1. Dryad and DryadLINQ Presented by Yin Zhu April 22, 2013 Slides taken from DryadLINQ project page: http://research.microsoft.com/en-us/ projects/dryadlinq/default.aspx

2. Distributed Data-Parallel Programming using Dryad, EuroSys’07 Andrew Birrell, Mihai Budiu, Dennis Fetterly, Michael Isard, Yuan Yu Microsoft Research Silicon Valley

3. Dryad goals General-purpose execution environment for distributed, data-parallel applications –Concentrates on throughput not latency –Assumes private data center Automatic management of scheduling, distribution, fault tolerance, etc.

4. Talk outline Computational model Dryad architecture Some case studies DryadLINQ overview Summary

5. A typical data-intensive query var logentries = from line in logs where !line.StartsWith("#") select new LogEntry(line); var user = from access in logentries where access.user.EndsWith(@"\ulfar") select access; var accesses = from access in user group access by access.page into pages select new UserPageCount("ulfar", pages.Key, pages.Count()); var htmAccesses = from access in accesses where access.page.EndsWith(".htm") orderby access.count descending select access; Ulfar’s most frequently visited web pages

6. Steps in the query var logentries = from line in logs where !line.StartsWith("#") select new LogEntry(line); var user = from access in logentries where access.user.EndsWith(@"\ulfar") select access; var accesses = from access in user group access by access.page into pages select new UserPageCount("ulfar", pages.Key, pages.Count()); var htmAccesses = from access in accesses where access.page.EndsWith(".htm") orderby access.count descending select access; Go through logs and keep only lines that are not comments. Parse each line into a LogEntry object. Go through logentries and keep only entries that are accesses by ulfar. Group ulfar ’s accesses according to what page they correspond to. For each page, count the occurrences. Sort the pages ulfar has accessed according to access frequency.

7. Serial execution var logentries = from line in logs where !line.StartsWith("#") select new LogEntry(line); var user = from access in logentries where access.user.EndsWith(@"\ulfar") select access; var accesses = from access in user group access by access.page into pages select new UserPageCount("ulfar", pages.Key, pages.Count()); var htmAccesses = from access in accesses where access.page.EndsWith(".htm") orderby access.count descending select access; For each line in logs, do… For each entry in logentries, do.. Sort entries in user by page. Then iterate over sorted list, counting the occurrences of each page as you go. Re-sort entries in access by page frequency.

8. Parallel execution var logentries = from line in logs where !line.StartsWith("#") select new LogEntry(line); var user = from access in logentries where access.user.EndsWith(@"\ulfar") select access; var accesses = from access in user group access by access.page into pages select new UserPageCount("ulfar", pages.Key, pages.Count()); var htmAccesses = from access in accesses where access.page.EndsWith(".htm") orderby access.count descending select access;

9. How does Dryad fit in? Many programs can be represented as a distributed execution graph –The programmer may not have to know this “SQL-like” queries: LINQ –Spark (OSDI’12) utilizes the same idea. Dryad will run them for you

10. Talk outline Computational model Dryad architecture Some case studies DryadLINQ overview Summary

11. Runtime Services –Name Server –Daemon Job Manager –Centralized coordinating process –User application to construct graph –Linked with Dryad libraries for scheduling vertices Vertex executable –Dryad libraries to communicate with JM –User application sees channels in/out –Arbitrary application code, can use local FS

12. Job = Directed Acyclic Graph Processing vertices Channels (file, pipe, shared memory) Inputs Outputs

13. What’s wrong with MapReduce? Literally Map then Reduce and that’s it… –Reducers write to replicated storage Complex jobs pipeline multiple stages –No fault tolerance between stages Map assumes its data is always available: simple! Output of Reduce: 2 network copies, 3 disks –In Dryad this collapses inside a single process –Big jobs can be more efficient with Dryad

14. What’s wrong with Map+Reduce? Join combines inputs of different types “Split” produces outputs of different types –Parse a document, output text and references Can be done with Map+Reduce –Ugly to program –Hard to avoid performance penalty –Some merge joins very expensive Need to materialize entire cross product to disk

15. How about Map+Reduce+Join+…? “Uniform” stages aren’t really uniform

16. How about Map+Reduce+Join+…? “Uniform” stages aren’t really uniform

17. Graph complexity composes Non-trees common E.g. data-dependent re-partitioning –Combine this with merge trees etc. Distribute to equal-sized ranges Sample to estimate histogram Randomly partitioned inputs

