Problems and Technical Issues with Rosetta@home

Message boards : Number crunching : Problems and Technical Issues with Rosetta@home

To post messages, you must log in.

Previous · 1 . . . 253 · 254 · 255 · 256 · 257 · 258 · 259 . . . 272 · Next

AuthorMessage
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107701 - Posted: 8 Nov 2022, 20:48:19 UTC - in response to Message 107699.  
Last modified: 8 Nov 2022, 20:48:51 UTC

My Verizon FiOS is fibre-optic to a box on the outside of my house. There the signal is converted to 4 voice lines, one ethernet line, and a co-ax for television ("cable"). I actually use only one voice line, but I did use two for a while. And I do not watch TV, so I do not need that. The ethernet line goes to a Verizon-supplied router that is hard-wired to my main computer, to a littlle Windows10 computer, and my UPS that keeps in occasional touch with the UPS vendor (APC).

In any case, I pay for 75 Megabits/second both up and down. And Verizon exceed those speeds almost all the time. I could pay more money and get ten times faster, but I would probably need a new router, and a new box on the side of my house (box now 18 years old). But my data rates to and from most servers is slower than what I have now to the various sites that that test program uses for measuring Internet performance, so I would not benefit from a faster connection.
You have fibre to the house and aren't using 1 Gbit? Can we swap houses please? "Wouldn't benefit" WTF? Everything can go faster.

What's a FiOS?
ID: 107701 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jean-David Beyer

Send message
Joined: 2 Nov 05
Posts: 168
Credit: 5,391,098
RAC: 3,766
Message 107702 - Posted: 8 Nov 2022, 20:49:16 UTC - in response to Message 107698.  

When I was working as an electrician's helper, I saw one on the problems with aluminum wire.

It SLOWLY moves away from whatever is squeezing it. Therefore, unless you keep constant spring tension on it, you'll eventually get an open circuit (a gap breaking the connection).


Ahh! Yes.

I used to work in a very fancy office building designed by Eero Saarinen who was a pretty good architect, but there were a few screw-ups. Aluminum wire for the main feeders, though copper to each office and each laboratory. I used the services of the plant department when installing a new computer for our work. And the main engineer in the plant department and I became pretty good friends. Several times a year he went to each and every power panel where the aluminum wire was present and measured the temperature of each connection with an optical infra-red pyrometer. Anything that was too hot got fightened or the wiring replaced.

https://www.midcenturyhome.com/eero-saarinens-bell-labs-holmdel-complex-collaboration-reignited/
ID: 107702 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107703 - Posted: 8 Nov 2022, 23:14:24 UTC - in response to Message 107702.  

Ahh! Yes.

I used to work in a very fancy office building designed by Eero Saarinen who was a pretty good architect, but there were a few screw-ups. Aluminum wire for the main feeders, though copper to each office and each laboratory. I used the services of the plant department when installing a new computer for our work. And the main engineer in the plant department and I became pretty good friends. Several times a year he went to each and every power panel where the aluminum wire was present and measured the temperature of each connection with an optical infra-red pyrometer. Anything that was too hot got fightened or the wiring replaced.

https://www.midcenturyhome.com/eero-saarinens-bell-labs-holmdel-complex-collaboration-reignited/
Ahhh, French electrics. Like in my car :-/
ID: 107703 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jean-David Beyer

Send message
Joined: 2 Nov 05
Posts: 168
Credit: 5,391,098
RAC: 3,766
Message 107704 - Posted: 8 Nov 2022, 23:30:39 UTC - in response to Message 107701.  

You have fibre to the house and aren't using 1 Gbit? Can we swap houses please? "Wouldn't benefit" WTF? Everything can go faster.

What's a FiOS?


What use is 1 GBit (my NIC will go that fast) to me when the servers at the other end, such as WCG, only go around 40 to 60KBytes per second. Increasing my peak rate from 75 Megabits/ second up to 1000 Megabits per second will make no difference at all. Everything at my end could go faster but the throughput is no faster than the slowest link in the chain, and I somehow do not think from here to Canada (where WCG is now located) wil go any faster no matter how fast my end is. Sending their signals through US border control incurs huge delays (just kidding).

