using Galaxy for map/reduce
Hi all, I've been investigating use of Galaxy for our lab and it has many attractive aspects -- a big thank you to all involved. We still have a couple of related sticking points, however, that I would like to get the Galaxy developers' feedback on. Basically, I want to use Galaxy to run Map/Reduce type analysis on many initial data files. What I mean is that I want to take many initial datasets (e.g. 250 or more), perhaps already stored in a library, and then apply a workflow to each and every one of them (the Map step). Then, on the many result datasets (one from each of the initial datasets), I want to run a Reduce step which creates a single dataset. I have achieved this in an imperfect and not-quite-working way with a few tricks, but I hope that with a little work, Galaxy could be much better for this type of use case. I have a couple of specific problems and a proposal for a general solution: 1) My first specific problem is that loading many datasets (e.g. 250) into history causes the javascript running locally withing a browser to be extremely slow. 2) My second specific problem is that applying a workflow with N steps to many datasets creates even more datasets (Nx250 additional datasets). In addition to the slow Javascript problem, there seems to be other issues I haven't diagnosed further, but the console in which I'm running run.sh indicates many errors of the type "Exception AssertionError: AssertionError('State <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990> is not present in this identity map',) in <bound method MutableAttrInstanceState._cleanup of <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990>> ignored". Furthermore the webserver gets slow and my nginx frontend proxy gives 504 gateway time-outs. 3) There's no good way to do reduce within Galaxy. Currently I work around this by having a tool type which takes as an input a dataset and then uploads this to a self-written webserver, which then collects such uploads, performs the reduce, and offers a download link for the user to collect the reduced dataset. The user must manually then upload this dataset back into Galaxy for further processing. My proposal for a general solution, and what I'd be interested in feedback on, is an idea of a "dataset container" (this is just a working name). It would look and act much like a dataset in the history, but would in fact be a logical construct that merely bundles together a homogeneous bunch of datasets. When a tool (or a workflow) is applied to a dataset container, Galaxy would automatically create a new container in which each dataset in this new container is the result of running the tool. (Workflows with N steps would thus generate N new containers.) The thing I like about this idea is that it preserves the ability to use tools and workflows on both individual datasets and, with some additional logic, on these new containers. In particular, I don't think the tools and workflows themselves would have to be modified. This would seemingly mitigate the slow Javascript issue by only showing a few items in the history window (even though Galaxy may have launched many jobs in the background). Furthermore, a new Reduce tool type could then act to take a dataset container as input and output a single dataset. A library doesn't seem a good candidate for the dataset container idea I have above. I realize that a library also bundles together datasets, but it has other attributes that don't play well with the above idea (the idea of hierarchically arranged folders and heterogeneous datasets) nor can it be represented in the history. I'm interested in thoughts on this proposal, as I think it would really help us, and I think our use case may be representative of what others might also like to do. I realize that in my text above I write "with some additional logic" to describe the work required to implement this idea, but the fact is that I have very little idea about how much work this would be. So, practically speaking, my question boils down to how hard would implementing this be, given the existing code base and goals? And, would such an implementation - if done to the taste of the Galaxy devs, of course - have a chance of making into the Galaxy distribution? Thanks, Andrew -- Andrew D. Straw, Ph.D. Research Institute of Molecular Pathology (IMP) Vienna, Austria http://strawlab.org/
On Tue, Aug 2, 2011 at 3:12 PM, Andrew Straw <andrew.straw@imp.ac.at> wrote:
...
My proposal for a general solution, and what I'd be interested in feedback on, is an idea of a "dataset container" (this is just a working name). It would look and act much like a dataset in the history, but would in fact be a logical construct that merely bundles together a homogeneous bunch of datasets. When a tool (or a workflow) is applied to a dataset container, Galaxy would automatically create a new container in which each dataset in this new container is the result of running the tool. (Workflows with N steps would thus generate N new containers.) The thing I like about this idea is that it preserves the ability to use tools and workflows on both individual datasets and, with some additional logic, on these new containers. In particular, I don't think the tools and workflows themselves would have to be modified. This would seemingly mitigate the slow Javascript issue by only showing a few items in the history window (even though Galaxy may have launched many jobs in the background). Furthermore, a new Reduce tool type could then act to take a dataset container as input and output a single dataset.
