My last WU was 7/25 it is now 8/6. In order to get WU's I had to increase my network connect time to 9 days. I got plenty of LHC units, a few SETI (no uploads now).
And therein lies the problem. You have 9 days of work for other projects, but Einstein has a 7 day deadline. Very simply you are downloading too much work. The scheduler recognizes your committement to other projects and won't give you more work until that commitement is met.
Try backing off to no more then 2 days and then give it a week without messing with it. (meaning change nothing, just leave it alone). It will sort itself out in the next week, and then will give all the projects their share as set in resource sharing. I run a 1-2 day cache and if you don't mess with it all the time, it does just fine. In science, patience, not constant tinkering is the key.
At least Seti lets you know their hardware/software to handle the uploads is down. Einstein makes you go looking here to find way they are hosed up.
You are kidding us, right??? Seti tells you (frequently) about their planned outages because they seem to have so many of them. Einstein also tells you about their planned outages - very infrequently - because they rarely need to have them.
But what about all the unplanned outages, server failures, configuration mishaps, hardware and software glitches?? Seti seems to have many and you often have to be patient for quite a while for the problem even to be acknowledged, let alone be fixed. On the other hand, EAH staff, particularly Bruce himself, are always highly responsive and visible whenever there is a problem.
Shaktai has given you an excellent explanation of your problem and excellent advice on how to fix it. Please note that EAH is NOT "hosed up" as you assert, and is far more reliable than Seti. Your problem is self made and flows from your misunderstanding of how BOINC (and not EAH) works. To better understand how BOINC works and find out about important things like LTD, STD, (long term debt, short term debt), EDF mode, (earliest deadline first), resource shares, connect to network interval, etc, you should go and have a good read in Paul D Buck's excellent Wiki.
Well as soon as I bitch I get 16 wu's, my laptop is still without any.
I did follow the directions of increasing my network connection period.
that gave me ~ 16 wu's that were all finished in 2 days.
My current dual CPU system does one WU per 6.5 hr or 1 wu / 3.25 hr. That should be enough to keep some wu's in the queue.
I write software for a living, so I know about how Seti started, and started with them almost as soon as they started. I understand where they must spend limited man-hours and hardware resources. I know Bonic is a new concept, but its set up is not as easy and is more confusing that it should be. (This is a Bonic Manager problem.) I also don't remember reading suggested settings when I added Einstein.
I know this is science, but all coders must idiot proof their software enough so the average user can use it. They must also consider that the average user will be running more than one project. (I run 4 of 5 on my server, 5 of 5 on my laptop, and 3 of 5 on my computer I gave my parents). So Einstein should work with a 20% share. I just wanted to crunch some numbers without having to do an in depth study of how to set up each of my 3 computers.
PS Predictor@home is still not sending out work.
ST
Quote:
Quote:
At least Seti lets you know their hardware/software to handle the uploads is down. Einstein makes you go looking here to find way they are hosed up.
You are kidding us, right??? Seti tells you (frequently) about their planned outages because they seem to have so many of them. Einstein also tells you about their planned outages - very infrequently - because they rarely need to have them.
But what about all the unplanned outages, server failures, configuration mishaps, hardware and software glitches?? Seti seems to have many and you often have to be patient for quite a while for the problem even to be acknowledged, let alone be fixed. On the other hand, EAH staff, particularly Bruce himself, are always highly responsive and visible whenever there is a problem.
Shaktai has given you an excellent explanation of your problem and excellent advice on how to fix it. Please note that EAH is NOT "hosed up" as you assert, and is far more reliable than Seti. Your problem is self made and flows from your misunderstanding of how BOINC (and not EAH) works. To better understand how BOINC works and find out about important things like LTD, STD, (long term debt, short term debt), EDF mode, (earliest deadline first), resource shares, connect to network interval, etc, you should go and have a good read in Paul D Buck's excellent Wiki.
Well as soon as I bitch I get 16 wu's, my laptop is still without any.
So you didn't follow Shaktai's advice?? Prepare yourself for the next deadline dance then :).
Quote:
I did follow the directions of increasing my network connection period.
that gave me ~ 16 wu's that were all finished in 2 days.
My current dual CPU system does one WU per 6.5 hr or 1 wu / 3.25 hr. That should be enough to keep some wu's in the queue.
Where did anyone give you directions to increase your connect interval??