FiOS is a data service from Verizon that is fibre-optic to the home. They offer different speeds for different prices, but I am paying only for 75 Megabits/second up and down.
ID: 107704 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1222
Credit: 13,706,757
RAC: 2,097
Message 107705 - Posted: 8 Nov 2022, 23:49:15 UTC - in response to Message 107700.  

When I was working as an electrician's helper, I saw one on the problems with aluminum wire.

It SLOWLY moves away from whatever is squeezing it. Therefore, unless you keep constant spring tension on it, you'll eventually get an open circuit (a gap breaking the connection).
Like trying to squeeze a worm and it pops out?

Slower than that, otherwise yes.
ID: 107705 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107706 - Posted: 9 Nov 2022, 0:08:35 UTC - in response to Message 107704.  
Last modified: 9 Nov 2022, 0:09:00 UTC

What use is 1 GBit (my NIC will go that fast) to me when the servers at the other end, such as WCG, only go around 40 to 60KBytes per second. Increasing my peak rate from 75 Megabits/ second up to 1000 Megabits per second will make no difference at all. Everything at my end could go faster but the throughput is no faster than the slowest link in the chain, and I somehow do not think from here to Canada (where WCG is now located) wil go any faster no matter how fast my end is. Sending their signals through US border control incurs huge delays (just kidding).
I assume you don't only use WCG? Most places will go way faster than your current connection. Especially piratebay!

FiOS is a data service from Verizon that is fibre-optic to the home.
I misread that as to the throne! (Toilet if throne only makes sense in the UK).
ID: 107706 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
.clair.

Send message
Joined: 2 Jan 07
Posts: 274
Credit: 26,399,595
RAC: 88
Message 107709 - Posted: 9 Nov 2022, 2:12:14 UTC - in response to Message 107705.  

When I was working as an electrician's helper, I saw one on the problems with aluminum wire.

It SLOWLY moves away from whatever is squeezing it. Therefore, unless you keep constant spring tension on it, you'll eventually get an open circuit (a gap breaking the connection).
Like trying to squeeze a worm and it pops out?

Slower than that, otherwise yes.

The worst of all is with fine multistrand copper wire that has its end soldered to a `nice` blob then screw fixed , it wont stay fixed for long :-(
ID: 107709 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107710 - Posted: 9 Nov 2022, 2:24:10 UTC - in response to Message 107709.  
Last modified: 9 Nov 2022, 2:24:59 UTC

The worst of all is with fine multistrand copper wire that has its end soldered to a `nice` blob then screw fixed , it wont stay fixed for long :-(
Just twisted up also comes loose, so does non-stranded solid copper. Maybe the screw undoes with vibration? It could be glued? Or is the wire moving or compressing? Big current stuff here is done up with two or more grub screws in a row. And you shouldn't have the wire under any tension so it pulls out when it gets loose.

Gah, that sentence is so untidy, this stupid forum removes my double spacing between sentences. And yet it can't get the line spacing right....
ID: 107710 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jean-David Beyer

Send message
Joined: 2 Nov 05
Posts: 168
Credit: 5,391,098
RAC: 3,766
Message 107711 - Posted: 9 Nov 2022, 2:27:29 UTC - in response to Message 107706.  
Last modified: 9 Nov 2022, 2:29:05 UTC

I assume you don't only use WCG? Most places will go way faster than your current connection. Especially piratebay!


My favorites now, in order, are ClimatePrediction, WCG, Rosetta. I also am doing universe and milkyWay to soak up unused processors because ClimatePrediction sent me only one work unit in the last 4 months (Nov 7) that bombed out with invalid theta. 4 other users bombed on that work unit too. Except for that one, the last work unit I got from CPDN was at the end of July. I had about an 8-month gap in WCG work units. And Rosetta does not keep up with my machine either. I used to do Seti@home and Malaria, but they are gone.