...
That is a very interesting idea. Note that in some of the usecases I had in mind the order of the sub-files was important, but in other cases not. So I think that internally I think Galaxy would have to store a "dataset collection" aka "homogeneous filetype collection" as an ordered list of filenames. As you observed, at the level of an individual tool, nothing changes - it gets given a single input file(s) as before, but now multiple copies of the tool will be running, each with a different input file (or files for more complex tools). I had been mulling over what is essentially a special case of this - a new datatype for "collection of BLAST XML files", and debating with myself if a zip file or simple concatenation would work here. In the case of BLAST XML files, there is precedent from early NCBI BLAST tools outputting concatenated XML files (which are not valid XML). My motivating example was the embarrassingly parallel task of multi-query BLAST searches. Here we can split up the input query file (*) and run the searches separately (the map step). The potentially hard part is merging the output (the reduce). Tabular output and plain text can basically be concatenated (note we should preserve the original query order). For XML (or -shudder- HTML output), a bit of data munging is needed. Your idea is much more elegant, and to me fits nicely with a general sub-task parallelization framework (as well as your example of running a single workflow on a collection of data files). Peter (*) You can also split the BLAST database/subject file, and there are options to adjust the e-value significance accordingly (so it is calculated using the full database size, not the partial database size). The downside is the merging of the results is much more complicated.
Hi I really like this proposal. We faced some of the similar issues you talk about below when we tried to use galaxy to use High Throughput computing techniques (using Condor) for sequencing close to 500 genomes (embarrassingly parallel problem). We leveraged (hacked) the dataset construct but it did not map very well into the problem we were trying to solve. We ended up taking an approach that involved using the Galaxy tool mechanism to create a "Composite Dataset" from a filesystem location. This approach required a configuration file to be updated with a directory path containing the datasets. The directory listing was then filtered and displayed to the user of the tool to allow them to select a Genome. The tool would then create a composite dataset consisting of a JSON document containing at least a list of all the files in the CG data directory. We are not sure how generally useful this tool would be. On Aug 2, 2011, at 10:12 AM, Andrew Straw wrote:
Hi all,
I've been investigating use of Galaxy for our lab and it has many attractive aspects -- a big thank you to all involved.
We still have a couple of related sticking points, however, that I would like to get the Galaxy developers' feedback on. Basically, I want to use Galaxy to run Map/Reduce type analysis on many initial data files. What I mean is that I want to take many initial datasets (e.g. 250 or more), perhaps already stored in a library, and then apply a workflow to each and every one of them (the Map step). Then, on the many result datasets (one from each of the initial datasets), I want to run a Reduce step which creates a single dataset. I have achieved this in an imperfect and not-quite-working way with a few tricks, but I hope that with a little work, Galaxy could be much better for this type of use case.
I have a couple of specific problems and a proposal for a general solution:
1) My first specific problem is that loading many datasets (e.g. 250) into history causes the javascript running locally withing a browser to be extremely slow.
2) My second specific problem is that applying a workflow with N steps to many datasets creates even more datasets (Nx250 additional datasets). In addition to the slow Javascript problem, there seems to be other issues I haven't diagnosed further, but the console in which I'm running run.sh indicates many errors of the type "Exception AssertionError: AssertionError('State <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990> is not present in this identity map',) in <bound method MutableAttrInstanceState._cleanup of <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990>> ignored". Furthermore the webserver gets slow and my nginx frontend proxy gives 504 gateway time-outs.