18. Scheduler state machine Scheduling is independent of semantics –Vertex can run anywhere once all its inputs are ready Constraints/hints place it near its inputs –Fault tolerance If A fails, run it again If A’s inputs are gone, run upstream vertices again (recursively) If A is slow, run another copy elsewhere and use output from whichever finishes first

19. Dryad DAG architecture Simplicity depends on generality –Front ends only see graph data-structures –Generic scheduler state machine Software engineering: clean abstraction Restricting set of operations would pollute scheduling logic with execution semantics Optimizations all “above the fold” –Dryad exports callbacks so applications can react to state machine transitions

20. Talk outline Computational model Dryad architecture Some case studies DryadLINQ overview Summary

21. SkyServer DB Query 3-way join to find gravitational lens effect Table U: (objId, color) 11.8GB Table N: (objId, neighborId) 41.8GB Find neighboring stars with similar colors: –Join U+N to find T = U.color,N.neighborId where U.objId = N.objId –Join U+T to find U.objId where U.objId = T.neighborID and U.color ≈ T.color

22. Took SQL plan Manually coded in Dryad Manually partitioned data SkyServer DB query u: objid, color n: objid, neighborobjid [partition by objid] select u.color,n.neighborobjid from u join n where u.objid = n.objid (u.color,n.neighborobjid) [re-partition by n.neighborobjid] [order by n.neighborobjid] [distinct] [merge outputs] select u.objid from u join where u.objid =.neighborobjid and |u.color -.color| < d

23. Optimization D M S Y X M S M S M S UN U

24. D M S Y X M S M S M S UN U

25. 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 0246810 Number of Computers Speed-up Dryad In-Memory Dryad Two-pass SQLServer 2005

26. Query histogram computation Input: log file (n partitions) Extract queries from log partitions Re-partition by hash of query (k buckets) Compute histogram within each bucket

27. Naïve histogram topology Pparse lines D hash distribute S quicksort C count occurrences MSmerge sort

28. Efficient histogram topology Pparse lines D hash distribute S quicksort C count occurrences MSmerge sort M non-deterministic merge Q' is:Each R is: Each MS C M P C S Q' RR k T k n T is: Each MS D C

29. RR T Q’ MS►C►D M►P►S►C MS►C Pparse linesDhash distribute SquicksortMSmerge sort Ccount occurrencesMnon-deterministic merge R

30. MS►C►D M►P►S►C MS►C Pparse linesDhash distribute SquicksortMSmerge sort Ccount occurrencesMnon-deterministic merge RR T R Q’

31. MS►C►D M►P►S►C MS►C Pparse linesDhash distribute SquicksortMSmerge sort Ccount occurrencesMnon-deterministic merge RR T R Q’ T

32. MS►C►D M►P►S►C MS►C Pparse linesDhash distribute SquicksortMSmerge sort Ccount occurrencesMnon-deterministic merge RR T R Q’ T

33. Pparse linesDhash distribute SquicksortMSmerge sort Ccount occurrencesMnon-deterministic merge MS►C►D M►P►S►C MS►C RR T R Q’ T

34. Pparse linesDhash distribute SquicksortMSmerge sort Ccount occurrencesMnon-deterministic merge MS►C►D M►P►S►C MS►C RR T R Q’ T

35. Final histogram refinement 1,800 computers 43,171 vertices 11,072 processes 11.5 minutes

36. Optimizing Dryad applications General-purpose refinement rules Processes formed from subgraphs –Re-arrange computations, change I/O type Application code not modified –System at liberty to make optimization choices High-level front ends hide this from user –SQL query planner, etc.

