This information is obsolete as of September 2017.¶
- Table of contents
- New GPGrid
- This information is obsolete as of September 2017.
- When can we use it?
Making this resource available to users is our top priority.
We are actively testing it and expect testing to complete by Oct 16' 2015
- What is the default allocation?
1 CPU/2GB Memory/35GB Disk
- What is the maximum allocation?
8 CPU/2GB Memory/35GB Disk
So, that would be 16 GB of memory with 280 GB of disk.
There are two very important caveats to this:
You must have all 8 CPUs on the worker node available in order to
get a slot that big (condor does some defragmentation, but it is not
fully efficient and still in develop) and so you will likely wait a
longer time to get a 16 GB slot than a 2 GB slot.
The condor accounting will charge you for utilizing 8 slots if you
request one 8 CPU/16 GB slot. 4 slots for a 4CPU/8 GB slot. MicroBooNE’s
and your personal allocation will reflect this utilization and end up
farther back in the queue the next time a job starts. Note that you can
request such a slot and run multiple processes in parallel but they must
collectively stay within the resource limitation.
- How does one determine, e.g., the number of available slots at any allocation level?
The slots are configured at the GlideInWMS factory and instantiated at run time.
So there is no fixed number of slots available for a fixed allocation level.
You can determine the number of single CPU/2 GB/280 MB slots that can be combined into
larger slots (the max slot being 8 CPU/16 GB/280 GB).
- Will our jobs run if we submit them requesting 4GB of memory?
- Will we be charged for two slots if we submit jobs requesting 4GB of memory?