3) There's no good way to do reduce within Galaxy. Currently I work around this by having a tool type which takes as an input a dataset and then uploads this to a self-written webserver, which then collects such uploads, performs the reduce, and offers a download link for the user to collect the reduced dataset. The user must manually then upload this dataset back into Galaxy for further processing.
My proposal for a general solution, and what I'd be interested in feedback on, is an idea of a "dataset container" (this is just a working name). It would look and act much like a dataset in the history, but would in fact be a logical construct that merely bundles together a homogeneous bunch of datasets. When a tool (or a workflow) is applied to a dataset container, Galaxy would automatically create a new container in which each dataset in this new container is the result of running the tool. (Workflows with N steps would thus generate N new containers.) The thing I like about this idea is that it preserves the ability to use tools and workflows on both individual datasets and, with some additional logic, on these new containers. In particular, I don't think the tools and workflows themselves would have to be modified. This would seemingly mitigate the slow Javascript issue by only showing a few items in the history window (even though Galaxy may have launched many jobs in the background). Furthermore, a new Reduce tool type could then act to take a dataset container as input and output a single dataset.
A library doesn't seem a good candidate for the dataset container idea I have above. I realize that a library also bundles together datasets, but it has other attributes that don't play well with the above idea (the idea of hierarchically arranged folders and heterogeneous datasets) nor can it be represented in the history.
I'm interested in thoughts on this proposal, as I think it would really help us, and I think our use case may be representative of what others might also like to do. I realize that in my text above I write "with some additional logic" to describe the work required to implement this idea, but the fact is that I have very little idea about how much work this would be. So, practically speaking, my question boils down to how hard would implementing this be, given the existing code base and goals? And, would such an implementation - if done to the taste of the Galaxy devs, of course - have a chance of making into the Galaxy distribution?
Thanks, Andrew
-- Andrew D. Straw, Ph.D. Research Institute of Molecular Pathology (IMP) Vienna, Austria http://strawlab.org/
___________________________________________________________ Please keep all replies on the list by using "reply all" in your mail client. To manage your subscriptions to this and other Galaxy lists, please use the interface at:
-- Ravi K Madduri The Globus Alliance | Argonne National Laboratory | University of Chicago http://www.mcs.anl.gov/~madduri
On Aug 2, 2011, at 10:12 AM, Andrew Straw wrote:
1) My first specific problem is that loading many datasets (e.g. 250) into history causes the javascript running locally withing a browser to be extremely slow.
What browser are you using?
2) My second specific problem is that applying a workflow with N steps to many datasets creates even more datasets (Nx250 additional datasets). In addition to the slow Javascript problem, there seems to be other issues I haven't diagnosed further, but the console in which I'm running run.sh indicates many errors of the type "Exception AssertionError: AssertionError('State <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990> is not present in this identity map',) in <bound method MutableAttrInstanceState._cleanup of <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990>> ignored". Furthermore the webserver gets slow and my nginx frontend proxy gives 504 gateway time-outs.
Yes, creating all the jobs and datasets for a workflow is relatively slow right now. We have some optimizations for this that are not in the mainline (not well tested) however there is a limit to how fast it can be with so many new datasets and objects being created. The better solution is probably to move workflow creation into a background process. Starting the workflow would just save the initial state, and a background process would actually create all the datasets and jobs and get it running. The downside is that the history would not be completely populated by the time the page had returned. -- jt James Taylor, Assistant Professor, Biology / Computer Science, Emory University
On Aug 2, 2011, at 10:12 AM, Andrew Straw wrote:
1) My first specific problem is that loading many datasets (e.g. 250) into history causes the javascript running locally withing a browser to be extremely slow. What browser are you using? Primarily Firefox 3.6.18 as packaged by Ubuntu 10.04 amd64. I know more recent browsers have faster JS interpreters, but I'm hoping that JS interpreter speed optimizations will be largely irrelevant with the
On 08/02/2011 06:43 PM, James Taylor wrote: proposal for the dataset container proposal I made. (I just tested, and it's certainly true that Google Chromium 12.0.742.112 on the same system is much faster.)
