Computational infrastructure requirement for the development of genomic data analysis framework

Question

[shuksi1984]

What are the requirements (computational infrastructure) to develop an application for the analysis of whole genome, exome and transcriptome data? Example: Galaxy like application.Would like to know the requirement for both cloud and local server.

[Devon Ryan]

This is dependent entirely upon the exact application and the number and size of samples.

[colindaven]

Do you really mean software development ? Or do you want to analyze genomic results ? Most developers will only work on lightweight dev machines with small test datasets. Full whole genome datasets (the biggest) might take 24-48 hours on modern hardware, even with 56 cores, SSDs and 10Gbit networks.

[shuksi1984]

We want to develop an application to analyze genomic results. Hence, to start with, is working on the infrastructure. Lets say, we get 2 WG, 3 WE, and 3 RNAseq sample everyday.

[Devon Ryan]

Right, and doing what exactly to analyze them? How big are the files? What are your time constraints? Etc. Typically for that amount a day you're going to want a cluster backing things just to ensure that everything is done before the next sample rolls in (unless the WGS coverage is low).

[colindaven]

Simple criteria.

• 2 servers (more is better) + 1 VM head node.
• SLURM job scheduler
• 50-100TB HD space (NFS for allow addition of further servers)
• Server config -48-56 core, at least 256 RAM, 10 Gbit Network ports
• 10Gbit switches
• Backup to tape/ cloud

If you have more money go for more servers and or a big SSD scratch of >10 TB

Hope that helps.