What would be recommended hardware (computing) for a NGS lab?

[Torst]

Hi All, We are in process of expanding our high-throughput Next Generation Sequencing facility. The sequencer we have. But we are thinking of going for better computing hardware. As far as the budget is concerned we are not running on a 'shoe-string' budget but we will have to justify all the expenses. Any pointers to number of nodes, amount of RAM per node, amount of scratch space per node, total storage and any specific pointers are welcome. Any suggestions?

As an aside, don't forget that without good staff members to actually analyse the data, a big compute system won't be worth much to you.

[drlukun]

I also would like to set up our own analysis platform.
More detailed information about the whole hardware system provided by company is greatly appreciated.

[ECO]

Merged ymc's separate thread with this. Seems I got an answer to my own question.

[ECO]

Do people want to keep hardware discussions in this forum? Do we need a hardware/infrastructure subforum?

[ymc]

Building a computer. Please advise

Hi I am building a system to do SNP calling from raw data and SNP imputation and maybe GWAS also. This is the system I have in mind:

2 x Intel Xeon E5520
2 x Intel BXSTS100C 130W CPU Fan for Xeon 5500s
Intel S5500HCV Board
3 x Kingston 8GB DDR3-1333 Registered ECC
LSI Mega RAID SAS 9260-8i
OS/Apps Storage (RAID0, 2 stripes)
2 x OCZ 30GB SATA II 64MB Cache SSD OCZSSD2-1VTX30G
Main Storage (RAID10, 2 stripes, 2 mirrors)
4 x Hitachi 1TB HDS721010CLA332 SATA II/32MB HDD
Swap Drive (RAID0, 2 stripes)
2 x Hitachi 300GB HUS153030VLS300 15K SAS HDD
Enermax FMA II 535W EG565P-VE DXX 2.2 ATX
Chenbro CA-SR20964 4-Bays E-ATX Server Case
-----------------------------------------------
Total US$4,800

Do you think this system is an overkill for my purposes? If you were me, how would you fix it? Thanks in advance!

[ymc]

AFAIK the SAS is preferred because of the sheer size of data that needs to be copied across the compute nodes. I think 3 Gbps is just not enough speed. But as a disclaimer, i am not really an expert and probably experts should comment on this, i am making this comment from some experience.

What do you think about the Seagate 7200RPM SAS 6Gb/s hard drives? Are they better than their SATA 3Gb/s equivalent?

Or do you think everything should be 15000RPM SAS 6Gb/s in NGS? No place for slow HDD???

[sameet]

AFAIK the SAS is preferred because of the sheer size of data that needs to be copied across the compute nodes. I think 3 Gbps is just not enough speed. But as a disclaimer, i am not really an expert and probably experts should comment on this, i am making this comment from some experience.

[ymc]

Is SAS hard drive necessary for NGS???

Is SATA 3Gb/s fast enough for NGS???

[sameet]

We are looking at about 2 SOLiD runs per week that will generate to the order of 8 TB data per week, but after post-processing it should reduce to about 100 - 200 GB of usable data. Most of the raw (image) data need not be stored for long term, it may be stored for at the most a month after preliminary analysis.

The tools are basically open source tools like the velvet pipeline and other chip-sep, rna-seq tools that are standard for the SOLiD. The memory requirements are variable, but they require decently high amounts of memory.

[thinkRNA]

I think it will help to know how much data you for see generating on a weekly basis? How long will you need to store the data for your users? The amount of memory required by the tools you are using for analysis.

We are also thinking about setting up hardware and are doing preliminary analysis of our needs (which seems very variable at this point). So, I would like input from the community on this as well.

[sameet]

Hi All,

We are in process of expanding our high-throughput Next Generation Sequencing facility. The sequencer we have. But we are thinking of going for better computing hardware. As far as the budget is concerned we are not running on a 'shoe-string' budget but we will have to justify all the expenses.

Any pointers to number of nodes, amount of RAM per node, amount of scratch space per node, total storage and any specific pointers are welcome.

Any suggestions?