2) My second specific problem is that applying a workflow with N steps to many datasets creates even more datasets (Nx250 additional datasets). In addition to the slow Javascript problem, there seems to be other issues I haven't diagnosed further, but the console in which I'm running run.sh indicates many errors of the type "Exception AssertionError: AssertionError('State <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990> is not present in this identity map',) in <bound method MutableAttrInstanceState._cleanup of <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990>> ignored". Furthermore the webserver gets slow and my nginx frontend proxy gives 504 gateway time-outs. Yes, creating all the jobs and datasets for a workflow is relatively slow right now. We have some optimizations for this that are not in the mainline (not well tested) however there is a limit to how fast it can be with so many new datasets and objects being created.
The better solution is probably to move workflow creation into a background process. Starting the workflow would just save the initial state, and a background process would actually create all the datasets and jobs and get it running. The downside is that the history would not be completely populated by the time the page had returned.
There could be a little checkbox in the run workflow page which allows the user to decide whether to fork the workflow creation process. Then the user could do so for a big job but keep synchronous behavior for which just a few jobs are scheduled. -- Andrew D. Straw, Ph.D. Research Institute of Molecular Pathology (IMP) Vienna, Austria http://strawlab.org/
I did something similar, but implemented as an evolution of the original "basic" parallelism (see BWA), that: - Moved the splitting of input files into the datatype classes - Allowed any number of inputs to be split, as long as they were the same datatype (so they were mutually consistent - think paired end fastq files) - Allowed other inputs to be shared among jobs - Merged any number of outputs, which merge code implemented in the datatype classes This worked functionally, but the IO required to split large files has proved too much for something like a whole genome (~500GB) I was thinking of something philosophically similar to your dataset container idea, but more in the idea that a dataset is no longer a "file", so the jobs running on subsets of the dataset would just ask for the parts they need. Galaxy would take care of preserving the abstraction that the subset of the dataset is a single input file, perhaps by extracting the subset to a temporary file on local storage. Similarly, the merged outputs would just be held in the target dataset, not copied, thus making the IO cost for the "merge" 0 for the simple case where it is mere concatenation. John Duddy Sr. Staff Software Engineer Illumina, Inc. 9885 Towne Centre Drive San Diego, CA 92121 Tel: 858-736-3584 E-mail: jduddy@illumina.com -----Original Message----- From: galaxy-dev-bounces@lists.bx.psu.edu [mailto:galaxy-dev-bounces@lists.bx.psu.edu] On Behalf Of Andrew Straw Sent: Tuesday, August 02, 2011 7:13 AM To: galaxy-dev@bx.psu.edu Subject: [galaxy-dev] using Galaxy for map/reduce Hi all, I've been investigating use of Galaxy for our lab and it has many attractive aspects -- a big thank you to all involved. We still have a couple of related sticking points, however, that I would like to get the Galaxy developers' feedback on. Basically, I want to use Galaxy to run Map/Reduce type analysis on many initial data files. What I mean is that I want to take many initial datasets (e.g. 250 or more), perhaps already stored in a library, and then apply a workflow to each and every one of them (the Map step). Then, on the many result datasets (one from each of the initial datasets), I want to run a Reduce step which creates a single dataset. I have achieved this in an imperfect and not-quite-working way with a few tricks, but I hope that with a little work, Galaxy could be much better for this type of use case. I have a couple of specific problems and a proposal for a general solution: 1) My first specific problem is that loading many datasets (e.g. 250) into history causes the javascript running locally withing a browser to be extremely slow. 2) My second specific problem is that applying a workflow with N steps to many datasets creates even more datasets (Nx250 additional datasets). In addition to the slow Javascript problem, there seems to be other issues I haven't diagnosed further, but the console in which I'm running run.sh indicates many errors of the type "Exception AssertionError: AssertionError('State <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990> is not present in this identity map',) in <bound method MutableAttrInstanceState._cleanup of <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990>> ignored". Furthermore the webserver gets slow and my nginx frontend