I don't know if this has ever been suggested before or considered by the developers, but might there be a way to change the function of various servers based on need? I notice that there are now six transitioners, which has reduced the "Transitioner backlog" to 0 hours, and there are 300K+ "results" (work) ready to send. A fairly simple set of conditions might be set (e.g. when "Ready to send" is G.T. 100K and "Transitioner backlog" is L.T. 1 hour), one or more of the of excess servers' might be switched by the software to a differentfunction, perhaps even on the fly.

I don't know if this has ever been suggested before or considered by the developers, but might there be a way to change the function of various servers based on need? I notice that there are now six transitioners, which has reduced the "Transitioner backlog" to 0 hours, and there are 300K+ "results" (work) ready to send. A fairly simple set of conditions might be set (e.g. when "Ready to send" is G.T. 100K and "Transitioner backlog" is L.T. 1 hour), one or more of the of excess servers' might be switched by the software to a differentfunction, perhaps even on the fly.

Good idea. Same sofware on some machines and they can switch between validation, backlog and splitter. Take some machines which do the same as now and take some which will switch. F.e. take splitter and assimilator and give them differnet jobs if possible.

---

WARR - Wissenschaftliche Arbeitsgemeinschaft für Raketentechnik und Raumfahrt
(WARR - scientific working group for rocket technology and space travel)

According to the Status Page there are presently six (6) transitioners and a transitioner backlog of zero (0) hours; there are four (4) validators and almost 700K results waiting for validation. Isn't there some way to switch the function of the servers to do what most needs to be done? Rom? Ron? Matt?

I don't know if this has ever been suggested before or considered by the developers, but might there be a way to change the function of various servers based on need? I notice that there are now six transitioners, which has reduced the "Transitioner backlog" to 0 hours, and there are 300K+ "results" (work) ready to send. A fairly simple set of conditions might be set (e.g. when "Ready to send" is G.T. 100K and "Transitioner backlog" is L.T. 1 hour), one or more of the of excess servers' might be switched by the software to a differentfunction, perhaps even on the fly.

I am sure that Mat would appreciate it if you donated your time to getting the systems to do that. When all the people working on it are also doing other things at SSL and have lives away from the SSL I am sure that they would appreciate it.

But then like most of us you may not know the programmig that goes into it. But maby you have the pull at Sun to get them to donate some NEW servers that will do it all faster. I am sure that they could be needed when Classic is closed down.

---

Pizza@Home Rays Place Rays place Forums

I don't know if this has ever been suggested before . . .

I am sure that Mat would appreciate it if you donated your time to getting the systems to do that. When all the people working on it are also doing other things at SSL and have lives away from the SSL I am sure that they would appreciate it.

But then like most of us you may not know the programmig that goes into it. But maby you have the pull at Sun to get them to donate some NEW servers that will do it all faster. I am sure that they could be needed when Classic is closed down.

Well, Ray, I simply put the idea out there for discussion. As for new servers, the ones they have are not being used as efficiently as they could be. Though I've done some programming (20-40 years ago) and I know the concept of setting conditions to do the task of switching jobs is not that difficult, I wasn't making demands, just talking.

I don't know if this has ever been suggested before . . .

I am sure that Mat would appreciate it if you donated your time to getting the systems to do that. When all the people working on it are also doing other things at SSL and have lives away from the SSL I am sure that they would appreciate it.

But then like most of us you may not know the programmig that goes into it. But maby you have the pull at Sun to get them to donate some NEW servers that will do it all faster. I am sure that they could be needed when Classic is closed down.

Well, Ray, I simply put the idea out there for discussion. As for new servers, the ones they have might not be used as efficiently as they could be. Though I've done some programming (20-40 years ago) and I know the concept of setting conditions to do the task of switching jobs is not that difficult, I wasn't making demands, just talking.

See next post

---

Pizza@Home Rays Place Rays place Forums

Well, Ray, I simply put the idea out there for discussion. As for new servers, the ones they have might not be used as efficiently as they could be. Though I've done some programming (20-40 years ago) and I know the concept of setting conditions to do the task of switching jobs is not that difficult, I wasn't making demands, just talking.

