1 / 47

Scaling Up Data Intensive Science to Campus Grids

In this talk, Douglas Thain from Clemson University explores the concept of campus grids, which aggregate institutional computing resources for large-scale computing problems. This approach enables efficient usage of idle cycles from desktops and clusters, providing robust batch queueing across distributed systems. Thain highlights the practical implications through real-world examples, emphasizing the development of abstractions for distributed computing that improve resource allocation and job execution. The goal is to facilitate significant biometric research by reducing computation time for complex workloads.

yolanda-sullivan
Télécharger la présentation

Scaling Up Data Intensive Science to Campus Grids

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Scaling UpData Intensive Scienceto Campus Grids Douglas Thain Clemson University 25 Septmber 2009

  2. http://www.cse.nd.edu/~ccl

  3. The Cooperative Computing Lab • We collaborate with people who have large scale computing problems. • We build new software and systems to help them achieve meaningful goals. • We run a production computing system used by people at ND and elsewhere. • We conduct computer science research, informed by real world experience, with an impact upon problems that matter.

  4. What is a Campus Grid? • A campus grid is an aggregation of all available computing power found in an institution: • Idle cycles from desktop machines. • Unused cycles from dedicated clusters. • Examples of campus grids: • 700 CPUs at the University of Notre Dame • 9000-11,000 CPUs at Clemson University • 20,000 CPUs at Purdue University

  5. Provides robust batch queueing on a complex distributed system. • Resource owners control consumption: • “Only run jobs on this machine at night.” • “Prefer biology jobs over physics jobs.” • End users express needs: • “Only run this job where RAM>2GB” • “Prefer to run on machines http://www.cs.wisc.edu/condor

  6. Clusters, clouds, and gridsgive us access to unlimited CPUs. How do we write programs that canrun effectively in large systems?

  7. F F 0.97 0.05 Example: Biometrics Research • Goal: Design robust face comparison function.

  8. Similarity Matrix Construction Challenge Workload: 60,000 iris images 1MB each .02s per F 833 CPU-days 600 TB of I/O

  9. I have 60,000 iris images acquired in my research lab. I want to reduce each one to a feature space, and then compare all of them to each other. I want to spend my time doing science, not struggling with computers. I own a few machines I can buy time from Amazon or TeraGrid. I have a laptop. Now What?

  10. Try 1: Each F is a batch job. Failure: Dispatch latency >> F runtime. Try 2: Each row is a batch job. Failure: Too many small ops on FS. F F F F F CPU CPU CPU CPU CPU F F F F F F F F F F CPU F CPU F CPU F CPU F CPU F F F F F F HN HN Try 3: Bundle all files into one package. Failure: Everyone loads 1GB at once. Try 4: User gives up and attempts to solve an easier or smaller problem. F F F F F F F F F F CPU F CPU F CPU F CPU F CPU F F F F F F HN Non-Expert User Using 500 CPUs

  11. Observation • In a given field of study, many people repeat the same pattern of work many times, making slight changes to the data and algorithms. • If the system knows the overall pattern in advance, then it can do a better job of executing it reliably and efficiently. • If the user knows in advance what patterns are allowed, then they have a better idea of how to construct their workloads.

  12. Abstractionsfor Distributed Computing • Abstraction: a declarative specification of the computation and data of a workload. • A restricted pattern, not meant to be a general purpose programming language. • Usesdata structures instead of files. • Provide users with a bright path. • Regular structure makes it tractable to model and predict performance.

  13. Working with Abstractions A1 A1 A2 A2 F An Bn AllPairs( A, B, F ) Custom Workflow Engine Cloud or Grid Compact Data Structure

  14. All-Pairs Abstraction AllPairs( set A, set B, function F ) returns matrix M where M[i][j] = F( A[i], B[j] ) for all i,j A1 A2 A3 A1 A1 allpairs A B F.exe An AllPairs(A,B,F) B1 F F F B1 B1 Bn B2 F F F F B3 F F F

  15. How Does the Abstraction Help? • The custom workflow engine: • Chooses right data transfer strategy. • Chooses the right number of resources. • Chooses blocking of functions into jobs. • Recovers from a larger number of failures. • Predicts overall runtime accurately. • All of these tasks are nearly impossible for arbitrary workloads, but are tractable (not trivial) to solve for a specific abstraction.

