For organizers, how many nodes and when?

Hello, organizers of summer 2014 competitions and events. Could you please post to this forum with how many nodes you think you may want, and the dates you think you want them? This will help us get started on planning how we will allocate the nodes.

Thanks,
Aaron

Hi Aaron,

For CoCo 2014 we would like to have 10 nodes from July 07 till July 14, if possible.

Thanks, Harald.

Hi Aaron,

for the QBF Gallery I would be great if we could use
15-20 nodes from May 20th to June 15th.

Thanks,
Martina

For CASC I'd like to have some, e.g., 10-20, nodes for testing from 10th July, then everything possible 19th-21st July for running the competition.

Hi Geoff,

same here for Termination,

We want to present the awards at the Vienna ceremony on July 21,
and run the competition as close as possible before that
(so it can be watched "live" during IJCAR),
and on as many nodes as we can get (at least 10, but 100 is better)

and we need to do some tests before (earlier in July).
I think at least one test run should be done on exactly set of nodes
that will be used in the actual competition.

- J.

For SMT-COMP, we are greedy - we would like as many as possible.
We will be running the competition from Jun 15 - July 17, with trial runs from May 15 - June 15. We'd like to run as many benchmarks as possible over the competition period.
So 32 nodes at least. 64 or more maybe if no one else is using them.
We are not going to try to run live during the conference, though the tail of the jobs might be running then.

(Sorry for the delayed response)

Hi Aaron,

For SyGuS-COMP it would be great if we can have
5-6 nodes from June 25 to July 2nd.

Thanks,
Dana

Hi, everyone.

Thanks for your replies. Here is the data in tabular form, ordered by start date:

event,how many,start,end
QBF Gallery,22,05/20/14,06/15/14
SMT-COMP,max possible,06/15/14,07/17/14
SyGuS-COMP,6 nodes,06/25/14,07/02/14
Confluence,10,07/07/14,07/14/14
Termination,max possible,07/19/14,07/21/14
CASC,max possible,07/19/14,07/21/14

QBF Gallery is a bit of a special case right now, because you wish to start right away (even a few days ago!). I have increased the number of nodes a little beyond what you asked for. Your request does not overlap with other events at all, so you can just have 22 nodes right now.

We have an elaborate queue reservation system in place, but testing it has revealed lots of problems. I probably will try to fix those and use the system for all the events besides QBF Gallery, because it will be more flexible in terms of adding nodes and removing nodes during a reservation.

For now, I have just created a permanent queue with community access only to QBF, for QBF Gallery.

We will have to figure out how to allocate the other resources for the later events. Any ideas are welcome.

Aaron

An update on queues for the competitions: at this point, I am probably not going to try to get the queue reservation system working. Instead, I will just create static queues for the different competitions. This is what I did for QBF Gallery, and it worked just fine.

One other note: I am thinking to give SyGuS-COMP and CoCo nodes from all2.q. These are the same as the nodes in all.q except they have "only" 128GB physical memory. I would like to use these nodes for SyGuS-COMP and CoCo because those competitions have requested fewer nodes than the remaining other three competitions, which are all seeking the max possible. Please let me know if you have concerns about this.

Aaron

Aaron,

From the table you published earlier (5/23), the contention during the period 6/15 - 7/17 is

SMT-COMP,max possible,06/15/14,07/17/14
SyGuS-COMP,6 nodes,06/25/14,07/02/14
Confluence,10,07/07/14,07/14/14

From this it appears that SyGuS and Confluence could use some of all.q2 nodes, with the rest left for other experiments, and, during this period, SMT-COMP could use all of all.q

From 7/19 on, CASC and Termination can split all.q

- David

David,

You are right. This sounds reasonable to me. I hope other organizers will speak up if they have concerns about that as a possible resource-allocation plan.

Aaron

Admin wrote:
You are right.  This sounds reasonable to me.  I hope other organizers will speak up if they have concerns about that as a possible resource-allocation plan.
Aaron

Hi,

For CoCo this plan is fine.

Harald.