It has actually been a very cool spring in the middle of the USA (I'm in Missouri). Lots of rain and cooler than normal temps. So, perhaps for the middle of the USA, the summer heat up has not even begun yet. The middle of the USA is not likely a vast source of power for BOINC projects thought :<. We had a big flood in 1993 and this year is beginning to look a little similar.
But, last year we had lots of rain early and then a very dry late summer. You just can never predict what the weather will do. Glad I'm not a farmer.
One more thing. I concur with Bikeman that it appears many science grids are joining the project. Many of the newer (higher #) computers I am sharing tasks with are anonymous and are running linux. Could be just a bunch of users want to stay hidden or that the science grids tend to stay anonymous.
It has actually been a very cool spring in the middle of the USA (I'm in Missouri). Lots of rain and cooler than normal temps. So, perhaps for the middle of the USA, the summer heat up has not even begun yet. The middle of the USA is not likely a vast source of power for BOINC projects thought :<. We had a big flood in 1993 and this year is beginning to look a little similar.
But, last year we had lots of rain early and then a very dry late summer. You just can never predict what the weather will do. Glad I'm not a farmer.
Happy Crunching to all.
I'm just south of the border from you in NW Arkansas. It's been a very cool summer here so far, but more rain than at any time since I was born. I'm glad I'm not a farmer either. Most of them in the eastern part of the state are currently underwater.
It has actually been a very cool spring in the middle of the USA (I'm in Missouri). Lots of rain and cooler than normal temps. So, perhaps for the middle of the USA, the summer heat up has not even begun yet. The middle of the USA is not likely a vast source of power for BOINC projects thought :<. We had a big flood in 1993 and this year is beginning to look a little similar.
But, last year we had lots of rain early and then a very dry late summer. You just can never predict what the weather will do. Glad I'm not a farmer.
Happy Crunching to all.
I'm just south of the border from you in NW Arkansas. It's been a very cool summer here so far, but more rain than at any time since I was born. I'm glad I'm not a farmer either. Most of them in the eastern part of the state are currently underwater.
Feel free to send some of that rain over here. For the most part, the whole state of Georgia has been in a drought condition for the past two years.
My personal bet would be more like 70-80 days. Remember, in summer, many users will power down some of their hosts. A lot will depend on when the SSE auto-detecting app for windows will be made official.
My personal bet would be more like 70-80 days. Remember, in summer, many users will power down some of their hosts. A lot will depend on when the SSE auto-detecting app for windows will be made official.
CU
Bikeman
Humm... I wonder. I just spent ten minutes clicking my way through work units returned by my quorum partners on my four hosts. In that obviously non-random sample, the dominant partner type was a grid machine running an up-to-date Linux ap, and the second most common partner type was a windows machine already running 4.36 ap. Windows 4.26 (stock) partners were appreciably less common.
Now I do understand that a single host gets a very non-random sample of partners, but some others have commented here on the rise in importance of the grids.
If the grids are adding hosts without adding new user accounts, it may be that much of the recent added new host count is new grid hosts. And if these new grid hosts tend to devote a much higher fraction of their capacity to Einstein than an average personal user host does, and are modern machines, they may account for a much bigger capacity boost then their fraction of total active hosts would suggest.
And with the Linux stock ap now greatly improved in productivity, the grid machines are at high output whether or not they are following the betas. Even though there is not yet a stock Windows ap released with the great improvements Bernd has released in various betas, the total project output may already have a huge benefit from them.
So my summary guess (I won't dignify it as a conclusion) is that the eventual issue of an improved Windows stock ap will have less impact than we might suppose--perhaps a 20% prompt boost in output. And, assuming that the grid machines are less likely to take summer vacation, I'll guess that the sag this summer will be less than last.
But I understand almost nothing about these grids--so if their computation might suddenly be moved to another purpose, perhaps we might get a precipitate drop.
The grids I saw as partners were:
AEI eScience group, for the German Grid (D-Grid) and the Open Science Grid (OSG)
You're probably right. In fact, if I remember correctly, hosts counts had already dropped appreciably by this time last year. This, we're hitting record numbers.
Quote:
Quote:
My personal bet would be more like 70-80 days. Remember, in summer, many users will power down some of their hosts. A lot will depend on when the SSE auto-detecting app for windows will be made official.
CU
Bikeman
Humm... I wonder. I just spent ten minutes clicking my way through work units returned by my quorum partners on my four hosts. In that obviously non-random sample, the dominant partner type was a grid machine running an up-to-date Linux ap, and the second most common partner type was a windows machine already running 4.36 ap. Windows 4.26 (stock) partners were appreciably less common.
Now I do understand that a single host gets a very non-random sample of partners, but some others have commented here on the rise in importance of the grids.
If the grids are adding hosts without adding new user accounts, it may be that much of the recent added new host count is new grid hosts. And if these new grid hosts tend to devote a much higher fraction of their capacity to Einstein than an average personal user host does, and are modern machines, they may account for a much bigger capacity boost then their fraction of total active hosts would suggest.
And with the Linux stock ap now greatly improved in productivity, the grid machines are at high output whether or not they are following the betas. Even though there is not yet a stock Windows ap released with the great improvements Bernd has released in various betas, the total project output may already have a huge benefit from them.
