BOINC Wish List


log in

Advanced search

Questions and Answers : Wish list : BOINC Wish List

1 · 2 · 3 · 4 . . . 5 · Next
Author Message
Profile Steven Meyer
Avatar
Send message
Joined: 24 Mar 08
Posts: 2301
Credit: 2,999,007
RAC: 4
United States
Message 928209 - Posted: 23 Aug 2009, 16:35:48 UTC
Last modified: 23 Aug 2009, 16:38:42 UTC

Put these items on System Tray menu next to the "Snooze" item.


    1) Advanced/Preferences menu, "Use GPU while computer is in use".
    2) Projects tab, Update button as "Update all projects" to send all "Ready to report" tasks.
    3) Advanced menu, "Do network communications".

If they cannot all be done, then my preferences are as shown here.

+------------------------------------+ | Open BOINC manager | +------------------------------------+ | Snooze | +------------------------------------+ | * Use GPU while computer is in use | | Update all projects | | Do network communications | +------------------------------------+ | About BOINC Manager | +------------------------------------+ | Exit | +------------------------------------+

____________
FireFox Personas


OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13615
Credit: 30,389,506
RAC: 20,894
United States
Message 928339 - Posted: 24 Aug 2009, 3:00:54 UTC - in response to Message 928209.

Being that people abuse the "Do network comms" and the "Update" button, I don't know if it would be a good idea to make it easier by putting in the context menu (right click menu).
____________

CryptokiD
Avatar
Send message
Joined: 2 Dec 00
Posts: 134
Credit: 2,814,936
RAC: 0
United States
Message 928361 - Posted: 24 Aug 2009, 7:33:05 UTC - in response to Message 928339.

respectfully, i do not agree with the abuse of the update button. if the update button can be abused, they would implement a longer minimum wait time between updates. as it is right now, you can updatre all day long if you just wait a few seconds. if you do it every 3 seconds, the server says you updated again too quickly.

on the other hand, i would like to see the server implement a longer minimum wait time between updates and network comms. a minimum of 1 hour. anything sooner and i would like the server to error back saying you updated in too quick of a sucession.

Norwich Gadfly
Avatar
Send message
Joined: 29 Dec 08
Posts: 100
Credit: 488,414
RAC: 0
United Kingdom
Message 928363 - Posted: 24 Aug 2009, 8:19:00 UTC - in response to Message 928361.

respectfully, i do not agree with the abuse of the update button. if the update button can be abused, they would implement a longer minimum wait time between updates. as it is right now, you can updatre all day long if you just wait a few seconds. if you do it every 3 seconds, the server says you updated again too quickly.

on the other hand, i would like to see the server implement a longer minimum wait time between updates and network comms. a minimum of 1 hour. anything sooner and i would like the server to error back saying you updated in too quick of a sucession.


How about increasing the interval before you are allowed to use the update button again ? This would start at say 1 minute, then it would make you wait 2 minutes, then 4 minutes and keep doubling. If you don't use it for a week, the interval would start to come down again.

____________

Profile Steven Meyer
Avatar
Send message
Joined: 24 Mar 08
Posts: 2301
Credit: 2,999,007
RAC: 4
United States
Message 928392 - Posted: 24 Aug 2009, 12:37:50 UTC
Last modified: 24 Aug 2009, 12:42:00 UTC

All good points.

I would rather see it simple to use these features and let the software make the feature unavailable for a period of time (greyed out) if it is being used too much.

+------------------------------------+ | Open BOINC manager | +------------------------------------+ | Snooze | +------------------------------------+ | * Use GPU while computer is in use |
| Update all projects |
| Do network communications | +------------------------------------+ | About BOINC Manager | +------------------------------------+ | Exit | +------------------------------------+

However, nothing more than an hour delay should be imposed.

Also, the delay should be imposed only when the last operation was successful. If the Update button was pressed but the servers are down, why would SETI care if the user presses it again every 10 minutes?
____________
FireFox Personas


Profile Steven Meyer
Avatar
Send message
Joined: 24 Mar 08
Posts: 2301
Credit: 2,999,007
RAC: 4
United States
Message 928396 - Posted: 24 Aug 2009, 12:54:44 UTC
Last modified: 24 Aug 2009, 13:00:46 UTC

By the way, I found that the Update button can be pressed more than once per second, but it appears that it does not actually do all of those operations.

Also, the current Update button is much easier to abuse than my proposal.

These are the steps to submit multiple Updates with the current Update button.

1) Move the mouse to the system try icon
2) Right click
3) Wait for the menu to pop up
4) Click on "Open BOINC Manager"
5) Wait for BM to open
6) Click on "Projects"
7) Select a project
8) Click on "Update"
9) Repeat step 8

While my proposal requires the following.

