[QUOTE=paulunderwood;518625]I pledge 2x 6 core 1090Ts [i]after[/i] the summer here in the UK. I will be running Debian 10 by then :smile:[/QUOTE]

Still awaiting for U.K. real summer...where’s the global warming after all?!

Summer there is July 23rd or so, correct? The party lasts a week or so?
With any luck (and support from Vebis and/or scole), we'll be into the 15e portion of the job by 15 August, with much lower memory requirements.

ETA currently 13 Oct, thanks to Vebis' 10 clients joining the parade. So far, I am very pleased with the support and interest in this factorization; pending successful filtering and a matrix that doesn't take a year, I will be happy to organize and shepherd another such GNFS project over winter. Teamwork is wayyy more fun than silently trudging through another C180 job on my own!

Here it’s always winter, summer for 2 mins.

[YOUTUBE]JsJynLpoT0U[/YOUTUBE]

[QUOTE=lukerichards;518588]LOVE this - it's like a leaderboard.

Tempted to set my 12 CPU cores at home onto the task for a bit in addition to the Google Cloud instance ;-)

I could of course get a LOT more out of Google Cloud were I using pre-emptible instances, but restarting it on a daily basis is a hassle.[/QUOTE]
I applaud and support that temptation, sir. Or will you wait until your cloud-dollars are used up?

As for Google-Cloud, how much cheaper are preemptible instances? Perhaps running two preemptible clients makes it worth the manual restart every day for not-much-more cost per day? Then again, I don't know how much work the manual daily startup would be.... but I'm cheap, so I'd be looking for best production per virtual-dollar.

I agree with the others- Seth's stats geekery is adding to the fun of this project. :) :tu:

[QUOTE=VBCurtis;518652]As for Google-Cloud, how much cheaper are preemptible instances? Perhaps running two preemptible clients makes it worth the manual restart every day for not-much-more cost per day? Then again, I don't know how much work the manual daily startup would be.... but I'm cheap, so I'd be looking for best production per virtual-dollar.[/QUOTE]

It's a lot. I think it's like 60% reduction. You make a fair point.

[QUOTE=lukerichards;518690]It's a lot. I think it's like 60% reduction. You make a fair point.[/QUOTE]

For GCP it looks that pre-emptible cost 31% so 3x as much computer for having to restart instances.

Also rocket fuel upgrade from Vebis has us heading straight to orbit.

Expected power outage tonight, server likely down a few hours
 

Reminder:
The campus building that houses our server is expected to lose power tonight due to construction.
The warning emailed to us says only "some time between 11pm and 7am" power will be cut for new electrical equipment to be installed.
I expect to arrive to campus around 7:30am USA-Pacific time, and will get the machine back up at that time.

Clients will simply stall while waiting for a reply, retrying every 10 seconds. You don't need to shut down nor restart anything, but your cores may be idle a few hours tonight.

This is the third such outage this academic year, so hopefully trend will hold and we'll make it another 3ish months before another one!

Currently 140M relations found, 5.2%. ETA is 29 Sep, still falling.

I am confused to report that power was not cut last night. I'll be annoyed if this means they cut it some time over this weekend, but I suppose we can't ask for too much when using free university power & IP-addresses.

Looking forward to seeing next stats breakdown table....

The server seemed to handle the rush fine; 'top' didn't indicate a backup or heavy load with gzip or python. Vebis was still running too, so there were ~50 concurrent clients including your boost, multiple results files per minute.

pumpkin also joined today.

I'll let Seth share the details.

Boost might come from our friend Thomas not from usual suspects since I know they are running a few challenges across the Boinc world.