Questions about hardware specifications for analysis

[sallusti10]

Hi,
i am new about this tecnology.
I woul know what must be the hardware caractersitic to use a Illumina Solexa.
I mean, can i use a personal computer, or i have need a cluster?
I think is correct to use a cluster.
What must be the OS? And ram and the space for storage?
Where can i find this kind of information?
Thank you,
and sorry for my banal question.
Bye

[drio]

Hi,
i am new about this tecnology.
I woul know what must be the hardware caractersitic to use a Illumina Solexa.
I mean, can i use a personal computer, or i have need a cluster?
I think is correct to use a cluster.
What must be the OS? And ram and the space for storage?
Where can i find this kind of information?
Thank you,
and sorry for my banal question.
Bye

Assuming you are getting a single instrument:

Start by looking at the specs of the minicluster that illumina sells with their sequencers (illumina compute). That's what you need to compute the data.

As for archiving, you want a MSL tape libraries or something similar.

[simonandrews]

The answer really depends on what you want to do with your data. An illumina machine comes with a control PC which now does all of the processing necessary to generate basecalled data. If you want to map this data to a reference genome then you'll need a separate system for that. You can do this on a PC as long as you have a lot of storage and RAM in it, but it won't be very quick (up to several hours per lane depending on your choice of mapping program). You can speed this process up by using a small cluster to parallelise the process.

For downstream analysis your computing requirements will change depending on what you're doing and how big your experiment is. For remapping experiments such as ChIP-Seq you can often just use a normal PC using software designed for that environment, or web based analysis systems. For assembly or mutation detection you're probably going to need something more beefy to do the computation.

In any case you'll also need to work out how to store your data. Even if you're not keeping Images from a run you can still end up with hundreds of gigabytes of data in a very short time. Our experience has been that storing, managing and backing up this data ends up being far more expensive than purchasing the computational hardware for its analysis.

[basickler]

I'm going to second Andrews here. With RTA you'll get basecalls off the machine directly so the days of needing a cluster to get to sequence off a GA are long gone. The problem comes from the fact that you now have GB of sequence and you generally want to do something with it. Alignment alone on MAQ or ELAND will take quite a while (few days) on a beefy PC ( 8 cores, 32GB ram) if you're working against a large database.

That's not to mention de-novo sequencing, re-assembly from reference, etc etc. You also have to figure in how long you're going to want to keep your data and plan storage for that accordingly. In order to answer your question I'd need a lot more info about what you plan on doing.

Brad

[drio]

I am seeing this question being asked many times, perhaps we should create a sticky thread or something like that?

[ECO]

I am seeing this question being asked many times, perhaps we should create a sticky thread or something like that?

I think a newegg.com wishlist would be more appropriate.

[sallusti10]

Thank you very much for your answers.
Now i am following a course of NGS in Breda and i hope to know more about it.
thank you

[basickler]

Currently we use clusters of dell blades with the following config
Dual Socket 4 core Nehelem
48GB Ram
Running CentOS + SGE
Connected to isilon storage as the backend.

This is pretty much what you'll get from Illumina's high end tier computing product as well.

Brad