[Bioclusters] Any issues porting applications to OS X?

Joe Landman bioclusters@bioinformatics.org
Fri, 05 Mar 2004 11:27:19 -0500

On Fri, 2004-03-05 at 10:32, Christopher Porter wrote:
> We're in the market for a cluster; most of our options are Xeon/Linux, 
> but one is a cluster of XServe G5s running OS X. We're going to run 
> some benchmarks to see how the performance compares, but some in of our 
> group have expressed concern that 'the vast majority bioinformatics 
> software is developed on Linux', and 'there may be a long time lag 
> before new software is available on OS X'.
> I have never had problems getting software I need to run on OS X, but I 
> wondered if anyone can provide me with examples of applications that 
> won't run on OS X, or are Linux only (only binaries released & no 
> source available).

Accelrys tools, many commercial chemistry applications (Gaussian,
Jaguar, Macromodel, ...), Tripos tools, Lion's platform, GeneData's
tools, ... (long list).

The Apples are good to consider in a specific range of applications,
where commercial binary code is not a requirement.  This is a
generalization, and there are specific counter-examples, but not
industry vertical counter-examples.  Where it is a requirement, you
likely would have better luck with Xeon and Opteron platforms.

> This is only one of the criteria we're judging on, and the performance 
> comparison will be interesting. Any insights on this issue would be 
> extremely useful, though.

Remember to look at the complete set of codes you will be running, in
the manner you will be running them.  Using those as criteria, you will
be able to distinguish between viable solutions and solutions while
possibly nice, do not completely cover the application set.  From that
point onwards, benchmarking is important (e.g. if the platform doesn't
run code X that you need, then benchmarks on code Y which you do not
need/use are not relevant to code X performance).

As a note: beware of hype.  Lots of folks talk about great performance,
but nothing speaks as loud as a real measurement of a real application
compiled/built the way you would run it, executing a real job.  


> Thanks,
> Chris
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters