Hardware requirements¶

MetAMOS was designed to work on any standard 64bit Linux or OSX environment. To use MetAMOS for tutorial/teaching purposes, a minimum of 16 GB RAM is recommended. To get started on real data sets a minimum of 64 GB of RAM is recommended, and up to 1 TB of RAM may be necessary for larger datasets. In our experience, for most 50-100 million read datasets, 64-128 GB is a good place to start.

Scenario #1: running locally on large memory server¶

Suggested all-purpose build:

256 GB RAM (16 x 16GB)
48 cores (96 HT, 4x cpu)
1 TB SSD temporary scratch space for running analyses
16 TB HDD archival space for storing analyses

Scenario #2: running on local cluster/grid¶

Notes:

Great for BLAST intense jobs
RAM will limit the supported assemblers
Grid job submission via SGE, others not supported
MPI install required for Ray Meta

Scenario #3: running on cloud via Amazon EC2 High Memory¶

Recommend for best price/performance ratio -> cr1.8xlarge (Memory optimized):

2 X 120 GB SSD
32 HT x 2.8GhZ
244 GB RAM
Spot instance currently at $0.361 per Hour (based on availability)
On demand instance at $3.500 per Hour (always available)
Reserved instance at $1.54 per Hour, $2474 upfront, approx. $2000 a month

More info on spot instances: http://aws.amazon.com/ec2/purchasing-options/spot-instances/

Or for smaller assembly jobs -> c3.8xlarge (Compute optimized):

2 X 320 GB SSD
32 HT x 2.8GhZ
60 GB RAM
On demand instance at $2.400 per Hour (spot instance same price!)

Here is a very useful cost calculator: https://www.scalyr.com/cloud/

Don’t forget to account for time/cost to upload data!