There are plenty more projects on WCG, especially now that scc is back.

Mike

Actually, something is completely nuts with MIP. A volunteer noted observation that one MIP job came with a 97 MEGABYTE file and I thought it may have been some database file used by multiple MIP tasks. Well, since I had suspended a task and forgot about it, this morning my queue was empty but for HST and ARP which I only allow 1 at a time. After unsuspending/resume 6(six) MIP came down. Each with a variable input file between 30 and 71 MEGABYTE. That much for just 2 hours or less of crunching. NUTS. Removed from the job selection.

I've had also some exception errors in the past. That is no more since I lowered the max mem usage by boinc.

Once i had a 250 mb upload for mip1. It was insane but hopefully neccesary.
and that while i have data-usage issues and noticed about downloading the following:

First boinc-client downloads the project files, executables and so on.
next it will download some task containing 2 numbers mip1_x_y
next it downloads the first number after mip1 x some files for more y tasks. (they used to be about 30 mb max but over time they have gotten bigger now like 90 mb )

And it downloads some small files for the second number y.

When downloading multiple tasks to fill task que, needing for example 3 tasks:
mip1_1_6
mip1_2_3
mip1_3_4
creates more traffic than downloading:
mip1_2_1
mip1_2_2
mip1_2_4
because the number x is more the same and only y varies on the client

I know more y numbers can be downloaded to a client, the tasks and files numbers are there because i have noticed them going to different computers at almost the same time. but instead it downloads some other x number creating another 90 mb traffic.

I wonder if data traffic can be limited somehow some if the server prefers to not changes the x number for a given client but only change the y number as much as possible when a subsequent task is requested by the same client

Server side there is (i think) no optimization in deciding wich wu to send based on the files that are allready downloaded(or download in progress) on the client.

i name mip1_x_y_z
the big project files are related to the number in X.

For example completing 4 wu(x 1 to 2, y 1 to 2), on 2 different computers.
forcing the distribution like:
computer1:
mip1_1_1
mip1_1_2
computer2:
mip1_2_1
mip1_2_2

is better than (randomly?)distributing:
computer1:
mip1_1_1
mip1_2_1
computer2:
mip1_1_2
mip1_2_2

in the first case the big files related to x are only downloaded once and then used again on the same client.
The second case the big files are downloaded not once but because of the way of distribution twice(once per client, 2 total).