Uhmm.... the 16 units you mention could only have been done in 2 days (actually 2 days 5 hours from your results list) because nothing else was given any access to either of your two CPUs during that time. Elsewhere you state that EAH has a 20% share so you are making it impossible for BOINC to honour your resource share without forever doing the deadline dance. Each time BOINC has to suspend your other projects just to get the excess EAH work done, the large negative LTD that EAH will earn will give you a prolonged deadline drought just like you have been bitching about. It's just gonna keep recurring unless you knock back your connect to network interval as Shaktai suggested.
Here's the maths: 7 days x 20% share = 33.6 hours. You need say 30hrs max EAH work.
Your box can do 2 work units in 6.6hrs. You need 30/3.3 = 9 work units max.
For safety you should not download more than 8 units to your machine if you want to avoid the deadline dance and the deadline drought. Of course you can download more and BOINC will handle it for you if you let it. Your previous postings suggest you don't really enjoy the way BOINC is forced to do it though.
In a way, because other projects often severely overestimate the time taken by their work units, you often have to live with a little excess of EAH just to get sufficient work for the other projects if those other projects (eg seti) are going to be unreliable. But why should you blame EAH for seti's unreliability? Why not simply adopt the attitude that an unreliable project misses out since you have plenty of others to take up the slack? Having multiple projects and a reasonable cache size (eg 1 day) is a perfect way to "reward" reliable projects and to use BOINC actually the way it was designed.
Quote:
I know this is science, but all coders must idiot proof their software enough so the average user can use it. They must also consider that the average user will be running more than one project. (I run 4 of 5 on my server, 5 of 5 on my laptop, and 3 of 5 on my computer I gave my parents). So Einstein should work with a 20% share. I just wanted to crunch some numbers without having to do an in depth study of how to set up each of my 3 computers.
You really need to read enough to understand the difference between BOINC and the projects that run on BOINC. BOINC is a work in progress that has gone way beyond classic Seti. BOINC is designed for multiple projects and that in itself introduces complexities that the developers are gradually learning how to deal with. They are making pretty reasonable progress and if you do decide to do a bit of reading up about it, you will learn to appreciate just how well they are progressing. The varying requirements of the projects that run on BOINC introduce additional complexities of their own. When BOINC has problems dealing with the requirements of a particular project, the developers of BOINC try to adapt perhaps in cooperation with the project developers. You can't just blame the project nor can you make ill-informed statements like "Einstein should work with a 20% share" because that just highlights how ill-informed you are. EAH works extremely well with a 20% share if only you will let it. A user does usually need to read the manual occasionally to understand the finer points of how any complex piece of software works best.
So, why don't you reduce your connect to network interval to 1 day max and patiently watch how BOINC is smart enough to sort it out and start honouring your resource share (eg 20% for EAH). I'm sure you will be pleasantly surprised at how well it will cope, if you are patient. Then go and read up about the changes in BOINC 4.72 which will allow BOINC to refine the often wildly erroneous completion time estimates. When this matures, we can look forward to much more reasonable behaviour in the amount of work downloaded for each project in your project mix.
I'm not suggesting that you install 4.72 as this is decidely an alpha style release but rather that you find out about the envisaged changes and how they will benefit your situation so that you can be encouraged to grab a stable version as soon as they release it. This shouldn't be too far away actually.
RE: My last WU was 7/25 it
)
And therein lies the problem. You have 9 days of work for other projects, but Einstein has a 7 day deadline. Very simply you are downloading too much work. The scheduler recognizes your committement to other projects and won't give you more work until that commitement is met.
Try backing off to no more then 2 days and then give it a week without messing with it. (meaning change nothing, just leave it alone). It will sort itself out in the next week, and then will give all the projects their share as set in resource sharing. I run a 1-2 day cache and if you don't mess with it all the time, it does just fine. In science, patience, not constant tinkering is the key.
Team MacNN - The best Macintosh team ever.
RE: At least Seti lets you
)
You are kidding us, right??? Seti tells you (frequently) about their planned outages because they seem to have so many of them. Einstein also tells you about their planned outages - very infrequently - because they rarely need to have them.
But what about all the unplanned outages, server failures, configuration mishaps, hardware and software glitches?? Seti seems to have many and you often have to be patient for quite a while for the problem even to be acknowledged, let alone be fixed. On the other hand, EAH staff, particularly Bruce himself, are always highly responsive and visible whenever there is a problem.
Shaktai has given you an excellent explanation of your problem and excellent advice on how to fix it. Please note that EAH is NOT "hosed up" as you assert, and is far more reliable than Seti. Your problem is self made and flows from your misunderstanding of how BOINC (and not EAH) works. To better understand how BOINC works and find out about important things like LTD, STD, (long term debt, short term debt), EDF mode, (earliest deadline first), resource shares, connect to network interval, etc, you should go and have a good read in Paul D Buck's excellent Wiki.