IIRC, none of those go faster than my connection.
ID: 107711 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107712 - Posted: 9 Nov 2022, 2:40:39 UTC - in response to Message 107711.  

My favorites now, in order, are ClimatePrediction, WCG, Rosetta. I also am doing universe and milkyWay to soak up unused processors because ClimatePrediction sent me only one work unit in the last 4 months (Nov 7) that bombed out with invalid theta.
They're planning loads of new work very shortly, definitely for Linux, maybe also Windows under Virtualbox. I see you have a machine running each.

4 other users bombed on that work unit too. Except for that one, the last work unit I got from CPDN was at the end of July. I had about an 8-month gap in WCG work units. And Rosetta does not keep up with my machine either. I used to do Seti@home and Malaria, but they are gone.
I can max out all my machines with WCG. Sometimes I can't get enough GPU work, but only 1/4 of the time. All I changed was to make Boinc auto-retry downloads every 30 seconds, and to get 1+0.25 days buffer.

IIRC, none of those go faster than my connection.
Surely you do other things on the internet besides Boinc?
ID: 107712 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1437
Credit: 13,693,695
RAC: 0
Message 107713 - Posted: 9 Nov 2022, 4:13:19 UTC - in response to Message 107692.  
Last modified: 9 Nov 2022, 4:15:52 UTC

Store at least 0.50 days of work
Store up to an additional 1.5 days of work.

There are 5 projects. Is that really so large a cache?
For more than a couple of projects, yes.
The whole idea of a cache is to keep your system busy if a project has issues, and you are attached to only that project. The odds of all your projects having issues at the same time with that many projects is bugger all, so there is no need for a cache.

With more than a couple of projects, Store at least 0.05 days of work and Store up to an additional 0.01 days of work is plenty, and it means your Resource share settings will be met, much, much sooner.
Grant
Darwin NT
ID: 107713 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107714 - Posted: 9 Nov 2022, 4:17:10 UTC - in response to Message 107713.  

For more than a couple of projects, yes.
The whole idea of a cache is to keep your system busy if a project has issues, and you are attached to only that project. The odds of all your projects having issues at the same time with that many projects is bugger all, so there is no need for a cache.

With more than a couple of projects, Store at least 0.05 days of work and Store up to an additional 0.01 days of work is plenty, and it means your Resource share settings will be met, much, much sooner.
I have about 15 projects, but I only run about 4 at a time, and sometimes half those are ones with rare work. So I have a big cache so I get some of the rare work instead of just the other projects filling in.
ID: 107714 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jean-David Beyer

Send message
Joined: 2 Nov 05
Posts: 168
Credit: 5,391,098
RAC: 3,766
Message 107735 - Posted: 10 Nov 2022, 14:17:18 UTC - in response to Message 107713.  

Store at least 0.50 days of work
Store up to an additional 1.5 days of work.

There are 5 projects. Is that really so large a cache?

For more than a couple of projects, yes.
The whole idea of a cache is to keep your system busy if a project has issues, and you are attached to only that project. The odds of all your projects having issues at the same time with that many projects is bugger all, so there is no need for a cache.

With more than a couple of projects, Store at least 0.05 days of work and Store up to an additional 0.01 days of work is plenty, and it means your Resource share settings will be met, much, much sooner.


You think odds of my projects having issues at the same times are "bugger all?" I have that problem right now. No work units from CPDN since end of July, No work units from Rosetta in about a week. Spotty work from WCG. Having a queue of about a week might take care of Rosetta and now (after six months) WCG. But I do not like that. I much preferred when all these projects supplied work units whenever my Boinc-client asked for it.
ID: 107735 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Link
Avatar

Send message
Joined: 4 May 07
Posts: 346
Credit: 382,349
RAC: 0
Message 107737 - Posted: 10 Nov 2022, 15:52:12 UTC - in response to Message 107693.  