proxy gives 504 gateway time-outs. 3) There's no good way to do reduce within Galaxy. Currently I work around this by having a tool type which takes as an input a dataset and then uploads this to a self-written webserver, which then collects such uploads, performs the reduce, and offers a download link for the user to collect the reduced dataset. The user must manually then upload this dataset back into Galaxy for further processing. My proposal for a general solution, and what I'd be interested in feedback on, is an idea of a "dataset container" (this is just a working name). It would look and act much like a dataset in the history, but would in fact be a logical construct that merely bundles together a homogeneous bunch of datasets. When a tool (or a workflow) is applied to a dataset container, Galaxy would automatically create a new container in which each dataset in this new container is the result of running the tool. (Workflows with N steps would thus generate N new containers.) The thing I like about this idea is that it preserves the ability to use tools and workflows on both individual datasets and, with some additional logic, on these new containers. In particular, I don't think the tools and workflows themselves would have to be modified. This would seemingly mitigate the slow Javascript issue by only showing a few items in the history window (even though Galaxy may have launched many jobs in the background). Furthermore, a new Reduce tool type could then act to take a dataset container as input and output a single dataset. A library doesn't seem a good candidate for the dataset container idea I have above. I realize that a library also bundles together datasets, but it has other attributes that don't play well with the above idea (the idea of hierarchically arranged folders and heterogeneous datasets) nor can it be represented in the history. I'm interested in thoughts on this proposal, as I think it would really help us, and I think our use case may be representative of what others might also like to do. I realize that in my text above I write "with some additional logic" to describe the work required to implement this idea, but the fact is that I have very little idea about how much work this would be. So, practically speaking, my question boils down to how hard would implementing this be, given the existing code base and goals? And, would such an implementation - if done to the taste of the Galaxy devs, of course - have a chance of making into the Galaxy distribution? Thanks, Andrew -- Andrew D. Straw, Ph.D. Research Institute of Molecular Pathology (IMP) Vienna, Austria http://strawlab.org/ ___________________________________________________________ Please keep all replies on the list by using "reply all" in your mail client. To manage your subscriptions to this and other Galaxy lists, please use the interface at: http://lists.bx.psu.edu/
Not intending to hijack the thread, but in response to John's comment -- I, too, made a general solution for embarassingly parallel problems but instead of splitting the large files on disk, I just use seek to move the file pointer so each task can grab it's part. On Tue, Aug 2, 2011 at 10:54 AM, Duddy, John <jduddy@illumina.com> wrote:
I did something similar, but implemented as an evolution of the original "basic" parallelism (see BWA), that: - Moved the splitting of input files into the datatype classes - Allowed any number of inputs to be split, as long as they were the same datatype (so they were mutually consistent - think paired end fastq files) - Allowed other inputs to be shared among jobs - Merged any number of outputs, which merge code implemented in the datatype classes
This worked functionally, but the IO required to split large files has proved too much for something like a whole genome (~500GB)
I was thinking of something philosophically similar to your dataset container idea, but more in the idea that a dataset is no longer a "file", so the jobs running on subsets of the dataset would just ask for the parts they need. Galaxy would take care of preserving the abstraction that the subset of the dataset is a single input file, perhaps by extracting the subset to a temporary file on local storage. Similarly, the merged outputs would just be held in the target dataset, not copied, thus making the IO cost for the "merge" 0 for the simple case where it is mere concatenation.
John Duddy Sr. Staff Software Engineer Illumina, Inc. 9885 Towne Centre Drive San Diego, CA 92121 Tel: 858-736-3584 E-mail: jduddy@illumina.com
-----Original Message----- From: galaxy-dev-bounces@lists.bx.psu.edu [mailto:galaxy-dev-bounces@lists.bx.psu.edu] On Behalf Of Andrew Straw Sent: Tuesday, August 02, 2011 7:13 AM To: galaxy-dev@bx.psu.edu Subject: [galaxy-dev] using Galaxy for map/reduce
Hi all,
I've been investigating use of Galaxy for our lab and it has many attractive aspects -- a big thank you to all involved.