1) Move the mouse to the system try icon
2) Right click
3) Wait for the menu to pop up
4) Click on "Update all projects"
5) Menu disappears
6) Repeat steps 1 through 5

8/24/2009 5:44:43 AM SETI@home update requested by user 8/24/2009 5:44:46 AM SETI@home update requested by user 8/24/2009 5:44:47 AM SETI@home Sending scheduler request: Requested by user. 8/24/2009 5:44:47 AM SETI@home Reporting 21 completed tasks, not requesting new tasks 8/24/2009 5:44:48 AM SETI@home update requested by user 8/24/2009 5:44:49 AM SETI@home update requested by user 8/24/2009 5:44:51 AM SETI@home update requested by user 8/24/2009 5:44:51 AM SETI@home update requested by user 8/24/2009 5:44:52 AM SETI@home update requested by user 8/24/2009 5:44:54 AM SETI@home Scheduler request completed: got 0 new tasks 8/24/2009 5:44:58 AM SETI@home update requested by user 8/24/2009 5:44:59 AM SETI@home update requested by user 8/24/2009 5:45:00 AM SETI@home update requested by user 8/24/2009 5:45:00 AM SETI@home Sending scheduler request: Requested by user. 8/24/2009 5:45:00 AM SETI@home Not reporting or requesting tasks 8/24/2009 5:45:01 AM SETI@home update requested by user 8/24/2009 5:45:01 AM SETI@home update requested by user 8/24/2009 5:45:01 AM SETI@home update requested by user 8/24/2009 5:45:02 AM SETI@home update requested by user 8/24/2009 5:45:02 AM SETI@home update requested by user 8/24/2009 5:45:06 AM SETI@home Scheduler request completed: got 0 new tasks

____________
FireFox Personas


Norwich Gadfly
Avatar
Send message
Joined: 29 Dec 08
Posts: 100
Credit: 488,414
RAC: 0
United Kingdom
Message 928397 - Posted: 24 Aug 2009, 13:00:55 UTC - in response to Message 928392.

All good points.

I would rather see it simple to use these features and let the software make the feature unavailable for a period of time (greyed out) if it is being used too much.
+------------------------------------+ | Open BOINC manager | +------------------------------------+ | Snooze | +------------------------------------+ | * Use GPU while computer is in use |
| Update all projects |
| Do network communications | +------------------------------------+ | About BOINC Manager | +------------------------------------+ | Exit | +------------------------------------+

However, nothing more than an hour delay should be imposed.

Also, the delay should be imposed only when the last operation was successful. If the Update button was pressed but the servers are down, why would SETI care if the user presses it again every 10 minutes?

Fairy nuff.

____________

OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13615
Credit: 30,389,506
RAC: 20,894
United States
Message 928403 - Posted: 24 Aug 2009, 14:05:30 UTC - in response to Message 928361.

respectfully, i do not agree with the abuse of the update button. if the update button can be abused, they would implement a longer minimum wait time between updates. as it is right now, you can updatre all day long if you just wait a few seconds. if you do it every 3 seconds, the server says you updated again too quickly.


...and how do you think the server sends a message to say that it's been too recent? It certainly isn't the DNS server giving this message.

In fact, the server has to respond to each and every request, even to say, "sorry, you already contacted me less than a minute ago. Please try again later". When the server is having comms problems for whatever reason, and everyone starts hitting this button, they are only adding to the load only to have the server tell them they've contacted too soon.

Sure, it can be argued there's no proof that too many people bother to do this, but I still don't believe it should be made easier for people to pound the servers.

on the other hand, i would like to see the server implement a longer minimum wait time between updates and network comms. a minimum of 1 hour. anything sooner and i would like the server to error back saying you updated in too quick of a sucession.


That would be fine and all, but it still requires a response from the server to tell the client that - which doesn't help if no one is getting through because of clogged pipes and dropped connections.
____________

OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13615
Credit: 30,389,506
RAC: 20,894
United States
Message 928406 - Posted: 24 Aug 2009, 14:13:42 UTC - in response to Message 928396.

By the way, I found that the Update button can be pressed more than once per second, but it appears that it does not actually do all of those operations.


It only ignores OCD-type presses because it actually takes time to send the information out over the internet and receive a response back, which happens far slower than somebody hammering the button. But when I say people "abuse" the button, I am not talking about them literally pressing the update button every couple of seconds. I am actually talking about those who are having problems getting work because of server comms problems who press the update button, get a failed request and then press the update button again; rinse, lather, repeat.

Also, the current Update button is much easier to abuse than my proposal.


