WRT grids and superclusters, aren't you overlooking one fact here?
Therefore, if he was adding nodes to the cluster, then that would show as an increase in the number of hosts.
We are seeing a recent (about the last 2.5 months) disparity in that new host signups have climbed dramatically while new user signups have not.
Here I'm speaking of the project aggregates, not of a particular account.
Relying on Boincstat user rollups, it appears that the credit contribution of Merlin/Morgane has actually gone down in the period in question, but that the AEI contribution has surged greatly. Possibly there are others, but appears there are none of remotely that scale of current contribution which appear under single user accounts.
LIGO Livermore did surge in mid-May, but that contribution is not on the scale to move project aggregate results.
Unfortunately the BOINCstat user host number is not very helpful on the big accounts, as it appears to clip its reporting at 127.
WRT grids and superclusters, aren't you overlooking one fact here?
Therefore, if he was adding nodes to the cluster, then that would show as an increase in the number of hosts.
We are seeing a recent (about the last 2.5 months) disparity in that new host signups have climbed dramatically while new user signups have not.
Here I'm speaking of the project aggregates, not of a particular account.
Unfortunately the BOINCstat user host number is not very helpful on the big accounts, as it appears to clip its reporting at 127.
OK, I get your drift now....
Agreed that when it comes to trying to figure out Participant and host usage and performance patterns, the readily available data makes it so you have to 'read between the lines' a little bit. ;-)
I think it's pretty clear that from the typical individual user POV, the advent of the Quads and really fast Duals have significantly upped the FLOP capacity of BOINC overall.
Likewise, it sure looks like many grid and supercluster owner/operators have been expanding their horsepower as well, as indicated by the increase in the Host/User Ratio, at least on some projects.
Either way, it sure makes it so the WU's get knocked off in pretty short order! :-D
Relying on Boincstat user rollups, it appears that the credit contribution of Merlin/Morgane has actually gone down in the period in question, but that the AEI contribution has surged greatly. Possibly there are others, but appears there are none of remotely that scale of current contribution which appear under single user accounts.
I recall seeing here a while ago that after several years of flawless running, Merlin was shedding a machine a week or so. With 180 machines in Merlin and 615 faster ones in Morgane, unless the Merlin machines have started dieing significantly faster or the MOrgane blades are starting to reach end of life I wouldn't expect a the decline to be very large.
The number of new users per month has remained almost constant. However the number of new hosts per month as seen in the quoted diagrams has risen sharply, I wonder whether this could be caused by the way the E@H app is used in some science grids. Is it possible that the hosts are attached and some time later detached again, maybe repeatedly for the same host, leading to a distortion of the statistics?
I finally looked at the units again, and have a new guess.
The problem is that an eyeball average of Stef's new hosts graph suggests that about 4000 new hosts are registering per day, which at even a modest success and retention rate would suggest a far higher rate of active hosts increase than we in fact see.
My guess is that a large fraction of the "new hosts" are involuntary new host IDs assigned during transient events to continuing use machines. I had such an episode during the last month for my host 1279025 (not its original host number). The result was that I contributed one of the thousands of new hosts on the graph for May 12, but had never intentionally stopped participating.
I no longer recall what event happened that day, but my new guess is that some source of such re-issues became abruptly more common in mid-March, and perhaps yet another in Mid-May.
The grids I know about simply have nowhere near enough hosts to have their genuine add rate be a material contribution to this number.
This effect would even falsely raise the active host account somewhat--depending on how the merger process is handled and on whether users experiencing this type of event take the time to merge their previous with their new host ID.
All other theories aside, in my situation I ran Cosmo at nearly 100% but that project has been unable to supply steady work, so I've been running EaH nearly 100%. Maybe other people find themselves in the same situation.
All other theories aside, in my situation I ran Cosmo at nearly 100% but that project has been unable to supply steady work, so I've been running EaH nearly 100%. Maybe other people find themselves in the same situation.
I agree with that, I was running a mix, but Cosmo has gotten to be unreliable. I also have been blessed with fairly cool weather so far this spring/summer. Once the temps start to stay above 75 drg, I shut down machines to avoid having to run the AC in my office.
All other theories aside, in my situation I ran Cosmo at nearly 100% but that project has been unable to supply steady work, so I've been running EaH nearly 100%. Maybe other people find themselves in the same situation.