We still have a couple of related sticking points, however, that I would like to get the Galaxy developers' feedback on. Basically, I want to use Galaxy to run Map/Reduce type analysis on many initial data files. What I mean is that I want to take many initial datasets (e.g. 250 or more), perhaps already stored in a library, and then apply a workflow to each and every one of them (the Map step). Then, on the many result datasets (one from each of the initial datasets), I want to run a Reduce step which creates a single dataset. I have achieved this in an imperfect and not-quite-working way with a few tricks, but I hope that with a little work, Galaxy could be much better for this type of use case.
I have a couple of specific problems and a proposal for a general solution:
1) My first specific problem is that loading many datasets (e.g. 250) into history causes the javascript running locally withing a browser to be extremely slow.
2) My second specific problem is that applying a workflow with N steps to many datasets creates even more datasets (Nx250 additional datasets). In addition to the slow Javascript problem, there seems to be other issues I haven't diagnosed further, but the console in which I'm running run.sh indicates many errors of the type "Exception AssertionError: AssertionError('State <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990> is not present in this identity map',) in <bound method MutableAttrInstanceState._cleanup of <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990>> ignored". Furthermore the webserver gets slow and my nginx frontend proxy gives 504 gateway time-outs.
3) There's no good way to do reduce within Galaxy. Currently I work around this by having a tool type which takes as an input a dataset and then uploads this to a self-written webserver, which then collects such uploads, performs the reduce, and offers a download link for the user to collect the reduced dataset. The user must manually then upload this dataset back into Galaxy for further processing.
My proposal for a general solution, and what I'd be interested in feedback on, is an idea of a "dataset container" (this is just a working name). It would look and act much like a dataset in the history, but would in fact be a logical construct that merely bundles together a homogeneous bunch of datasets. When a tool (or a workflow) is applied to a dataset container, Galaxy would automatically create a new container in which each dataset in this new container is the result of running the tool. (Workflows with N steps would thus generate N new containers.) The thing I like about this idea is that it preserves the ability to use tools and workflows on both individual datasets and, with some additional logic, on these new containers. In particular, I don't think the tools and workflows themselves would have to be modified. This would seemingly mitigate the slow Javascript issue by only showing a few items in the history window (even though Galaxy may have launched many jobs in the background). Furthermore, a new Reduce tool type could then act to take a dataset container as input and output a single dataset.
A library doesn't seem a good candidate for the dataset container idea I have above. I realize that a library also bundles together datasets, but it has other attributes that don't play well with the above idea (the idea of hierarchically arranged folders and heterogeneous datasets) nor can it be represented in the history.
I'm interested in thoughts on this proposal, as I think it would really help us, and I think our use case may be representative of what others might also like to do. I realize that in my text above I write "with some additional logic" to describe the work required to implement this idea, but the fact is that I have very little idea about how much work this would be. So, practically speaking, my question boils down to how hard would implementing this be, given the existing code base and goals? And, would such an implementation - if done to the taste of the Galaxy devs, of course - have a chance of making into the Galaxy distribution?
