I have read some of the threads here on BioStars concerning clusters versus BASS approaches, etc. I am looking at drawing up a proposal for the computer requirements for a local hospital that is moving in to NGS, specifically doing custom amplicon or panel sequencing in an oncology setting. We will be looking at a benchtop sequencer, probably MiSeq or Ion Torrent, although the new NextSeq 500 may not be totally out of the picture either.
While a workstation will be sufficient for doing mapping, variant calling, and annotations for MiSeq runs, we are thinking about future proofing and scalability. So we have been thinking about starting with a single compute node and single storage node type set-up, which can organically grow into a larger cluster as the amount of sequencing expands and we need more resources. Right now looking at storage the enterprise solutions are very expensive (Ilsilon, which I know many clinical sequencing centres use, starts at 30-40K for 30TB of NAS). Does anyone have any thoughts on building out our own NAS solution? Right now we have a total budget of ~20K for the starting compute solution.
Thoughts, opinions, etc very much welcome!
Update: We have a consultants report which of course states what they currently use, but they are comparable to a large sequencing centre. Clinical analysts have workstations with 16GB of RAM and Quad Core machines as well as having a cluster with compute nodes that have 256GB of RAM, etc. While having something that big is certainly future proof, it would be out of budget for a pilot project and later clinical implementation using initially only a MiSeq for generating sequence data.
Update2: I'll add that this is physically located in Canada, so we have different legal and privacy issues to deal with. Taking into consideration HIPPA requirements is still useful, as our patient privacy rules are similar, but my particular province also has stricter policies regarding data/samples leaving the country. So cloud providers aren't feasible.
We have been in contact with some providers for NAS storage and that is looking promising. Unfortunately the hospital IT department, as far as I know, is an all Windows shop. So we will likely be on our own for managing and supporting our own compute infrastructure. Only final reports need to interface with the hospital system. Currently we are looking at a ZFS-based system which will give us high-availability storage at a much cheaper price than most other enterprise solutions we have seen while still maintaining good redundancy and data backup.