New Hardware Setup a) Server with Linux b) Server ESX and on top Linux or b) Workstation Z820 with Linux

Question

[it]

I don't even know here to post so please move if wrong Tag.

Good evening,

Disappointed to find out that all the open source software for NGS analysis does not Support TESLA GPU we go now for bare CPU power.

We run ESX Vmware 5.X Setup and 95% Windows Servers and some Centos. For two new Bioinformatics working place we would like to know following. Because of the noise and also heat the workstation will be placed in the Server room. Since we don't like standing things and not 19" the question would be if we can't use a Regular HP Server with Local Server.

Most of the stuff was done external until now at some large clusters so Amazon AWS is not an opinion because of confidential data that they handle now.

Software: BWA,Bowtie,Velvet,Mira,HMMER (Target seem only Opens source from users side)

OS: Will be Linux Red hat, Centos or Ubuntu

Setups:

a) HP Z820, 96 GB RAM, Two Xeon, Linux Direct on machine

b) HP DL 385 Intel, 96 GB RAM, Two Xeon, Linux direct on machine

c) Hypervisor ESX 5.5. HP DL385 Intel, 96GB RAM, Two Xeon, local Storage

We from IT-corporate side would prefer the ESX Vsphere soltuion because we can do snapshots, Veeam backup and other things. I have seen Hypervisor studies for Bowtie and the lost was 5% to 20% on heavy load in time that calculations need. That no problem from our side because we can then from the Windows Team manage things better.

If the BOX anyway goes into the Clima room because of Noise and heat would it be not better to take a regular Server?

We are Intel Reseller and no I don't want any NON HP/DELL brand because we need 4+ hrs repair time

Budget Price ranges is around EUR/USD 9500.- to 12'000.- per machine.

---

I come from large pharma where we had team with 30+ Bio IT Informatics so I don't expect too much wonders. The department was fully isolated from corporate network because people having almost full access on those machines which is a problem for IT-security mostly.

Thank you for any tips or direct answer if the Server solution or Server and ESX Hypervisor layer in between is absolute stupid idea. From management, backup and security it would be perfect for us.

Answer 1

[Devon Ryan]

You'll end up needing to chat with the end users and discuss which of (B) and (C) to go for. In the systems I've used, utilization wasn't 100% (or particularly close) so a loss of 5% or so on top wouldn't be a big deal if that means that disaster recovery and support from IT are better.

However, if the bioinfo folks have already spec'd out their minimum needs then you don't want to undershoot that by adding hypervisor overhead.

Edit: I should probably add that the number of bioinformaticians who will even notice that you're using a hypervisor is really low. They'll likely just be happy with the additional CPU cycles.

Answer 2

[william.bill.langdon]

BarraCUDA is compatible with BWA indexes but runs on nVidia CUDA GPUs. We just released version 0.7.107f http://sourceforge.net/projects/seqbarracuda/ which has also been used with Epigenetics data

Bill http://www.cs.ucl.ac.uk/staff/W.Langdon/

Answer 3

[Michael]

Your setup contains too little RAM for running assemblies most likely, try to get a setup with at least 128GB of RAM or more. Rather get a single machine with sufficient RAM than 3 with insufficient.