BOINC Pentathlon 2019

[Windows7ge]

4 minutes ago, leadeater said:

Yea screw that, ESXi is just a tick box

I haven't even gone over what the procedure is for PROXMOX . (I won't)

Still if I bridged the VM's with my local network none of the MAC spoofing or NAT configuring should be necessary...in theory...I'd have to test it. Which I may just in case LHC comes up next year.

 

I will have ALL the credits. I realized that within the time frame that comes before a event starts you could (for all intensive purposes) bunker effectively an infinite number of WU's with VM's just run as many as the system will let you at once. When they run out of work shut them down spin up another 4, 6, or 8, rinse, repeat until the event starts. Just like that you'll drop a million WU's.

 

Did anybody think of this previously? Of course it's only really going to work with the right OS's/software so it's not a plausible method for everyone.

[leadeater]

4 minutes ago, Windows7ge said:

When they run out of work shut them down spin up another 4, 6, or 8, rinse, repeat until the event starts. Just like that you'll drop a million WU's.

 

Did anybody think of this previously? Of course it's only really going to work with the right OS's/software so it's not a plausible method for everyone.

Most people do it via multiple boinc instances, same method without the VMs. Both work, which is better I don't know. I find multi instance management annoying, had to do that to bunker different projects and not clash/early release etc.

[Windows7ge]

23 minutes ago, leadeater said:

Most people do it via multiple boinc instances, same method without the VMs. Both work, which is better I don't know. I find multi instance management annoying, had to do that to bunker different project and no clash/early release etc.

I tried the multi-instance and Windows Server 2016 just went "Nope." shat out a bunch of errors and then proceeded to break my working instance of BOINC so I don't plan on trying that again.

 

As far as my knowledge goes native is always better but if tasks clash fighting over processor time then it doesn't do you any good. Running containers/docker may be the happy medium. Perhaps that would do it's own thing to manage who gets to communicate with the kernel when and keep clashes from happening but maybe not. I'd have to test it. At that point the only other option is full blown VMs. The downside being there is definite overhead. However if you have the time available before the event it would allow you to hold more WU's than running 2 or 3 instances natively.

 

Of course you could take a HDD, convert it to GPT, divide it into 8+ partitions, install Linux to all of them then fill up your WU queue complete them all then start the next OS, rinse repeat. That would probably be ultimately ideal.

 

...hmn...

 

Unrelated:

Running 13 days now and I'm over 1/2 to Diamond. Sweet.

 

I'm also quickly approaching Ithanul's current BOINC combined record.

Mine: 

I can't compete with leadeater (yet) so I guess I have to compete with you.

 

Scratch that apparently I can look up Ithanul's name on the BOINCstats website. That's just his WCG score his combined is over 42M.

 

jeez I can't compete with either of you.

[leadeater]

@Windows7ge

So..... bad news, every ATLAS task on the VMs get computation errors so nested vbox seems like a no go.

[leadeater]

16 minutes ago, Windows7ge said:

Running 13 days now and I'm over 1/2 to Diamond. Sweet.

 

I'm also quickly approaching Ithanul's current BOINC combined record.

I can actually do over 6mil WCG in 24 hours 

[Windows7ge]

43 minutes ago, leadeater said:

@Windows7ge

So..... bad news, every ATLAS task on the VMs get computation errors so nested vbox seems like a no go.

Is that Windows in ESXi or Linux in ESXi? Do you know what VB version? It's possible with some testing it could still work under certain circumstances.

 

30 minutes ago, leadeater said:

I can actually do over 6mil WCG in 24 hours 

That's only because you have access to the surplus of a data-center. Speaking of. One of the places I work I spotted two HP DL380p Gen 8's on a office cart. I get the feeling they were getting decommissioned. There's a whole 0.01% chance I could convince the network guy to let me take them off their hands. I have to give it a shot.

[leadeater]

1 minute ago, Windows7ge said:

Is that Windows in ESXi or Linux in ESXi? Do you know what VB version? It's possible with some testing it could still work under certain circumstances.

ESXi host, Linux VM guest, vbox 5.2

[leadeater]

2 minutes ago, Windows7ge said:

I work I spotted two HP DL380p Gen 8's on a office cart. I get the feeling they were getting decommissioned. There's a whole 0.01% chance I could convince the network guy to let me take them off their hands. I have to give it a shot.

Worth a try, those are great

[porina]

One machine fell over overnight but rest are going strong on WCG. Is it still worth adding more to LHC? Now concerned as time is tight if it will return and validate in time to count? This would be in addition to existing firepower, could add more quad cores although older ones... Broadwell/Haswell.

