Server Specs for small Bioinformatics Server

Created on 2016-10-31 (Mon)

There seems to be very little information on how to build a small workhorse computer for running bioniformatics pipelines on the internets. Most of the information is focused on big expensive Next Gen Sequencing centers and there is none for small university labs or companies.

Here's my contribution to this lack of data.

Disclaimer - this post does not address the storage solution for bioinformatics labs. For large data sets this is a non trivial solution though for smaller data sets (under 50TB) getting an 8 bay NAS with 8x 6TB HDD (or even 8TB HDD) might be a decent start.

Now regarding the work horse computer build.

The architecture I chose is based around high-end gaming Intel architecture - as this is what I find I have moderate experience with.

I propose two solutions that differ only in the CPU model used. If you find you can parallelize a lot of the pipeline the more expensive architecture is going to be probably faster at a pretty high cost.

Also all component prices are in $CAD and are based on NCIX.com list price.

Without further ado:
Component Model Price Notes
Motherboard EVGA X99 Classified E-ATX X99 $529.99 8x DDR4 slots (128GB); M.2 connectors; 10x SATA interfaces; socket supports both Xenon and i7 CPU.
CPU (cheap) Intel Core I7-6800K BROADWELL-E $589.99 Marginally more expensive than the i7 4 core Skylake. But has 2x L3 cache and 6 cores.
CPU (expensive) Intel CPU BX80660E52680V4 Xeon E5-2680V4 $2,523.40 14 core (28 threads) with 3.5MB L2 and 35MB L3 (about 4x i7 CPU at about 5x price)
RAM 128GiB 4x Corsair Vengeance Lpx 32GB 2X16GB DDR4 3000MHZ $999.92 Maxed out the MB - the more the better.
SSD Samsung 950 Pro 512GB M.2 2280 NVMe $429.99 Place holder until the Samsung 960Pro is released later in the year.
HDD 2x Western Digital WD60EFRX 6TB Red $639.98 In RAID 0 for increased data transfer.
Power Supply EVGA SUPERNOVA 750 G2 80 PLUS $134.99 750W is OKish for this MB (more would be better)
Case Phanteks Enthoo Pro Full Tower EATX $129.99 nothing fancy

Total price (cheap): $3,454.85 (before tax) [~$3,900 after tax]

Total price (expensive): $5,388.26 (before tax) [~$6,100 after tax]

Note: You'll also need to buy a video card - I will use one that I have lying around so I didn't include it in the price. As I don't think the bioinformatics tools are GPU optimized the graphics card can be as cheap as possible. Also there might be some cables you need to buy.

A few other considerations:
Instead of using a single powerful computer you can build a computing cluster. The major issue with this is that someone needs to administer the cluster. The personnel is a recurring cost that completely dominates the cost of the cluster itself in the long run. Alternatively you could look into DNA Nexus for a cloud alternative for your computing needs.