In article <Pine.SOL.3.96.1000225113218.4072V-100000 at cress>,
Keith Bradnam <keith at thale.nott.ac.uk> wrote:
>>Does anybody have any experience of performing intensive bioinformatics
>tasks (e.g. running blast searches, serving databases to the outside
>world) on PCs running linux as oppposed to Suns or DECs?
>>We are currently looking to upgrade our Sun servers and are curious
>whether switching to Linux on high-end PCs is a better (and maybe cheaper)
>option for doing bioinformatics.
There are a couple of performance issues:
1) SMP performance on Linux is not as good as something like Solaris.
2) Most PC motherboards can't take much RAM (usually only 768 MB or
1GB). This is simply not enough for many bioinformatics programs.
3) On Intel platforms, Linux is a 32-bit OS, which means that it can't
handle files greater than 2.1 GB in size. This could be a problem for
you. Linux on alpha (I believe) is 64-bit and does not have this
Despite these limitations, the price/performancs of Linux on Intel
still beats any commercial vendor.
If you can split your work into small enough chunks that the low RAM
amount of the PC architecture is not a problem, then the Intel solution
is a winner.