Agreed, which is why I think it should be disabled for as long as the backoff time is randomly chosen. However, there are a few scenarios where this could actually harm troubleshooting a problem when there aren't server issues.
____________

Profile Steven Meyer
Avatar
Send message
Joined: 24 Mar 08
Posts: 2301
Credit: 2,999,007
RAC: 4
United States
Message 928549 - Posted: 25 Aug 2009, 3:14:03 UTC - in response to Message 928403.
Last modified: 25 Aug 2009, 3:21:37 UTC

respectfully, i do not agree with the abuse of the update button. if the update button can be abused, they would implement a longer minimum wait time between updates. as it is right now, you can updatre all day long if you just wait a few seconds. if you do it every 3 seconds, the server says you updated again too quickly.


...and how do you think the server sends a message to say that it's been too recent? It certainly isn't the DNS server giving this message.

In fact, the server has to respond to each and every request, even to say, "sorry, you already contacted me less than a minute ago. Please try again later". When the server is having comms problems for whatever reason, and everyone starts hitting this button, they are only adding to the load only to have the server tell them they've contacted too soon.

Sure, it can be argued there's no proof that too many people bother to do this, but I still don't believe it should be made easier for people to pound the servers.

on the other hand, i would like to see the server implement a longer minimum wait time between updates and network comms. a minimum of 1 hour. anything sooner and i would like the server to error back saying you updated in too quick of a sucession.


That would be fine and all, but it still requires a response from the server to tell the client that - which doesn't help if no one is getting through because of clogged pipes and dropped connections.

Um... I was assuming that the local client would be the place to put the timeout, because you certainly don't want to occupy the network, or the server, with this trivial task, or store the info in the database. In any case, the whole point is to avoid traffic to the server.
____________
FireFox Personas


OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13615
Credit: 30,389,506
RAC: 20,894
United States
Message 928562 - Posted: 25 Aug 2009, 4:06:49 UTC - in response to Message 928549.

Um... I was assuming that the local client would be the place to put the timeout, because you certainly don't want to occupy the network, or the server, with this trivial task, or store the info in the database. In any case, the whole point is to avoid traffic to the server.


The local client provides the random time out, but it has to be the server that responds and says, "You last talked to me too soon, please wait a while". Otherwise, if the serer does not respond, then BOINC says an entirely different message about the server not being able to be contacted, then it tries a reference site (google.com) and if successful, concludes that the servers must be down (not entirely true, just a presumption in the wording).
____________

1mp0£173
Volunteer tester
Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 928565 - Posted: 25 Aug 2009, 4:12:57 UTC - in response to Message 928562.

Um... I was assuming that the local client would be the place to put the timeout, because you certainly don't want to occupy the network, or the server, with this trivial task, or store the info in the database. In any case, the whole point is to avoid traffic to the server.


The local client provides the random time out, but it has to be the server that responds and says, "You last talked to me too soon, please wait a while". Otherwise, if the serer does not respond, then BOINC says an entirely different message about the server not being able to be contacted, then it tries a reference site (google.com) and if successful, concludes that the servers must be down (not entirely true, just a presumption in the wording).

This is part of every scheduler response:

8/24/2009 9:10:18 PM SETI@home Project requested delay of 11 seconds

Note that this does not mean that CryptokiD is correct. The update button can be abused, and hammering on it so that the BOINC client queries every 11 seconds is still abuse.

I'd also like to see the project send everyone away for a longer time after a successful scheduler request, but that might upset those with dialup or otherwise intermittent connections.
____________

Profile Steven Meyer
Avatar
Send message
Joined: 24 Mar 08
Posts: 2301
Credit: 2,999,007
RAC: 4
United States
Message 928574 - Posted: 25 Aug 2009, 7:22:42 UTC - in response to Message 928562.
Last modified: 25 Aug 2009, 7:23:19 UTC

The local client provides the random time out, but it has to be the server that responds and says, "You last talked to me too soon, please wait a while". Otherwise, if the serer does not respond, then BOINC says an entirely different message about the server not being able to be contacted, then it tries a reference site (google.com) and if successful, concludes that the servers must be down (not entirely true, just a presumption in the wording).


Again, since the local client knows when it was last used to contact the server, and knows if that attempt was successful, it has all the info that it needs in order to make the decision to put the request through, or to respond with "You have contacted the server 11 hundred and 49 times in the last 60 minutes. Please wait for a few minutes before contacting the server again."

Given that the server is contacted just to find out if I can contact the server again, it would make sense to off-load that task to the local client, right?

Thus it will not load the server at all even if millions of users are hammering on their "Update" buttons.

If BOINC does their code correctly, those who abuse their "Update" buttons will only be hurting themselves.

So BOINC would then be free to make it easier for those of us who are not abusive, by providing a System Tray Menu like this:

+------------------------------------+ | Open BOINC manager | +------------------------------------+ | Snooze | +------------------------------------+ | * Use GPU while computer is in use |
| Update all projects |
| Do network communications | +------------------------------------+ | About BOINC Manager | +------------------------------------+ | Exit | +------------------------------------+

____________
FireFox Personas


OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13615
Credit: 30,389,506
RAC: 20,894
United States
Message 928591 - Posted: 25 Aug 2009, 10:21:37 UTC - in response to Message 928574.

