1 / 28

Large-Scale Iterative Data Processing CS525 Big Data Analytics

Large-Scale Iterative Data Processing CS525 Big Data Analytics. paper “ HaLoop : Efficient Iterative Data Processing On Large Scale Clusters” by Yingyi Bu, UC Irvine; Bill Howe, UW; Magda Balazinska , UW; and Michael Ernst, UW in VLDB’2010. VLDB 2010, Singapore. Observation.

howard
Download Presentation

Large-Scale Iterative Data Processing CS525 Big Data Analytics

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. Large-Scale Iterative Data ProcessingCS525 Big Data Analytics paper “HaLoop: Efficient Iterative Data Processing On Large Scale Clusters” by Yingyi Bu, UC Irvine;Bill Howe, UW; Magda Balazinska, UW; and Michael Ernst, UW in VLDB’2010 VLDB 2010, Singapore

  2. Observation • Observation:MapReduce has proven successful as a common runtime for non-recursive declarative languages • HIVE (SQL-like language) • Pig (Rel algebra with nested types) • Observation: Loops and iteration everywhere : • Graphs, clustering, mining • Problem: • MapReduce can’t express recursion/iteration • Observation: Many roll their own loops: • Iteration managed by hard-coded script outside MapReduce Bill Howe, UW

  3. O(N) in the size of the graph Twister – PageRank [Ekanayake HPDC 2010] while (!complete) { // start the pagerank map reduce process monitor = driver.runMapReduceBCast(new BytesValue(tmpCompressedDvd.getBytes())); monitor.monitorTillCompletion(); // get the result of process newCompressedDvd = ((PageRankCombiner) driver.getCurrentCombiner()).getResults(); // decompress the compressed pagerank values newDvd = decompress(newCompressedDvd); tmpDvd = decompress(tmpCompressedDvd); totalError = getError(tmpDvd, newDvd); // get the difference between new and old pagerank values if (totalError < tolerance) { complete = true; } tmpCompressedDvd = newCompressedDvd; } run MR term. cond. Bill Howe, UW

  4. Key idea • When the loop output is large… • transitive closure • connected components • PageRank (with a convergence test as the termination condition) • …need a distributed fixpoint operator • typically implemented as yet another MapReduce job -- on every iteration Bill Howe, UW

  5. Fixpoint • A fixpoint of a function f is a value x such that f(x) = x • Fixpoint queries can be expressed with relational algebra plus a fixpoint operator • Map - Reduce - Fixpoint • Hypothesis: model for all recursive queries Bill Howe, UW

  6. Key Vision • Thought: • Hypothesis: making MapReduce loop-aware affords optimization • …and lays foundation for scalable implementations of recursive languages • More precisely: • With minimal extensions, we provide efficient common runtime for recursive languages: • Map, Reduce, Fixpoint Bill Howe, UW

  7. Three Examples With Iteration Bill Howe, UW

  8. Example 1: PageRank (Web Link Analysis) Rank Table R0 Linkage Table L Ri+1 π(url_dest, γurl_destSUM(rank)) Ri.rank = Ri.rank/γurlCOUNT(url_dest) Rank Table R3 Ri.url = L.url_src Ri L Bill Howe, UW

  9. A MapReduce Implementation Join & compute rank Aggregate fixpoint evaluation Ri M M r M r r L-split0 M r M r M r L-split1 M Converged? i=i+1 Client done Bill Howe, UW

  10. What’s the problem? Ri m M r M r r L-split0 m r M r M r 3. L-split1 2. m 1. L (Link Table) is loop invariant, but plus Bill Howe, UW L is loaded on each iteration L is shuffled on each iteration Fixpoint evaluated as a separate MapReduce job per iteration

  11. Example 2: Transitive Closure Friend Find all transitive friends of Eric R0 {Eric, Eric} {Eric, Elisa} R1 {Eric, Tom Eric, Harry} R2 R3 {} (semi-naïve evaluation) Bill Howe, UW

  12. Example 2 in MapReduce (compute next generation of friends) (remove the ones we’ve already seen) Join Dupe-elim Si M M r r Friend0 M r M r Friend1 M Anything new? Client i=i+1 done Bill Howe, UW

