Welcome to MilkyWay@home

Travis: Please set a minimum update interval

Message boards : Number crunching : Travis: Please set a minimum update interval
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
Bill & Patsy
Avatar

Send message
Joined: 7 Jul 08
Posts: 47
Credit: 13,629,944
RAC: 0
Message 17512 - Posted: 4 Apr 2009, 4:33:00 UTC

LHC uses a minimum 15 minute interval between client updates. Attempted updates before the interval completes result in no work and the clock is reset for another full interval instead. Such server-side minimum update intervals prevent spamming and hammering the server, which has become a sport at Milkyway and a growing problem for everyone. If you enable this feature, it should stop this nonsense and assure work availability for everyone.

Thanks.
--Bill

ID: 17512 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile The Gas Giant
Avatar

Send message
Joined: 24 Dec 07
Posts: 1947
Credit: 240,884,648
RAC: 0
Message 17514 - Posted: 4 Apr 2009, 5:04:07 UTC
Last modified: 4 Apr 2009, 5:59:27 UTC

No problems with me. It takes about 12 minutes for me to complete the maximum number of wu's (6/core) that are allocated. I'd love to be able to get another 24 tasks 3 minutes after completing them versus the approx 20min it takes now!
ID: 17514 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dan T. Morris
Avatar

Send message
Joined: 17 Mar 08
Posts: 165
Credit: 410,228,216
RAC: 0
Message 17515 - Posted: 4 Apr 2009, 6:56:12 UTC

I don't feel that it would be a good idea.

I have a great idea lets just all stop complaining and crunch what you get and help out the project.
ID: 17515 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MontagsMeeting

Send message
Joined: 12 Mar 09
Posts: 61
Credit: 9,214,340
RAC: 0
Message 17516 - Posted: 4 Apr 2009, 7:15:50 UTC

It's a great idea

i get sometimes only one or two WUs, i do them in 30 seconds, so this or less has to be the min update interval.

I'll set that in my update script at once - Bill & Patsy, thanks for the suggestion

Or not?

Perhaps it would be better to reduce the deadline from 3 days to 10 minutes, this would make sure those who can do will be supplied with work and the project would not be thwarted by the slow crunchers.

Travis, please!

just kidding of course, but some should really think about what they say before they say

Thanks Dan, that's the way :)
ID: 17516 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Glenn Rogers
Avatar

Send message
Joined: 4 Jul 08
Posts: 165
Credit: 364,966
RAC: 0
Message 17517 - Posted: 4 Apr 2009, 7:22:41 UTC - in response to Message 17515.  

Here Here Dr Dan, think its about time those who are whining about the wu situation go GET A LIFE and wait for Travis to sort something out...
ID: 17517 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
ExtraTerrestrial Apes
Avatar

Send message
Joined: 1 Sep 08
Posts: 204
Credit: 219,354,537
RAC: 0
Message 17519 - Posted: 4 Apr 2009, 9:09:04 UTC

A 15 min timeout would be stupid IMO, as the maximum cache you can get with a quad lasts only 12 mins with a 150€ 4870.

However, there is something which oculd be done: after a server request BOINC goes into an automatic timeout of 7 s at MW, whereas this time interval is 1 min by default. Well, a GPU is fast, but it's not that fast.. so if one does actually get a full cache, the next server requests are wasted as BOINC doesn't understand this "reached maximum quota per CPU"-message. If this time is increased to 1 min quite some unneccessary server requests would be eleminated.

It wouldn't stop update scripts or manual updates, though. If you went for the solution Bill & Patsy suggested I recommend a minimum time between connects of not more than 1 min. That should keep people from hammering the server in an unresponsible manner and would still give machines with 2 GPUs or an X2 at least a chance to get new work in time.

MrS
Scanning for our furry friends since Jan 2002
ID: 17519 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile The Gas Giant
Avatar

Send message
Joined: 24 Dec 07
Posts: 1947
Credit: 240,884,648
RAC: 0
Message 17520 - Posted: 4 Apr 2009, 9:18:13 UTC

abc@home use a 2 min delay.
ID: 17520 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile verstapp
Avatar

Send message
Joined: 26 Jan 09
Posts: 589
Credit: 497,834,261
RAC: 0
Message 17524 - Posted: 4 Apr 2009, 10:09:07 UTC

>give machines with 2 GPUs or an X2 at least a chance to get new work in time.
But first there has to be work ready to d/l...