The message is given by the client only after a successful communication with the server. It would make sense to build it into the client to disable the "update" button if the last communication was too recent (I mentioned this earlier), but (as also mentioned earlier) there are scenarios where doing so would make troubleshooting connection issues very difficult.

I'm not certain that having the client automatically back off without contacting the server if comms happened too recently wouldn't make things more difficult for those people, so we are left with a sort of "honor" system that people will not abuse it, but many have admitted to doing this anyway despite information against doing so.
____________

Profile Steven Meyer
Avatar
Send message
Joined: 24 Mar 08
Posts: 2301
Credit: 2,999,007
RAC: 4
United States
Message 928601 - Posted: 25 Aug 2009, 12:31:28 UTC - in response to Message 928591.

The message is given by the client only after a successful communication with the server. It would make sense to build it into the client to disable the "update" button if the last communication was too recent (I mentioned this earlier), but (as also mentioned earlier) there are scenarios where doing so would make troubleshooting connection issues very difficult.

I'm not certain that having the client automatically back off without contacting the server if comms happened too recently wouldn't make things more difficult for those people, so we are left with a sort of "honor" system that people will not abuse it, but many have admitted to doing this anyway despite information against doing so.


Well then, if an honor system is what we are left with, then so be it.

But this should not prevent the suggestion from being implemented like this:

+------------------------------------+ | Open BOINC manager | +------------------------------------+ | Snooze | +------------------------------------+ | * Use GPU while computer is in use | +------------------------------------+ | About BOINC Manager | +------------------------------------+ | Exit | +------------------------------------+

Thus providing a 2-click way to turn on or off the CUDA code.
____________
FireFox Personas


OzzFan
Volunteer tester
Avatar
Send message
Joined: 9 Apr 02
Posts: 13615
Credit: 30,389,506
RAC: 20,894
United States
Message 928608 - Posted: 25 Aug 2009, 13:24:19 UTC - in response to Message 928601.

I've got no argument against the GPU option. I can certainly see how it could be very beneficial for many users.
____________

John McLeod VII
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 15 Jul 99
Posts: 24520
Credit: 521,562
RAC: 62
United States
Message 928669 - Posted: 26 Aug 2009, 2:08:35 UTC - in response to Message 928591.

The message is given by the client only after a successful communication with the server. It would make sense to build it into the client to disable the "update" button if the last communication was too recent (I mentioned this earlier), but (as also mentioned earlier) there are scenarios where doing so would make troubleshooting connection issues very difficult.

I'm not certain that having the client automatically back off without contacting the server if comms happened too recently wouldn't make things more difficult for those people, so we are left with a sort of "honor" system that people will not abuse it, but many have admitted to doing this anyway despite information against doing so.

Each message response from the server contains the minimum time to the next connection to the server. If the client would remember this in addition to the backoff time, the manager could implement a disabled update button.
____________


BOINC WIKI

Profile Lint trapProject donor
Send message
Joined: 30 May 03
Posts: 859
Credit: 26,660,950
RAC: 16,598
United States
Message 928708 - Posted: 26 Aug 2009, 4:29:08 UTC
Last modified: 26 Aug 2009, 4:34:17 UTC

Any figures available on abusive behaviour? Does it occur only (or mostly) after extended server downtime or just during connection issues? Maybe longer (edit) time intervals could be enforced only when really needed?

Does UPDATE really impact the servers in a big way? I mean, compared to having Tasks open all day, every day, so I can browse through them at any time, or scanning the forums? Refresh gets used a lot, too.

Is there a table of "costs" (server work required) for user initiated events?

Martin

Profile Steven Meyer
Avatar
Send message
Joined: 24 Mar 08
Posts: 2301
Credit: 2,999,007
RAC: 4
United States
Message 928783 - Posted: 26 Aug 2009, 12:53:26 UTC - in response to Message 928669.

Each message response from the server contains the minimum time to the next connection to the server. If the client would remember this in addition to the backoff time, the manager could implement a disabled update button.


Sounds ideal to me.

  • The server makes the decision on when the next connection will be allowed
  • The client enforces that decision by disabling the button/menu-item/what-have-you that makes the connection happen, without contacting the server again just to get the same decision back again.



Note that the decision on when the next connection will be allowed could be made in part based on the amount of network traffic and other load on the server.

More load on the server = longer timeout periods.

____________
FireFox Personas


1 · 2 · 3 · 4 . . . 5 · Next

Questions and Answers : Wish list : BOINC Wish List

Copyright © 2014 University of California