[leadeater]

53 minutes ago, porina said:

One machine fell over overnight but rest are going strong on WCG. Is it still worth adding more to LHC? Now concerned as time is tight if it will return and validate in time to count? This would be in addition to existing firepower, could add more quad cores although older ones... Broadwell/Haswell.

Looking at current output from the teams above us in Sprint I think it's unrealistic to expect to gain any places so maybe WCG is better?

[porina]

2 minutes ago, leadeater said:

Looking at current output from the teams above us in Sprint I think it's unrealistic to expect to gain any places so maybe WCG is better?

Ok, I have got one running with estimated 4 hours. I'll let it go WCG after that.

[GOTSpectrum]

Yesterday's report has been published.

 

https://www.seti-germany.de/forum/content/1066-BOINC-Pentathlon-2019-Day-13

[leadeater]

5 minutes ago, Ben Quigley said:

Yesterday's report has been published.

 

https://www.seti-germany.de/forum/content/1066-BOINC-Pentathlon-2019-Day-13

AF dropped 2.76mil a couple of hours ago pretty much destroying our marathon hopes of passing them.

[GOTSpectrum]

1 minute ago, leadeater said:

AF dropped 1.76mil a couple of hours ago pretty much destroying our marathon hopes of passing them.

Realistically it would have made little difference to the overall position

[leadeater]

1 minute ago, Ben Quigley said:

Realistically it would have made little difference to the overall position

True, just would have been nice.

[GOTSpectrum]

Just now, leadeater said:

True, just would have been nice.

There's always next year

[leadeater]

1 minute ago, Ben Quigley said:

There's always next year

Will have to do much better at recruiting next year, I highly doubt I will be able to squirrel away a similar amount of hardware again. I'd be happy to put up a contribution towards prizes, prizes that get better the higher the team ranks i.e. outside top 10 best prize is free game, inside top 10 free CPU/GPU, inside top 10 I'll ship out a server (or something).

 

Nothing quite gets people moving like free stuff 

[GOTSpectrum]

1 minute ago, leadeater said:

Will have to do much better at recruiting next year, I highly doubt I will be able to squirrel away a similar amount of hardware again. I'd be happy to put up a contribution towards prizes, prizes that get better the higher the team ranks i.e. outside top 10 best prize is free game, inside top 10 free CPU/GPU, inside top 10 I'll ship out a server (or something).

 

Nothing quite gets people moving like free stuff 

Agreed, I will plan it better next year with Jake

[Windows7ge]

2 minutes ago, leadeater said:

I'll ship out a server (or something).

Free server you say?  I've been looking to grow my collection.

[leadeater]

3 minutes ago, Windows7ge said:

Free server you say?  I've been looking to grow my collection.

Yep, and it fell off the back of a truck  

[Windows7ge]

5 minutes ago, leadeater said:

Yep, and it feel off the back of a truck  

There's a lot you can salvage from a server that fell off the back of a truck ^2+1

[GOTSpectrum]

22 minutes ago, Windows7ge said:

Free server you say?  I've been looking to grow my collection.

 

I need to expand mine too... only doing to have 2 processors soon

[leadeater]

5 minutes ago, Ben Quigley said:

I need to expand mine too... only doing to have 2 processors soon

What's interesting is how the landscape of CPUs and old servers has changed within the last 1-2 years. Now with Ryzen and TR old servers aren't actually as great anymore, you can see the effects of that looking at last years Pent scores. This year seems to have much, much, higher hourly outputs comparatively.

[Windows7ge]

3 minutes ago, leadeater said:

What's interesting is how the landscape of CPUs and old servers has changed within the last 1-2 years. Now with Ryzen and TR old servers aren't actually as great anymore, you can see the effects of that looking at last years Pent scores. This year seems to have much, much, higher hourly outputs comparatively.

AMD may have leveled the playing field in that regard but I still haven't seen them designing their RYZEN or TR chips to support multiple sockets and when density is important you still can't beat the price/density of a used dual or quad socket Intel server. Of course that depends on your use case.

 

What I'm surprised about is people are still recycling X5000 series servers for gaming PC's. They'd get more out of RYZEN.

[leadeater]

1 minute ago, Windows7ge said:

What I'm surprised about is people are still recycling X5000 series servers for gaming PC's. They'd get more out of RYZEN.

Yea it's these and the lower end LGA2011 I was meaning, when your dual socket system has 12/16 cores and TR has that or 24/32 with higher clocks it takes the shine off that lovely server. Still a lot of other niceties about a proper server platform but you can do a heck of a lot with Ryzen used or new, even in a rackmount chassis.