  16. Choose the Right # of CPUs

  17. Resources Consumed

  18. All-Pairs in Production • Our All-Pairs implementation has provided over 57 CPU-years of computation to the ND biometrics research group over the last year. • Largest run so far: 58,396 irises from the Face Recognition Grand Challenge. The largest experiment ever run on publically available data. • Competing biometric research relies on samples of 100-1000 images, which can miss important population effects. • Reduced computation time from 833 days to 10 days, making it feasible to repeat multiple times for a graduate thesis. (We can go faster yet.)

  19. Are there other abstractions?

  20. All-Pairs Abstraction AllPairs( set A, set B, function F ) returns matrix M where M[i][j] = F( A[i], B[j] ) for all i,j A1 A2 A3 A1 A1 allpairs A B F.exe An AllPairs(A,B,F) B1 F F F B1 B1 Bn B2 F F F F B3 F F F

  21. M[0,4] M[2,4] M[3,4] M[4,4] F x d y M[0,3] M[3,2] M[4,3] x F F x d y d y M[0,2] M[4,2] x F x F F x d y d y d y M[0,1] F F F F x x x x d y d y d y d y M[0,0] M[1,0] M[2,0] M[3,0] M[4,0] Wavefront( matrix M, function F(x,y,d) ) returns matrix M such that M[i,j] = F( M[i-1,j], M[I,j-1], M[i-1,j-1] ) Wavefront(M,F) M F

  22. Some-Pairs Abstraction SomePairs( set A, list (i,j), function F(x,y) ) returns list of F( A[i], A[j] ) A1 A2 A3 A1 A1 An SomePairs(A,L,F) A1 F (1,2) (2,1) (2,3) (3,3) A2 F F F A3 F

  23. What if your application doesn’t fit a regular pattern?

  24. Makeflow part1 part2 part3: input.data split.py ./split.py input.data out1: part1 mysim.exe ./mysim.exe part1 >out1 out2: part2 mysim.exe ./mysim.exe part2 >out2 out3: part3 mysim.exe ./mysim.exe part3 >out3 result: out1 out2 out3 join.py ./join.py out1 out2 out3 > result

  25. Makeflow Implementation 100s of workers dispatched to the cloud worker worker worker bfile: afile prog prog afile >bfile worker worker worker queue tasks detail of a single worker: put prog put afile exec prog afile > bfile get bfile makeflow master work queue worker tasks done Two optimizations: Cache inputs and output. Dispatch tasks to nodes with data. prog afile bfile

  26. Experience with Makeflow • Reusing a good old idea in a new way. • Easy to test and debug on a desktop machine or a multicore server. • The workload says nothing about the distributed system. (This is good.) • Graduate students in bioinformatics running codes at production speeds on hundreds of nodes in less than a week. • Student from Clemson got complex biometrics workload running in a few weeks.

  27. Putting it All Together Web Portal Abstraction Y F X Z Data Repository Campus Grid

  28. BXGrid Schema Immutable Replicas Scientific Metadata replicaid=423 state=ok replicaid=105 state=ok replicaid=293 state=creating General Metadata fileid = 24305 size = 300K type = jpg sum = abc123… replicaid=102 state=deleting

  29. Results from Campus Grid

  30. Biocompute

  31. Parallel BLAST Makeflow

  32. Abstractions as a Social Tool • Collaboration with outside groups is how we encounter the most interesting, challenging, and important problems, in computer science. • However, often neither side understands which details are essential or non-essential: • Can you deal with files that have upper case letters? • Oh, by the way, we have 10TB of input, is that ok? • (A little bit of an exaggeration.) • An abstraction is an excellent chalkboard tool: • Accessible to anyone with a little bit of mathematics. • Makes it easy to see what must be plugged in. • Forces out essential details: data size, execution time.

  33. Conclusion • Grids, clouds, and clusters provide enormous computing power, but are very challenging to use effectively. • An abstraction provides a robust, scalable solution to a narrow category of problems; each requires different kinds of optimizations. • Limiting expressive power, results in systems that are usable, predictable, and reliable. • Portal + Repository + Abstraction + Grid = New Science Capabilities

  34. Acknowledgments • Cooperative Computing Lab • http://www.cse.nd.edu/~ccl • Undergrads • Mike Kelly • Rory Carmichael • Mark Pasquier • Christopher Lyon • Jared Bulosan • Kameron Srimoungach • Rachel Witty • Ryan Jansen • Joey Rich • Faculty: • Patrick Flynn • Nitesh Chawla • Kenneth Judd • Scott Emrich • Grad Students • Chris Moretti • Hoang Bui • Li Yu • Mike Olson • Michael Albrecht • NSF Grants CCF-0621434 and CNS-0643229

More Related