Combined sieving discussion

[Citrix]

Originally Posted by ltd

At least the server is reachable. ( I am updating the stats at the moment.)
I can not check how good/bad the connectivity is from the outside world cause the server is located at the same provider where i have my DSL connection.
But at least i have no reports of an outage of the server.

Lars

do you want us to test your server or something? Let us know! (We will bring it down )


Citrix

[drakkar67]

voting for the third option but are we strong enough to finish that range in a reasonable time period or will we just be a burden for SOB sievers? I mean will they always wait for us to complete the range we reserved?
And what will be the account name for us? Any ideas?

drakkar67

[NeoGen]

Originally Posted by drakkar67

And what will be the account name for us? Any ideas?

PSP Sievers?

[ltd]

Originally Posted by drakkar67

voting for the third option but are we strong enough to finish that range in a reasonable time period or will we just be a burden for SOB sievers? I mean will they always wait for us to complete the range we reserved?
And what will be the account name for us? Any ideas?

drakkar67

There is no problem with the size of the reservation at the moment. Most of the SOB siever do there first pass sieving at the moment, so we have enough time. I got the confirmation that 101T to 125T is reserved for PSP.

About the account where the results are booked we have to wait and see. Problem is that they normaly are registered against a SOB contributor but joe told me that me might has an idea and that he will contact me again.

[hhh]

Sob needs still 3 months at least to reach 2^50. Until then, the new client might be out and they can just continue. Everybody who is interested (from sob side) is welcome to do joined sieving. I regularily make some publicity for that. No worry. H.

[Brucifer]

buuuummp


Well we are into the third month as it is now March. Our first pass stuff is about finished, as well at the 2nd pass to the point that Lars/Citrix. I would guestimate that no later than the middle of April that we should be ready to move into the combined sieving in a big way

So is this still a "go" with SOB and everyone here?

Bruce

[Mystwalker]

From my side, I don't see why we shouldn't continue the combined sieve.

The efficiency point of view:
It just makes sense to combine those sieves.

The strategical point of view:
SoB has 75T left for their 20M sieve. This range will soon be exhausted. Afterwards, they have two possiblities:
1. Use the new siever to sieve deeper.
2. Put more emphasis on their 50M sieving.

Assuming we (PSP) intensify our combined effort (which will be the case once the last 21T are assigned) and find quite some factors for SoB, I think the chances increase that a lot of SoB sievers also perform a combined search, because they see that it really benefits both projects and not only PSP to a great extend.
Finding a prime would be another positive event to make combined sieving for SoB'ers more attractive.

[hhh]

Originally Posted by Mystwalker

SoB has 75T left for their 20M sieve. This range will soon be exhausted.

I count 85T. And with the 1T/day rule that makes almost three months. It gives us time to close the hole. And perhaps the new client will come out.

Originally Posted by Mystwalker

Afterwards, they have two possiblities:
1. Use the new siever to sieve deeper.
2. Put more emphasis on their 50M sieving.

I would like it to be up to the user, as it is right now at PSP. And, also similar to PSP, I would apreciate a decision to stop non-combined secondpass. We cannot demand SoB to include PSP to their firstpass. It is just too advanced. But joint DC is not asked too much, IMO.
I feel SOB made bad experiences with its forced DC period. Many users just went away, probably never to come back. If there is no choice (no new siever)ok, but not if not absolutely necessary.

Originally Posted by Mystwalker

Finding a prime would be another positive event to make combined sieving for SoB'ers more attractive.

Thats for sure true, but not a necessary condition at all. On the contrary, everybody should be happy to be able to sieve this project in an early phase, not when half of the tests is already done.

We have a good pace; I think we can just wait another month or two and see how we come along.
Then, a poll/discussion might be justified.
H.

[Mystwalker]

Originally Posted by hhh

I count 85T.

I see 1014920 -> 1089800 = 74880.

We cannot demand SoB to include PSP to their firstpass.

First first pass you mean the n<20M sieving, right? In this case, I totally agree, combining probably doesn't even make sense when it comes to efficiency.

I feel SOB made bad experiences with its forced DC period. Many users just went away, probably never to come back.

I wouldn't directly compare PRPing and sieving at SoB - for two reasons:

1. Sieving is a rather manual process and not controlled by the creators of SoB. Hence, it's not really possible to force people to do something. Most decisions seem to be agreements after a thorough discussion.

2. I believe the sievers are more dedicated to the project, not to its stats. Thus, it would be easier to convince them when such a move proves to be more efficient.

Long story short, I predict that after a discussion of efficiency (SoB is already at n>10M with PRPing), they will sieve up to n=50M, which means that a lot of their sievers will join Joe_O et al. starting at (currently) p=134T.

After all, we have no concrete information about the new siever when it comes to sieving p>2^50, 8 vs. 20 k's and so on...

[hhh]

Originally Posted by Mystwalker

I see 1014920 -> 1089800 = 74880.

And 1090000-1100000, isn't it?

You are aware that right know, there is no such thing as 0-20M sieve at SoB, aren't you? All firstpass ranges (the above cited) are done up to 50M; the only thing is this large hole between 134000G and 500 000G, as for p. That just to avoid ambiguities.

The only point about starting over the sieve at 134T would be the reduction of the dat sizeby perhaps 10%, without new, immediatly useful factors. The speed increase is about \sqrt(1,1).
I guess from the SoB perspective, this big hole is not so much a matter as long as they are not at N=15M yet. But perhaps one could comvince them to push it up to 250T.
Anyway, as I said, wait and see.
Yours H.

[Mystwalker]

Originally Posted by hhh

And 1090000-1100000, isn't it?

Those are already assigned and/or completed:

Originally Posted by SoB forum, sieve thread

1090000-1092000 Mat67 [complete] (with 991-50M dat)
1092000-1093000 priwo [complete] (with 991-50M dat)
1093000-1098000 Maniacken [reserved] (with 991-50M dat)
1098000-1099000 royanee [reserved] (with 991-50M dat)
1099000-1100000 engracio [complete] (with 991-50M dat)

You are aware that right know, there is no such thing as 0-20M sieve at SoB, aren't you?

Actually, there is, and I have participated in it from the beginning, and for a long time. Only recently (compared to the overall sieve effort), the 0-50M range has been sieved.
But it seems like most, but not everybody sticks to the new dat.

All firstpass ranges (the above cited) are done up to 50M; the only thing is this large hole between 134000G and 500 000G, as for p.

With a few exceptions, 0-50M doesn't noticably show up until ~800T. Hence, more than 50% of the range have been sieved only with the 0-20M sob.dat.

The only point about starting over the sieve at 134T would be the reduction of the dat sizeby perhaps 10%, without new, immediatly useful factors.

At least in the sub-100T range, vjs (or was it Joe_O) reported that they found more relevant (--> n < 20M) factors per time than with p around 1000T, because some factors weren't found by earlier versions of proth_sieve (or Sobsieve?).

I guess from the SoB perspective, this big hole is not so much a matter as long as they are not at N=15M yet.

That's right, n=20M still needs approx. 3 more years. But then, this hole becomes very important, because the factor density is much higher than at >2^50...