  13. What’s the problem? (compute next generation of friends) (remove the ones we’ve already seen) Join Dupe-elim Si M M r r Friend0 M r M r Friend1 2. M 1. Friend is loop invariant, but Bill Howe, UW Friend is loaded on each iteration Friend is shuffled on each iteration

  14. Example 3: k-means ki = k centroids at iteration i ki P0 M r ki P1 ki+1 M r ki P2 M ki - ki+1 < threshold? Client i=i+1 done Bill Howe, UW

  15. What’s the problem? ki = k centroids at iteration i ki P0 M r ki P1 ki+1 M r ki P2 M 1. ki - ki+1 < threshold? Client i=i+1 done P is loop invariant, but • P is loaded on each iteration Bill Howe, UW

  16. Optimizations Enabled • Caching (and Indexing) Bill Howe, UW

  17. M … r M r M Approach: Inter-iteration caching Loop body Reducer output cache (RO) Reducer input cache (RI) Mapper output cache (MO) Mapper input cache (MI) Bill Howe, UW

  18. Experiments Bill Howe, UW • Amazon EC2 • 20, 50, 90 default small instances • Datasets • Billions of Triples (120GB) [1.5B nodes 1.6B edges] • Freebase (12GB) [7M ndoes 154M edges] • Livejournal social network (18GB) [4.8M nodes, 67M edges] • Queries • Transitive Closure • PageRank • k-means [VLDB 2010]

  19. RI: Reducer Input Cache • Provides: • Access to loop invariant data without map/shuffle • Used By: • Reducer function • Assumes: • Mapper output for a given table constant across iterations • Static partitioning (implies: no new nodes) • PageRank • Avoid shuffling the network at every step • Transitive Closure • Avoid shuffling the graph at every step • K-means • No help … Bill Howe, UW

  20. Reducer Input Cache Benefit Transitive Closure like PageRank Billion Triples Dataset (120GB) 90 small instances on EC2 (on Amazon Cloud EC2) Overall run time Bill Howe, UW

  21. Reducer Input Cache Benefit Transitive Closure Billion Triples Dataset (120GB) Join step only (key step optimized) Bill Howe, UW

  22. RO: Reducer Output Cache • Provides: • Distributed access to output of previous iterations • Used By: • Fixpoint evaluation • Assumes: • Partitioning constant across iterations • Reducer output key functionally determines Reducer input key • PageRank • Allows distributed fixpoint evaluation • Obviates extra MapReduce job • Transitive Closure • No help? • K-means • No help? … Bill Howe, UW

  23. Reducer Output Cache Benefit Fixpoint evaluation (s) Iteration # Livejournal dataset 50 EC2 small instances

  24. MI: Mapper Input Cache • Provides: • Access to non-local mapper input on later iterations • Used: • During scheduling of map tasks • Assumes: • Mapper input does not change • PageRank • Subsumed by use of Reducer Input Cache • Transitive Closure • Subsumed by use of Reducer Input Cache • K-means • Avoids non-local data reads on iterations > 0 … Bill Howe, UW

  25. Mapper Input Cache Benefit 5% non-local data reads; ~5% improvement Bill Howe, UW

  26. HaLoop Architecture Bill Howe, UW

  27. Programming Interface: Iteration Support Job job = new Job(); job.AddMap(Map Rank, 1); job.AddReduce(Reduce Rank, 1); job.AddMap(Map Aggregate, 2); job.AddReduce(Reduce Aggregate, 2); job.AddInvariantTable(#1); job.SetInput(IterationInput); job.SetFixedPointThreshold(0.1); job.SetDistanceMeasure(ResultDistance); job.SetMaxNumOfIterations(10); job.SetReducerInputCache(true); job.SetReducerOutputCache(true); job.Submit(); define loop body Declare an input as invariant Specify loop body input, parameterized by iteration # Termination condition Turn on caches Bill Howe, UW

  28. Conclusions • Relatively simple changes to MapReduce/Hadoop can support arbitrary recursive programs • TaskTracker (Cache management) • Scheduler (Cache awareness) • Programming model (multi-step loop bodies, cache control) • Optimizations • Caching loop invariant data realizes largest gain • Eliminate extra MapReduce step for termination checks Bill Howe, UW

More Related