20-40 years ago for the programming.

That sounds like me, but I was on a Mainframe aslo, a completely differant world than a PC. But you are correct, if the systems could switch to whare they are needed the most things would have keep close to caught up and not in this mess.

With the current problem being too many directories look like the only way may be to close the program down for a day or two and have people come back on line again slowley. But that would not work as almost everyone would want to be in the first wave to reconnect and end up with the same thing again.

In another thread someone said that thay could limit the new work to each PC while this is strightening out. That would cause the ones not getting enoughf work to just keep trying to connect over and over.

Ray

---

Pizza@Home Rays Place Rays place Forums

I was the person who suggested limiting the new work If you think about it a computer connecting every so often to request new work but being rejected until he is allowed more is a lot less strain that for example my computer being assigned 60 WU's and each of them connecting when they countdown to try downloading but not being sucessful and having to start the countdown again

---

According to the Status Page there are presently six (6) transitioners and a transitioner backlog of zero (0) hours; there are four (4) validators and almost 700K results waiting for validation. Isn't there some way to switch the function of the servers to do what most needs to be done? Rom? Ron? Matt?

The problem isn't the number of validators (or file_deleters) as much as it is just plain too many files in one directory making the process glacially slow, no matter how many machines are assigned to the task.

... and at this point, changing the storage method still means sucking up a whole stack of files and dumping 'em in a different spot.

---

The problem isn't the number of validators (or file_deleters) as much as it is just plain too many files in one directory making the process glacially slow, no matter how many machines are assigned to the task.

... and at this point, changing the storage method still means sucking up a whole stack of files and dumping 'em in a different spot.

You may be right about too many files slowing down the process, but shifting assignments of lightly used machines for backlogged tasks might still help. Some weeks ago the validator queue got backed up for days, but dissappeared in a very short time when they did something (I didn't see what was done, or maybe I didn't understand it). Still, with 6 transitioners assigned and a transitioner backlog of zero hours, I would hope that some developer would acknowlege my suggestion. Do you, Ned, see anything wrong with my logic?

The problem isn't the number of validators (or file_deleters) as much as it is just plain too many files in one directory making the process glacially slow, no matter how many machines are assigned to the task.

... and at this point, changing the storage method still means sucking up a whole stack of files and dumping 'em in a different spot.

You may be right about too many files slowing down the process, but shifting assignments of lightly used machines for backlogged tasks might still help. Some weeks ago the validator queue got backed up for days, but dissappeared in a very short time when they did something (I didn't see what was done, or maybe I didn't understand it). Still, with 6 transitioners assigned and a transitioner backlog of zero hours, I would hope that some developer would acknowlege my suggestion. Do you, Ned, see anything wrong with my logic?

See a problem with your logic -- not really.

A good friend long ago taught me something that has been very useful over the years: there are two reasons to try anything:

<li>you think it'll fix the problem</li>
<li>you don't think it'll fix the problem, but you might get lucky (and it's easy to try)</li>

To which I'll add: sometimes "lucky" is better than "good"....

I'm not sure that the transitioners really add load when there is nothing to transition, so turning off the transitioners may not make a difference, but the machines with "napping" validators waiting for work could run another process (maybe another file deleter).

But if we assume that 95% of the "bogged down" process is waiting for files, then adding a machine might be just enough to tip from a "growing queue" to a shrinking queue.

... and once those pesky files are out, life will get a whole lot better.

... and it wouldn't be that hard to try.

I trust that "my" idea about too many files slowing things down is correct because it isn't my idea -- Matt Lebofsky said that was the problem, based on experiments he and his co-workers performed on Wednesday.

---

I trust that "my" idea about too many files slowing things down is correct because it isn't my idea -- Matt Lebofsky said that was the problem, based on experiments he and his co-workers performed on Wednesday.

Yes, I read the Technical News page (and I understand that at this time they are too busy with other things to do anything about my proposal). But someday, maybe they can implement a software solution (temporary though it may be) to shift the load of backed up servers to underutilized machines when a particular process gets bogged down.