LLRnet and PRPnet servers for automated LLR

[rogue]

I don't know what you mean by "manually submit"? Did the client actually perform those tests?

[whengryphonsfly]

Originally Posted by rogue

I don't know what you mean by "manually submit"? Did the client actually perform those tests?

No, and that's the issue. The server thinks the client is doing those 12 tests, but the client threw an error and rejected them without properly abandoning them. So, those 12 tests are stuck in limbo. I manually ran LLR, so I know what the results are, but I don't know how to get those results to the server.


I think I read somewhere that the server will automatically cancel tests that take 48 hours, and if that's the case it won't matter. But I may be misremembering that, and if so then manual intervention will be required. Hence, my question: is there some way I can bypass prpclient on my end and report the results? Or do we just wait?

[rogue]

Originally Posted by whengryphonsfly

No, and that's the issue. The server thinks the client is doing those 12 tests, but the client threw an error and rejected them without properly abandoning them. So, those 12 tests are stuck in limbo. I manually ran LLR, so I know what the results are, but I don't know how to get those results to the server.


I think I read somewhere that the server will automatically cancel tests that take 48 hours, and if that's the case it won't matter. But I may be misremembering that, and if so then manual intervention will be required. Hence, my question: is there some way I can bypass prpclient on my end and report the results? Or do we just wait?

There is no way to manually report test results. The server will eventually expire the test (there is a Pending Tests page accessible via a browser) that will tell you when it will expire. Once expired it will be sent out again.

[gd_barnes]

920k-960k has been loaded into port 12000.

[MooMoo2]

Originally Posted by gd_barnes

920k-960k has been loaded into port 12000.

Thanks!

Over 19,000 tests were done today, which is TPS's 6th highest daily total of all-time. If we can maintain this, we'll wrap up the effort and hit k=1M in less than 4 weeks. I'll upload the quad-sieved candidates once we reach 990K.

[Bottom Quark]

Originally Posted by gd_barnes

920k-960k has been loaded into port 12000.

Thanks, looks like that batch contains approximately a quarter million candidates.

Looking forward, an n=1.7M test takes about 12 times as long as an n=490K test. For n=3.322M, that increases to 46 times as long as n=490K. Will batches of ~250,000 candidates continue to be released, or will it drop to ~20,000 for n=1.7M and ~5500 for n=3.322M?

[gd_barnes]

Originally Posted by Bottom Quark

Thanks, looks like that batch contains approximately a quarter million candidates.

Looking forward, an n=1.7M test takes about 12 times as long as an n=490K test. For n=3.322M, that increases to 46 times as long as n=490K. Will batches of ~250,000 candidates continue to be released, or will it drop to ~20,000 for n=1.7M and ~5500 for n=3.322M?

The fast tests are why we load so many into the server for n=480K-500K.

There will certainly be fewer loaded at a time for n=1.7M and even less for 3.332M. I/we just find logical cutoffs for loading the servers based on a k-value range due to the nature of the project. There's no hard and fast rule. I prefer not to have to load them in more than about once every 2 weeks so I'll adjust how much is loaded in as we progress based on how much work is being done by everyone.

[MooMoo2]

Originally Posted by Bottom Quark

Thanks, looks like that batch contains approximately a quarter million candidates.

Looking forward, an n=1.7M test takes about 12 times as long as an n=490K test. For n=3.322M, that increases to 46 times as long as n=490K. Will batches of ~250,000 candidates continue to be released, or will it drop to ~20,000 for n=1.7M and ~5500 for n=3.322M?

Originally Posted by gd_barnes

The fast tests are why we load so many into the server for n=480K-500K.

There will certainly be fewer loaded at a time for n=1.7M and even less for 3.332M. I/we just find logical cutoffs for loading the servers based on a k-value range due to the nature of the project. There's no hard and fast rule. I prefer not to have to load them in more than about once every 2 weeks so I'll adjust how much is loaded in as we progress based on how much work is being done by everyone.

For n=1.7M, I'm planning on releasing a 100G k-range at a time (we'll start with k=1-100G, then k=100G-200G, then k=200G-300G, and so on). For n=3.322M, 50G batches will be released.

The amount of work per batch will be more than k=920K-960K for the variable n-range, but the number of candidates loaded at a time will be much smaller than 250,000. Like Gary said, we'll adjust the batch size as we go along.

[MooMoo2]

Originally Posted by MooMoo2

Thanks!

Over 19,000 tests were done today, which is TPS's 6th highest daily total of all-time. If we can maintain this, we'll wrap up the effort and hit k=1M in less than 4 weeks. I'll upload the quad-sieved candidates once we reach 990K.

There were 27,566 completed tests and 7 primes found today, woohoo!

If this can be maintained (fingers crossed), we'll reach k=1M and move on to the quad-sieved candidates in just under 2 weeks.

Gary, could you load k=960K-1M into port 12000? We still have 4-5 days of work if this rate holds, but I'm eager to see that final countdown in the "Count Untested" column.

[gd_barnes]

Port 12000 has been loaded up to k=1M.

[MooMoo2]

Originally Posted by gd_barnes

Port 12000 has been loaded up to k=1M.

Thanks! Just 297,413 tests to go (as of this post) before we can start testing the quad-sieved candidates