 18:37:39 UTC - in response to Message 1346205.  

Greetings all,

We're having a problem and I need some advice on what to do about it. It pertains to what this thread is all about.

-[ snip ]-

The Tcp1323Opts will only give you a higher throughput benchmark if you were previously having a high rate of packet loss & then have a reduced rate of packet loss afterward.
However to change the settings back you can use the "Windows Default" button or do it manually.
http://www.hal6000.com/seti/images/Tcp1323Opts_disabled.png
http://www.hal6000.com/seti/images/Tcp1323Opts_enabled.png

Greetings Hal,

Ok. This is an area of computing that I have not spent any great amount of time in in trying to learn about. I know just enough to get me by (and probably to be of danger to myself as well). ;)

Let me see if I understand what you said: My PC was having major packet loss so the fix fixed the problem. Our other computer had little if any packet loss so the fix unfixed it and broke it. Is that essentially correct? So what I need to do is unfix the fix, or fix the unfix. Right? ;) lol

Thanks Hal! :)

Keep on BOINCing...! :)

Basically yes.
Enabling timestamps is a "fix". If the conditions is fixes don't exist then you are just adding extra data to all of your packets. Sort of like tossing a few bricks in the back of a real wheel drive car in the winter and leaving them there in the summer.

The TCPOptimizer software can change other settings that may speed up or slow down your internet connection. Vendors, such as Dell or HP, may tweak those settings to make them better optimized than the Microsoft defaults. If you used it and things got worse. Restore the previous settings or change back to MS defaults & see how things go.

Another tool I use is a program call Visual Route. It is like running a trace but provides more info.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1346246 · Report as offensive
Previous · 1 . . . 6 · 7 · 8 · 9 · 10 · 11 · 12 . . . 14 · Next

Message boards : Number crunching : Windows TCP Settings - Follow up - Help with server communication


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.