Cheers,

PeterV

.
ID: 17524 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Odd-Rod

Send message
Joined: 7 Sep 07
Posts: 444
Credit: 5,712,451
RAC: 0
Message 17525 - Posted: 4 Apr 2009, 10:09:13 UTC

I agree with the idea, although from the responses about crunch times, perhaps 15 min is a bit long.

But the point is, as soon as it becomes pointless to use scripts to hammer the server more often than the min update interval, crunchers will hopefully stop using them, and the server may just be able to cope better.

Just so you know, I have no GPU and I don't use a script. If I don't get work from MW, Boinc gets from other projects.
ID: 17525 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile [XTBA>XTC] ZeuZ

Send message
Joined: 27 Dec 07
Posts: 14
Credit: 5,089,974
RAC: 0
Message 17527 - Posted: 4 Apr 2009, 10:35:08 UTC

15 minutes? Is it a joke?

My 4850 overclocked work on a E5300 (dualcore) and 6 Wu per core do 12 wu in cache !

12 wu in cache does mean a maximum of 6 min of work ! What about the other nine ones?
And what about my 3850 on my celeron single core?

Everybody don't have a big quadcore or Core i7

I don't buy graphics cards for fun !

Why doesn't Travis increase the length again? by five or ten for exemple? This won't be disturbing for CPUs
ID: 17527 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MontagsMeeting

Send message
Joined: 12 Mar 09
Posts: 61
Credit: 9,214,340
RAC: 0
Message 17528 - Posted: 4 Apr 2009, 11:13:50 UTC
Last modified: 4 Apr 2009, 11:17:34 UTC

Hello ???

The interval has to be less than one WU can be done by any client!!!

This is far less than a minute!!!

So the interval has to be set smaller than this - easy to understand, i think

So the update interval can NOT be greater than 30 seconds and i don't know if this caps some hardware already.

This idea lacks any sense!

It's a simple me me me i don't get enough, someone got my pacifier idea, nothing else
ID: 17528 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Lord Tedric
Avatar

Send message
Joined: 9 Nov 07
Posts: 151
Credit: 8,391,608
RAC: 0
Message 17529 - Posted: 4 Apr 2009, 11:37:47 UTC - in response to Message 17528.  
Last modified: 4 Apr 2009, 11:43:51 UTC

Hello ???

The interval has to be less than one WU can be done by any client!!!

This is far less than a minute!!!

So the interval has to be set smaller than this - easy to understand, i think

So the update interval can NOT be greater than 30 seconds and i don't know if this caps some hardware already.

This idea lacks any sense!

It's a simple me me me i don't get enough, someone got my pacifier idea, nothing else


How about making the interval - the culminative time of the number of units downloaded by the given host.

In that, each batch of 6 (per core) is taken as a 'whole' and no new units issued until the 'batch' is complete.
ID: 17529 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MontagsMeeting

Send message
Joined: 12 Mar 09
Posts: 61
Credit: 9,214,340
RAC: 0
Message 17533 - Posted: 4 Apr 2009, 12:19:09 UTC
Last modified: 4 Apr 2009, 12:26:54 UTC

won't work too i think
1. Implementation: If i request 48 WU (i7 HT) the server has to wait until there are 48 WU ready to send any subsequent requests will be dropped, that will not help against update scripts
2. Implementation: When the Server has 6 WU ready the first client that requests 6(slowest PCs - 1 core) will get them - this way fast CPUs/GPUs will never get work until slower PCs have enough work

The problem is the behaviour of BOINC it is not made for a situation where no work is available. If a request fails BOINC will increase the time to the next request. This penalizes fast machines as they have to request more frequently than slower ones and so their time to the next request increases very fast and its very likely that they run dry for days. That's why especially GPU-users have to use such scripts to get at least the amount that slower PCs get percentaged.

I've two 4850 GPUs that starve at about 50% of their potential - not funny, but that's the situation. And if i don't use a script i will not get anything because of time to next request is always many hours

I think there can nothing be done than to wait that the project will be able to serve more work and on that solution is worked at the moment. Should they stop their work only to implement some disputable things and we have to wait even longer for a real solution?

I have a great idea lets just all stop complaining and crunch what you get and help out the project.

That is the only thing that can be done at the moment, anything else will delay a real solution and that will end in even more complaining and whining.