Thanks, Andrew
-- Andrew D. Straw, Ph.D. Research Institute of Molecular Pathology (IMP) Vienna, Austria http://strawlab.org/
___________________________________________________________ Please keep all replies on the list by using "reply all" in your mail client. To manage your subscriptions to this and other Galaxy lists, please use the interface at:
___________________________________________________________ Please keep all replies on the list by using "reply all" in your mail client. To manage your subscriptions to this and other Galaxy lists, please use the interface at:
Many of the tools out there work on files, and assume they are supposed to work on the whole file (or take arguments for subsets that vary from tool to tool). I'm working on a way for Galaxy to handle all these tools transparently, even if, as in my case, the files are compressed but the tools cannot read compressed files. John Duddy Sr. Staff Software Engineer Illumina, Inc. 9885 Towne Centre Drive San Diego, CA 92121 Tel: 858-736-3584 E-mail: jduddy@illumina.com -----Original Message----- From: Edward Kirton [mailto:eskirton@lbl.gov] Sent: Friday, August 26, 2011 12:34 PM To: Duddy, John Cc: galaxy-dev@bx.psu.edu Subject: Re: [galaxy-dev] using Galaxy for map/reduce Not intending to hijack the thread, but in response to John's comment -- I, too, made a general solution for embarassingly parallel problems but instead of splitting the large files on disk, I just use seek to move the file pointer so each task can grab it's part. On Tue, Aug 2, 2011 at 10:54 AM, Duddy, John <jduddy@illumina.com> wrote:
I did something similar, but implemented as an evolution of the original "basic" parallelism (see BWA), that: - Moved the splitting of input files into the datatype classes - Allowed any number of inputs to be split, as long as they were the same datatype (so they were mutually consistent - think paired end fastq files) - Allowed other inputs to be shared among jobs - Merged any number of outputs, which merge code implemented in the datatype classes
This worked functionally, but the IO required to split large files has proved too much for something like a whole genome (~500GB)
I was thinking of something philosophically similar to your dataset container idea, but more in the idea that a dataset is no longer a "file", so the jobs running on subsets of the dataset would just ask for the parts they need. Galaxy would take care of preserving the abstraction that the subset of the dataset is a single input file, perhaps by extracting the subset to a temporary file on local storage. Similarly, the merged outputs would just be held in the target dataset, not copied, thus making the IO cost for the "merge" 0 for the simple case where it is mere concatenation.
John Duddy Sr. Staff Software Engineer Illumina, Inc. 9885 Towne Centre Drive San Diego, CA 92121 Tel: 858-736-3584 E-mail: jduddy@illumina.com
-----Original Message----- From: galaxy-dev-bounces@lists.bx.psu.edu [mailto:galaxy-dev-bounces@lists.bx.psu.edu] On Behalf Of Andrew Straw Sent: Tuesday, August 02, 2011 7:13 AM To: galaxy-dev@bx.psu.edu Subject: [galaxy-dev] using Galaxy for map/reduce
Hi all,
I've been investigating use of Galaxy for our lab and it has many attractive aspects -- a big thank you to all involved.
We still have a couple of related sticking points, however, that I would like to get the Galaxy developers' feedback on. Basically, I want to use Galaxy to run Map/Reduce type analysis on many initial data files. What I mean is that I want to take many initial datasets (e.g. 250 or more), perhaps already stored in a library, and then apply a workflow to each and every one of them (the Map step). Then, on the many result datasets (one from each of the initial datasets), I want to run a Reduce step which creates a single dataset. I have achieved this in an imperfect and not-quite-working way with a few tricks, but I hope that with a little work, Galaxy could be much better for this type of use case.
I have a couple of specific problems and a proposal for a general solution:
1) My first specific problem is that loading many datasets (e.g. 250) into history causes the javascript running locally withing a browser to be extremely slow.
2) My second specific problem is that applying a workflow with N steps to many datasets creates even more datasets (Nx250 additional datasets). In addition to the slow Javascript problem, there seems to be other issues I haven't diagnosed further, but the console in which I'm running run.sh indicates many errors of the type "Exception AssertionError: AssertionError('State <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990> is not present in this identity map',) in <bound method MutableAttrInstanceState._cleanup of <sqlalchemy.orm.state.MutableAttrInstanceState object at 0x7f5c18c47990>> ignored". Furthermore the webserver gets slow and my nginx frontend proxy gives 504 gateway time-outs.
3) There's no good way to do reduce within Galaxy. Currently I work around this by having a tool type which takes as an input a dataset and then uploads this to a self-written webserver, which then collects such uploads, performs the reduce, and offers a download link for the user to collect the reduced dataset. The user must manually then upload this dataset back into Galaxy for further processing.