So my summary guess (I won't dignify it as a conclusion) is that the eventual issue of an improved Windows stock ap will have less impact than we might suppose--perhaps a 20% prompt boost in output. And, assuming that the grid machines are less likely to take summer vacation, I'll guess that the sag this summer will be less than last.
But I understand almost nothing about these grids--so if there computation might suddenly be moved to another purpose, perhaps we might get a precipitate drop.
The grids I saw as partners were:
AEI eScience group, for the German Grid (D-Grid) and the Open Science Grid (OSG)
WRT grids and superclusters, aren't you overlooking one fact here?
IINM, just adding computing nodes to an existing 'cluster' Host ID won't help a lot from a throughput POV. IOW's BOINC doesn't support each node as a separate host, so you would quickly hit the max daily quota. (currently 16 x 4 (or what ever the max CPUs is for EAH now).
Recall when Dr. Allen had Nemo crunching away hard on EAH. Each node was setup as a separate host under an unique HID. I looked at Steffen's account and this appears to be the way he has Merlin setup. Therefore, if he was adding nodes to the cluster, then that would show as an increase in the number of hosts.
It has actually been a very
)
It has actually been a very cool spring in the middle of the USA (I'm in Missouri). Lots of rain and cooler than normal temps. So, perhaps for the middle of the USA, the summer heat up has not even begun yet. The middle of the USA is not likely a vast source of power for BOINC projects thought :<. We had a big flood in 1993 and this year is beginning to look a little similar.
But, last year we had lots of rain early and then a very dry late summer. You just can never predict what the weather will do. Glad I'm not a farmer.
Happy Crunching to all.
One more thing. I concur
)
One more thing. I concur with Bikeman that it appears many science grids are joining the project. Many of the newer (higher #) computers I am sharing tasks with are anonymous and are running linux. Could be just a bunch of users want to stay hidden or that the science grids tend to stay anonymous.
RE: It has actually been a
)
I'm just south of the border from you in NW Arkansas. It's been a very cool summer here so far, but more rain than at any time since I was born. I'm glad I'm not a farmer either. Most of them in the eastern part of the state are currently underwater.
RE: RE: It has actually
)
Feel free to send some of that rain over here. For the most part, the whole state of Georgia has been in a drought condition for the past two years.
Good news: S5R3 search will
)
Good news: S5R3 search will be finished in 40 days with the power of new hosts and users. :)
RE: Good news: S5R3 search
)
Yeah, the new hosts and users are definitely making a difference.
My personal bet would be more
)
My personal bet would be more like 70-80 days. Remember, in summer, many users will power down some of their hosts. A lot will depend on when the SSE auto-detecting app for windows will be made official.
CU
Bikeman
RE: My personal bet would
)
Humm... I wonder. I just spent ten minutes clicking my way through work units returned by my quorum partners on my four hosts. In that obviously non-random sample, the dominant partner type was a grid machine running an up-to-date Linux ap, and the second most common partner type was a windows machine already running 4.36 ap. Windows 4.26 (stock) partners were appreciably less common.
Now I do understand that a single host gets a very non-random sample of partners, but some others have commented here on the rise in importance of the grids.
If the grids are adding hosts without adding new user accounts, it may be that much of the recent added new host count is new grid hosts. And if these new grid hosts tend to devote a much higher fraction of their capacity to Einstein than an average personal user host does, and are modern machines, they may account for a much bigger capacity boost then their fraction of total active hosts would suggest.
And with the Linux stock ap now greatly improved in productivity, the grid machines are at high output whether or not they are following the betas. Even though there is not yet a stock Windows ap released with the great improvements Bernd has released in various betas, the total project output may already have a huge benefit from them.
So my summary guess (I won't dignify it as a conclusion) is that the eventual issue of an improved Windows stock ap will have less impact than we might suppose--perhaps a 20% prompt boost in output. And, assuming that the grid machines are less likely to take summer vacation, I'll guess that the sag this summer will be less than last.
But I understand almost nothing about these grids--so if their computation might suddenly be moved to another purpose, perhaps we might get a precipitate drop.
The grids I saw as partners were:
AEI eScience group, for the German Grid (D-Grid) and the Open Science Grid (OSG)
and
Steffen Grunewald, for Merlin/Morgane
[edited a typo]
You're probably right. In
)
You're probably right. In fact, if I remember correctly, hosts counts had already dropped appreciably by this time last year. This, we're hitting record numbers.
Hmmm... WRT grids and
)
Hmmm...
WRT grids and superclusters, aren't you overlooking one fact here?
IINM, just adding computing nodes to an existing 'cluster' Host ID won't help a lot from a throughput POV. IOW's BOINC doesn't support each node as a separate host, so you would quickly hit the max daily quota. (currently 16 x 4 (or what ever the max CPUs is for EAH now).
Recall when Dr. Allen had Nemo crunching away hard on EAH. Each node was setup as a separate host under an unique HID. I looked at Steffen's account and this appears to be the way he has Merlin setup. Therefore, if he was adding nodes to the cluster, then that would show as an increase in the number of hosts.
Alinator