Cheers,
Gary.
Well as soon as I bitch I get
)
Well as soon as I bitch I get 16 wu's, my laptop is still without any.
I did follow the directions of increasing my network connection period.
that gave me ~ 16 wu's that were all finished in 2 days.
My current dual CPU system does one WU per 6.5 hr or 1 wu / 3.25 hr. That should be enough to keep some wu's in the queue.
I write software for a living, so I know about how Seti started, and started with them almost as soon as they started. I understand where they must spend limited man-hours and hardware resources. I know Bonic is a new concept, but its set up is not as easy and is more confusing that it should be. (This is a Bonic Manager problem.) I also don't remember reading suggested settings when I added Einstein.
I know this is science, but all coders must idiot proof their software enough so the average user can use it. They must also consider that the average user will be running more than one project. (I run 4 of 5 on my server, 5 of 5 on my laptop, and 3 of 5 on my computer I gave my parents). So Einstein should work with a 20% share. I just wanted to crunch some numbers without having to do an in depth study of how to set up each of my 3 computers.
PS Predictor@home is still not sending out work.
ST
RE: Well as soon as I bitch
)
So you didn't follow Shaktai's advice?? Prepare yourself for the next deadline dance then :).
Where did anyone give you directions to increase your connect interval??
Uhmm.... the 16 units you mention could only have been done in 2 days (actually 2 days 5 hours from your results list) because nothing else was given any access to either of your two CPUs during that time. Elsewhere you state that EAH has a 20% share so you are making it impossible for BOINC to honour your resource share without forever doing the deadline dance. Each time BOINC has to suspend your other projects just to get the excess EAH work done, the large negative LTD that EAH will earn will give you a prolonged deadline drought just like you have been bitching about. It's just gonna keep recurring unless you knock back your connect to network interval as Shaktai suggested.
Here's the maths: 7 days x 20% share = 33.6 hours. You need say 30hrs max EAH work.
Your box can do 2 work units in 6.6hrs. You need 30/3.3 = 9 work units max.
For safety you should not download more than 8 units to your machine if you want to avoid the deadline dance and the deadline drought. Of course you can download more and BOINC will handle it for you if you let it. Your previous postings suggest you don't really enjoy the way BOINC is forced to do it though.
In a way, because other projects often severely overestimate the time taken by their work units, you often have to live with a little excess of EAH just to get sufficient work for the other projects if those other projects (eg seti) are going to be unreliable. But why should you blame EAH for seti's unreliability? Why not simply adopt the attitude that an unreliable project misses out since you have plenty of others to take up the slack? Having multiple projects and a reasonable cache size (eg 1 day) is a perfect way to "reward" reliable projects and to use BOINC actually the way it was designed.
You really need to read enough to understand the difference between BOINC and the projects that run on BOINC. BOINC is a work in progress that has gone way beyond classic Seti. BOINC is designed for multiple projects and that in itself introduces complexities that the developers are gradually learning how to deal with. They are making pretty reasonable progress and if you do decide to do a bit of reading up about it, you will learn to appreciate just how well they are progressing. The varying requirements of the projects that run on BOINC introduce additional complexities of their own. When BOINC has problems dealing with the requirements of a particular project, the developers of BOINC try to adapt perhaps in cooperation with the project developers. You can't just blame the project nor can you make ill-informed statements like "Einstein should work with a 20% share" because that just highlights how ill-informed you are. EAH works extremely well with a 20% share if only you will let it. A user does usually need to read the manual occasionally to understand the finer points of how any complex piece of software works best.
So, why don't you reduce your connect to network interval to 1 day max and patiently watch how BOINC is smart enough to sort it out and start honouring your resource share (eg 20% for EAH). I'm sure you will be pleasantly surprised at how well it will cope, if you are patient. Then go and read up about the changes in BOINC 4.72 which will allow BOINC to refine the often wildly erroneous completion time estimates. When this matures, we can look forward to much more reasonable behaviour in the amount of work downloaded for each project in your project mix.
I'm not suggesting that you install 4.72 as this is decidely an alpha style release but rather that you find out about the envisaged changes and how they will benefit your situation so that you can be encouraged to grab a stable version as soon as they release it. This shouldn't be too far away actually.
Cheers,
Gary.
I've been uploading for over
)
I've been uploading for over 6 hours. Any thoughts? My messages don't reveal anything helpful, just saying that my computer is requesting work.