Allocation
How to select the resources to use in MiGA @ XSEDE
Once you have selected the appropriate application, it's a good idea to make sure the allocation settings are optimal for your task. We provide default settings that work well in most cases for a given application, but it's always better to double-check.

Total core count

The first step is to determine the number of cores you will need. The maximum core count is 4,096. The maximum number of cores MiGA will use is the number of genomes. If you have fewer than 4,096 genomes, set this value to the number of genomes you have.
Use the smaller value between the number of genomes and 4,096

Select a Queue

For a total core count < 128: use shared (default) For a total core count ≥ 128: use compute

Node Count

In the shared queue

Use 1
In the compute queue
You need to estimate the minimum number of nodes that can accomodate your job. For that, simply take the total core count above, divide it by 128 (maximum cores por node), and round it up. For example, if you have 300 cores: 300/128 = 2.34, so you should request 3 nodes.
Divide the total core count by 128 and round up

Wall Time Limit

This is the maximum number of minutes that the task will be running. If your task exceeds this number, it will be terminated without completing. However, if you pick a number too high, the task will stay a very long time in queue, so it's important to get this approximately right (but slightly overestimating).
Estimate the limit based on the type of application used

For Database Indexing and Genome Dereplication

The total time of these tasks grows quadratically with the number of genomes. A rule-of-thumb estimate using Fast indexing is:
time=5n+n2/10corestime = \frac{5n + n^2/10}{cores}
Where time is the total wall time (in minutes), n is the number of genomes, and cores is the total core count. We recommend you never set a total wall time below 60 minutes.

For Genome Classification and Genome Typing

The total time of these tasks grows linearly with the number of genomes. A quick estimate is:
time=10ncorestime = \frac{10n}{cores}
Where time is the total wall time (in minutes), n is the number of genomes, and cores is the total core count. We recommend you never set a total wall time below 60 minutes.
Export as PDF
Copy link
On this page
Total core count
Select a Queue
Node Count
Wall Time Limit
For Database Indexing and Genome Dereplication
For Genome Classification and Genome Typing