My proposal for a general solution, and what I'd be interested in feedback on, is an idea of a "dataset container" (this is just a working name). It would look and act much like a dataset in the history, but would in fact be a logical construct that merely bundles together a homogeneous bunch of datasets. When a tool (or a workflow) is applied to a dataset container, Galaxy would automatically create a new container in which each dataset in this new container is the result of running the tool. (Workflows with N steps would thus generate N new containers.) The thing I like about this idea is that it preserves the ability to use tools and workflows on both individual datasets and, with some additional logic, on these new containers. In particular, I don't think the tools and workflows themselves would have to be modified. This would seemingly mitigate the slow Javascript issue by only showing a few items in the history window (even though Galaxy may have launched many jobs in the background). Furthermore, a new Reduce tool type could then act to take a dataset container as input and output a single dataset.
A library doesn't seem a good candidate for the dataset container idea I have above. I realize that a library also bundles together datasets, but it has other attributes that don't play well with the above idea (the idea of hierarchically arranged folders and heterogeneous datasets) nor can it be represented in the history.
I'm interested in thoughts on this proposal, as I think it would really help us, and I think our use case may be representative of what others might also like to do. I realize that in my text above I write "with some additional logic" to describe the work required to implement this idea, but the fact is that I have very little idea about how much work this would be. So, practically speaking, my question boils down to how hard would implementing this be, given the existing code base and goals? And, would such an implementation - if done to the taste of the Galaxy devs, of course - have a chance of making into the Galaxy distribution?
Thanks, Andrew
-- Andrew D. Straw, Ph.D. Research Institute of Molecular Pathology (IMP) Vienna, Austria http://strawlab.org/
___________________________________________________________ Please keep all replies on the list by using "reply all" in your mail client. To manage your subscriptions to this and other Galaxy lists, please use the interface at:
___________________________________________________________ Please keep all replies on the list by using "reply all" in your mail client. To manage your subscriptions to this and other Galaxy lists, please use the interface at:
yes, many tools don't read from stdin, you're right. in practice, i actually have each task write it's part to the node's local scratch disk and also do implicit conversions in this step as well (e.g. scatter fastq as fasta). but not all clusters have a local scratchdisk. also, as you mentioned, the seek solution wouldn't work for compressed infiles. as i try to avoid working on the galaxy internals, i implemented this as a command-line utility. e.g. <command>psub --fastqToFasta $infile --cat $outfile qctool.py $infile $outfile</command> instead of the nonparallel: <command>qctool.py $infile $outfile</command> but it would be nice to see this functionality in galaxy. i thought about reimplementing this as a drmaa_epc.py job runner but noticed there was already tasks.py. On Fri, Aug 26, 2011 at 12:41 PM, Duddy, John <jduddy@illumina.com> wrote:
Many of the tools out there work on files, and assume they are supposed to work on the whole file (or take arguments for subsets that vary from tool to tool).
I'm working on a way for Galaxy to handle all these tools transparently, even if, as in my case, the files are compressed but the tools cannot read compressed files.
John Duddy Sr. Staff Software Engineer Illumina, Inc. 9885 Towne Centre Drive San Diego, CA 92121 Tel: 858-736-3584 E-mail: jduddy@illumina.com
-----Original Message----- From: Edward Kirton [mailto:eskirton@lbl.gov] Sent: Friday, August 26, 2011 12:34 PM To: Duddy, John Cc: galaxy-dev@bx.psu.edu Subject: Re: [galaxy-dev] using Galaxy for map/reduce
Not intending to hijack the thread, but in response to John's comment -- I, too, made a general solution for embarassingly parallel problems but instead of splitting the large files on disk, I just use seek to move the file pointer so each task can grab it's part.
participants (6)
-
Andrew Straw -
Duddy, John -
Edward Kirton -
James Taylor -
Peter Cock -
Ravi Madduri