[Bioclusters] BLAT users?

[David Adelson]

It was not OSX specific and would probably only have affected you if 
you were using the blast8 output option.  It looks like the problems 
were solved as determined from a sample job I am still running on one 
node.  I need to check the cluster version next, but do not anticipate 
problems if the single node job completes without problems.

Dave

On Nov 2, 2004, at 3:23 AM, Ran Rubinstein wrote:

> I was asking because I am using BLAT with bigger databases (but 
> smaller query load) on IRIX, with great success - but this could still 
> affect me when the system goes into production. Is it an OSX specific 
> issue? Were your problems solved?
>
> Thanks,
>
> David Adelson wrote:
>
>> Thanks.  Stops working means no output generated for a considerable 
>> time and 0%cpu usage for the process for some time.
>>
>> Jim Kent worked on this over the weekend and actually found 3 memory 
>> leaks that he fixed.  The new source and binaries are available for 
>> download from his website/ftpsite as v31.
>>
>> Dave Adelson
>>
>> On Nov 1, 2004, at 10:18 AM, Ran Rubinstein wrote:
>>
>>> Hi David
>>>
>>> Please elaborate on what does 'stop working' mean. The leaks program 
>>> can find a lot more memory leaks and the program can still be 
>>> correct - try looking at the process memory usage, and see if it 
>>> grows over time, and whether it reaches the size of the physical 
>>> RAM.
>>>
>>> Regards,
>>>
>>> David Adelson wrote:
>>>
>>>> This may be the wrong forum to look for an answer, but I thought I 
>>>> would start here.
>>>>
>>>> We have been using BLAT on G4 Xserves (2GB RAM) for some time now.  
>>>> Given the memory limitations, we carve up the database into chunks 
>>>> so that we don't get a malloc error message telling us we have run 
>>>> out of huge memory.  We then parallelize the job and run the chunks 
>>>> in parallel on our cluster.  This works for modest query sizes 
>>>> (15,000 ESTs vs 1.2GB database-hs_chroms1-6).  However, when we 
>>>> scaled up to larger queries (493365 ESTs vs 1.3GB database-bovine 
>>>> genome 200,000scaffolds) we see blat just sort of stop working 
>>>> after a while.  I ran the leaks program and found 91832 memory 
>>>> leaks, so the problem seems to be memory related.  I would be 
>>>> curious to know if anyone else has seen this type of behavior with 
>>>> BLAT.  I am in the process of contacting Jim Kent about this, but 
>>>> thought it would not hurt to ask some other knowledgeable 
>>>> computational biologists about this.
>>>>
>>>> thanks in advance,
>>>>
>>>> Dave Adelson
>>>> Texas A&M University
>>>>
>>>> _______________________________________________
>>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>>
>>>>
>>>
>>>
>>> -- 
>>> Ran Rubinstein
>>> Dept. of Molecular Biology
>>> Faculty of Medicine, Hebrew University, Ein Karem, Jerusalem, Israel
>>> Tel +972-2-6757906 Fax +972-2-6758992
>>>
>>> _______________________________________________
>>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>>
>>
>> _______________________________________________
>> Bioclusters maillist  -  Bioclusters@bioinformatics.org
>> https://bioinformatics.org/mailman/listinfo/bioclusters
>>
>>
>
>
> -- 
> Ran Rubinstein
> Dept. of Molecular Biology
> Faculty of Medicine, Hebrew University, Ein Karem, Jerusalem, Israel
> Tel +972-2-6757906 Fax +972-2-6758992
>
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters
>