I've been running Milkyway & Einstein 50/50 on my Mac, but with Milkyway having problems the last week the majority of my time has been spent on Einstein. Though I have noticed more and more of a lag in my credits trickling in...seems other hosts paired with my WU's are taking longer to return results. I've never had 6,000 pending credits before! It will be interesting to see how overall the project holds up with the summer heat waves...
Even as temperatures are rising in Europe, there's another record in credits granted per day for the E@H project: ca 17 Mio credits, and 16.mio credits averaged over the last few days ==> 160 TFlops displayed on the server status page.
RE: WRT grids and
)
We are seeing a recent (about the last 2.5 months) disparity in that new host signups have climbed dramatically while new user signups have not.
Here I'm speaking of the project aggregates, not of a particular account.
Relying on Boincstat user rollups, it appears that the credit contribution of Merlin/Morgane has actually gone down in the period in question, but that the AEI contribution has surged greatly. Possibly there are others, but appears there are none of remotely that scale of current contribution which appear under single user accounts.
LIGO Livermore did surge in mid-May, but that contribution is not on the scale to move project aggregate results.
Unfortunately the BOINCstat user host number is not very helpful on the big accounts, as it appears to clip its reporting at 127.
RE: RE: WRT grids and
)
OK, I get your drift now....
Agreed that when it comes to trying to figure out Participant and host usage and performance patterns, the readily available data makes it so you have to 'read between the lines' a little bit. ;-)
I think it's pretty clear that from the typical individual user POV, the advent of the Quads and really fast Duals have significantly upped the FLOP capacity of BOINC overall.
Likewise, it sure looks like many grid and supercluster owner/operators have been expanding their horsepower as well, as indicated by the increase in the Host/User Ratio, at least on some projects.
Either way, it sure makes it so the WU's get knocked off in pretty short order! :-D
Alinator
RE: Relying on Boincstat
)
I recall seeing here a while ago that after several years of flawless running, Merlin was shedding a machine a week or so. With 180 machines in Merlin and 615 faster ones in Morgane, unless the Merlin machines have started dieing significantly faster or the MOrgane blades are starting to reach end of life I wouldn't expect a the decline to be very large.
RE: The number of new users
)
I finally looked at the units again, and have a new guess.
The problem is that an eyeball average of Stef's new hosts graph suggests that about 4000 new hosts are registering per day, which at even a modest success and retention rate would suggest a far higher rate of active hosts increase than we in fact see.
My guess is that a large fraction of the "new hosts" are involuntary new host IDs assigned during transient events to continuing use machines. I had such an episode during the last month for my host 1279025 (not its original host number). The result was that I contributed one of the thousands of new hosts on the graph for May 12, but had never intentionally stopped participating.
I no longer recall what event happened that day, but my new guess is that some source of such re-issues became abruptly more common in mid-March, and perhaps yet another in Mid-May.
The grids I know about simply have nowhere near enough hosts to have their genuine add rate be a material contribution to this number.
This effect would even falsely raise the active host account somewhat--depending on how the merger process is handled and on whether users experiencing this type of event take the time to merge their previous with their new host ID.
Well, it's now June, when
)
Well, it's now June, when people normally would be cutting back on their BOINC participation. Yet, I see that we've hit a new record of 152 TFLOP's.
If this keeps up, we'll be done with the 'R3 run before we know it.
I just hope that the next run is ready. ;)
RE: Good news: S5R3 search
)
I was too optimistic. The WUs will run out only before the middle of July.
All other theories aside, in
)
All other theories aside, in my situation I ran Cosmo at nearly 100% but that project has been unable to supply steady work, so I've been running EaH nearly 100%. Maybe other people find themselves in the same situation.
RE: All other theories
)
I agree with that, I was running a mix, but Cosmo has gotten to be unreliable. I also have been blessed with fairly cool weather so far this spring/summer. Once the temps start to stay above 75 drg, I shut down machines to avoid having to run the AC in my office.
RE: All other theories
)
I've been running Milkyway & Einstein 50/50 on my Mac, but with Milkyway having problems the last week the majority of my time has been spent on Einstein. Though I have noticed more and more of a lag in my credits trickling in...seems other hosts paired with my WU's are taking longer to return results. I've never had 6,000 pending credits before! It will be interesting to see how overall the project holds up with the summer heat waves...
Human Stupidity Is Infinite...
Even as temperatures are
)
Even as temperatures are rising in Europe, there's another record in credits granted per day for the E@H project: ca 17 Mio credits, and 16.mio credits averaged over the last few days ==> 160 TFlops displayed on the server status page.
Bikeman