Page Worklog_Hauswedell

Project Work log Hauswedell

Friday, 25.06.2010:

• second "real" meeting with Thieme to do planning and division of work
• got access to svn, built benchmark tools, razers2
• together with thieme setup directories, reorganised data

Saturday 26.06.2010:

• tried to do RazerS2 runs but failed due to memory being insufficient
• wrote mail to Holtgrewe to discuss whether we can get access to higher performance computers

Sunday, 27.06.2010:

• received response from Holtgrewe, but didn't agree with strategy of reducing data, as I was unsure of how to sample in a way that preserves the expected behaviour (distribution of reads producing deviating results not known)

Monday, 28.06.2010:

• met with Holtgrewe and Thieme, discussed various issues
• was convinced by Holtgrewe that sampling is indeed a good idea and that there is a seqan tool for uniformly sampling from a given file
• reorganised file and directory structure to accomodate for more different sample sets
• resolved issues with permissions/umasks
• built seq_sample and created 0.1%, 1% and 10% sets from the original input data
• started razers2-run on 0.1% of set I

Tuesday, 29.06.2010 - Thursday 01.07.2010:

• started and finished razers2-runs on 0.1% and 1% of I, and on 0.1% of II
• razers2-run on 1% of II still running

Thursday 01.07.2010:

• done bwa run on 0.1% of I
• got benchmark running on that bwa-run
• Result: {"total_intervals": 0, "found_intervals": 0, "superflous_intervals": 0, "additional_intervals": 0}
• → something went wrong, time to debug

Friday 02.07.2010 - Sunday 04.07.2010:

• spent weekend in correspondence with Holtgrewe to debug this issue
• it turned out that neither the benchmark nor bwa verify their input
• → the output generated by bwa aln is not SAM and not BAM, it is another kind of index
• after bwa aln it is neccessary to run bwa samse that creates SAM
• there now are some bwa results

Monday 05.07.2010:

• razers2-run on 1% of II still running
• since run-time is so much longer than with I, more investigation
• run on I was started without paramchooser → data is useless and also time information
• met with Thieme, Holtgrewe and Weese to discuss current issues
• Weese and Holtgrewe agree that a run-time of now over 130hours is not usual for mapping 30MiB vs 3GiB
• checked whether razers2 binary was built with optimizations and without testing/debugging stuff
• binary seems to be ok, problem has to be somewhere else

Tuesday 06.07.2010:

• started razers2-run on 0.1% of I again to see if it takes equally long with activated paramchooser
• takes 4.5 hours, which is better than for II, but still longer than it should
• built razers1 and started a run to check for regression in RazerS
• successfully did bwa-runs for 0.1% and 1% of both I and II
• tried to do a benchmark of bwa 0.1%/I against RazerS 01%/I
• benchmark segfaults

Comments