That and backup projects should be set to 0% and not something like 0.xx%.
Can't do that with Milkyway, you get 40 seconds of work.
That highly depends on wether you ask for CPU or GPU work and also on your GPU, I'd get there at least 9.5 minutes of work for my GTX 275.
.
ID: 107737 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Link
Avatar

Send message
Joined: 4 May 07
Posts: 346
Credit: 382,349
RAC: 0
Message 107738 - Posted: 10 Nov 2022, 16:16:14 UTC - in response to Message 107714.  
Last modified: 10 Nov 2022, 16:18:17 UTC

I have about 15 projects, but I only run about 4 at a time, and sometimes half those are ones with rare work. So I have a big cache so I get some of the rare work instead of just the other projects filling in.
I follow a different strategy: projects with rare work or overloaded servers (so in my case WCG & Rosetta) have obviously more than enough computing ressources at the moment, so i set them to NNT and allow new work from those, that can offer continuous supply with work, i.e. they actually need more computing power than they currently get.
.
ID: 107738 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jean-David Beyer

Send message
Joined: 2 Nov 05
Posts: 168
Credit: 5,391,098
RAC: 3,766
Message 107739 - Posted: 10 Nov 2022, 21:47:57 UTC - in response to Message 107738.  

I follow a different strategy: projects with rare work or overloaded servers (so in my case WCG & Rosetta) have obviously more than enough computing ressources at the moment, so i set them to NNT and allow new work from those, that can offer continuous supply with work, i.e. they actually need more computing power than they currently get.


That is a very interesting strategy. I must think more about it.
ID: 107739 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Link
Avatar

Send message
Joined: 4 May 07
Posts: 346
Credit: 382,349
RAC: 0
Message 107741 - Posted: 11 Nov 2022, 9:37:16 UTC - in response to Message 107739.  
Last modified: 11 Nov 2022, 9:38:50 UTC

That is a very interesting strategy. I must think more about it.
Well, it has lots of advantages, for myself and the projects:

- I don't need to watch BOINC too carefully or even spend my time on pressing the retry button, forcing additional scheduler requests etc.
- I don't add load to already overloaded project servers
- I don't "steal" work from unattended computers without (working) backup projects, so less idle computers, more science done

Of course there are some limits for this strategy, Moo! for example was the last BOINC project supporting my HD3850, so even if it would have some issues, there was simply no other project to which I could move that GPU (except crunching directly for distributed.net of course). Users of for example Macs with Apple CPUs might face similar issues. But a standard x86_64 CPU running Windows or Linux can be moved nearly to any project.
.
ID: 107741 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107743 - Posted: 11 Nov 2022, 10:53:34 UTC - in response to Message 107737.  

That highly depends on wether you ask for CPU or GPU work and also on your GPU, I'd get there at least 9.5 minutes of work for my GTX 275.
Even my 10 year old 280X cards do a MW task in under 40 seconds.
ID: 107743 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mr P Hucker
Avatar

Send message
Joined: 12 Aug 06
Posts: 1600
Credit: 9,529,341
RAC: 66
Message 107744 - Posted: 11 Nov 2022, 10:55:42 UTC - in response to Message 107741.  

- I don't "steal" work from unattended computers without (working) backup projects, so less idle computers, more science done
Not my fault Krembil can't get it's act together.
ID: 107744 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
kotenok2000
Avatar

Send message
Joined: 22 Feb 11
Posts: 211
Credit: 244,189
RAC: 5
Message 107745 - Posted: 11 Nov 2022, 10:56:41 UTC - in response to Message 107743.  
Last modified: 11 Nov 2022, 11:04:39 UTC

My GTX 1650 does Milkyway task in about 6 minutes.
That feeling when 2019 GPU is slower than 2013 gpu.
ID: 107745 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 . . . 253 · 254 · 255 · 256 · 257 · 258 · 259 . . . 272 · Next

Message boards : Number crunching : Problems and Technical Issues with Rosetta@home



©2024 University of Washington
https://www.bakerlab.org