If you read between the lines of the complainers, all that they say is: stop the others give it to me! - no need for any attention, sadly if one starts all whiners that could restrain themself start to sing from the same hymn sheet. Thats the point when they stop thinking because some base instict takes control: "I probably miss something" - They don't care about the project, they don't care about others, it's just them
ID: 17533 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
kimmerin

Send message
Joined: 14 Mar 09
Posts: 16
Credit: 35,833,011
RAC: 0
Message 17546 - Posted: 4 Apr 2009, 15:27:06 UTC - in response to Message 17527.  

I don't buy graphics cards for fun !


No? So you run CAD-applications on your Celeron-based box?

Sorry, but buying new equipment for BOINC only is in my eyes contradicting the idea of using "unused" cycles.

Save your money for something else, the project itself will most likely not finish much faster with your contribution.


Regards, Lothar
ID: 17546 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile [XTBA>XTC] ZeuZ

Send message
Joined: 27 Dec 07
Posts: 14
Credit: 5,089,974
RAC: 0
Message 17551 - Posted: 4 Apr 2009, 15:56:21 UTC

This computer is used for an office automation use and for some little games as Fifa 08 so ...

I don't know why I tell you this, I do what I want with my money that's all. A lot of people do same thing and so what? It's our hobby ! Can you Respect this?

ID: 17551 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile [XTBA>TSA] Poulpito
Avatar

Send message
Joined: 19 Oct 08
Posts: 1
Credit: 20,664,945
RAC: 0
Message 17553 - Posted: 4 Apr 2009, 16:05:13 UTC - in response to Message 17551.  

ZeuZ stop buying stupid things like a single graphic card ....
take it at least by two

ID: 17553 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile speedimic
Avatar

Send message
Joined: 22 Feb 08
Posts: 260
Credit: 57,387,048
RAC: 0
Message 17556 - Posted: 4 Apr 2009, 16:32:42 UTC - in response to Message 17527.  

...
I don't buy graphics cards for fun !
...


I don't get it... Can we get money for credits now?

mic.


ID: 17556 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile banditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
Message 17557 - Posted: 4 Apr 2009, 16:48:40 UTC - in response to Message 17556.  

...
I don't buy graphics cards for fun !
...


I don't get it... Can we get money for credits now?


I kept trying for candy but I am still waiting!! :(
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 17557 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Bill & Patsy
Avatar

Send message
Joined: 7 Jul 08
Posts: 47
Credit: 13,629,944
RAC: 0
Message 17558 - Posted: 4 Apr 2009, 16:53:43 UTC
Last modified: 4 Apr 2009, 16:59:57 UTC

Getting back to the point...

In my initial post, I purposely did not suggest what the interval should be for MilkyWay. Travis will have to work that out based upon performance results and everyone's feedback. I cited LHC as an example, but yes, the LHC 15 min interval is clearly too long for MilkyWay.

As for "no work available", I don't understand why some people keep saying that. Every time I check the server it has plenty of work available. (The Home Page has a hot link "Server Status" where you can check this.)

Some have suggested that the "no work available" update problem is that the available work cannot load into the scheduler or upload/download server, or whatever, fast enough because the server is getting hammered so hard by all these requests. So if everyone backed off to some minimum interval, then we all could get work, maybe even on the first request, because the server would have time to actually serve us.

If this works, then everyone will be served, and those of you who are pushing to the front of the line will be fed plenty too without having to be so pushy. Those scripts may help some individuals, but they're hurting everyone else. A proper minimum update interval could fix this.

It's sure worth a try.
--Bill

ID: 17558 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 302,524,931
RAC: 15
Message 17565 - Posted: 4 Apr 2009, 18:02:21 UTC - in response to Message 17558.  

Interesting concept 'having *everyone* back off to some minimum interval' sounds like a view toward credit crunching altruism which, if applied to real social problems would eliminate poverty and war. Probably would require some modification of the human genome though....



Some have suggested that the "no work available" update problem is that the available work cannot load into the scheduler or upload/download server, or whatever, fast enough because the server is getting hammered so hard by all these requests. So if everyone backed off to some minimum interval, then we all could get work, maybe even on the first request, because the server would have time to actually serve us.

If this works, then everyone will be served, and those of you who are pushing to the front of the line will be fed plenty too without having to be so pushy. Those scripts may help some individuals, but they're hurting everyone else. A proper minimum update interval could fix this.

It's sure worth a try.


ID: 17565 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
1 · 2 · 3 · Next

Message boards : Number crunching : Travis: Please set a minimum update interval

©2024 Astroinformatics Group