37. Talk outline Computational model Dryad architecture Some case studies DryadLINQ overview Summary

38. DryadLINQ (Yuan Yu et. al, OSDI’08) LINQ: Relational queries integrated in C# More general than distributed SQL –Inherits flexible C# type system and libraries –Data-clustering, EM, inference, … Uniform data-parallel programming model –From SMP to clusters

39. LINQ System Architecture Parallel LINQ Local machine.Net program (C#, VB, F#, etc) Execution engines Query Objects LINQ-to-SQL DryadLINQ LINQ-to-Obj LINQ provider interface Scalability Single-core Multi-core Cluster

40. DryadLINQ System Architecture 40 DryadLINQ Client machine (11) Distributed query plan.NET program Query Expr Data center Output Tables Result s Input Tables Invoke Query Output DryadTable Dryad Executio n.Net Objects JM ToTable foreach Vertex code

41. An Example: PageRank Ranks web pages by propagating scores along hyperlink structure Each iteration as an SQL query: 1.Join edges with ranks 2.Distribute ranks on edges 3.GroupBy edge destination 4.Aggregate into ranks 5.Repeat

42. One PageRank Step in DryadLINQ // one step of pagerank: dispersing and re-accumulating rank public static IQueryable PRStep(IQueryable pages, IQueryable ranks) { // join pages with ranks, and disperse updates var updates = from page in pages join rank in ranks on page.name equals rank.name select page.Disperse(rank); // re-accumulate. return from list in updates from rank in list group rank.rank by rank.name into g select new Rank(g.Key, g.Sum()); }

43. The Complete PageRank Program var pages = DryadLinq.GetTable (“file://pages.txt”); var ranks = pages.Select(page => new Rank(page.name, 1.0)); // repeat the iterative computation several times for (int iter = 0; iter < iterations; iter++) { ranks = PRStep(pages, ranks); } ranks.ToDryadTable (“outputranks.txt”); public struct Page { public UInt64 name; public Int64 degree; public UInt64[] links; public Page(UInt64 n, Int64 d, UInt64[] l) { name = n; degree = d; links = l; } public Rank[] Disperse(Rank rank) { Rank[] ranks = new Rank[links.Length]; double score = rank.rank / this.degree; for (int i = 0; i < ranks.Length; i++) { ranks[i] = new Rank(this.links[i], score); } return ranks; } } public struct Rank { public UInt64 name; public double rank; public Rank(UInt64 n, double r) { name = n; rank = r; } } public static IQueryable PRStep(IQueryable pages, IQueryable ranks) { // join pages with ranks, and disperse updates var updates = from page in pages join rank in ranks on page.name equals rank.name select page.Disperse(rank); // re-accumulate. return from list in updates from rank in list group rank.rank by rank.name into g select new Rank(g.Key, g.Sum()); }

44. One Iteration PageRank J S G C D M G R J S G C D M G R J S G C D Join pages and ranks Disperse page’s rank Group rank by page Accumulate ranks, partially Hash distribute Merge the data Group rank by page Accumulate ranks M G R … … Dynamic aggregation

45. Multi-Iteration PageRank pagesranks Iteration 1 Iteration 2 Iteration 3 Memory FIFO

46. Expectation Maximization 190 lines 3 iterations shown

47. Understanding Botnet Traffic using EM 3 GB data 15 clusters 60 computers 50 iterations 9000 processes 50 minutes

48. Image Processing Cosmos DFSSQL Servers Software Stack 48 Windows Server Cluster Services Azure Platform Dryad DryadLINQ Windows Server Other Languages CIFS/NTFS Machine Learning Graph Analysis Data Mining Applications … Other Applications

49. Lessons Deep language integration worked out well –Easy expression of massive parallelism –Elegant, unified data model based on.NET objects –Multiple language support: C#, VB, F#, … –Visual Studio and.NET libraries –Interoperate with PLINQ, LINQ-to-SQL, LINQ-to- Object, … Key enablers –Language side LINQ extensibility: custom operators/providers.NET reflection, dynamic code generation, … –System side Dryad generality: DAG model, runtime callback Clean separation of Dryad and DryadLINQ

50. Future Directions Goal: Use a cluster as if it is a single computer –Dryad/DryadLINQ represent a modest step On-going research –What can we write with DryadLINQ? Where and how to generalize the programming model? –Performance, usability, etc. How to debug/profile/analyze DryadLINQ apps? –Job scheduling How to schedule/execute N concurrent jobs? –Caching and incremental computation How to reuse previously computed results? –Static program checking A very compelling case for program analysis? Better catch bugs statically than fighting them in the cloud?

51. Summary General-purpose platform for scalable distributed data-processing of all sorts Very flexible –Optimizations can get more sophisticated Designed to be used as middleware –Slot different programming models on top –LINQ is very powerful

52. Related work inside MSR Dryad, EuroSys’07. DryadLINQ, OSDI’08. Steno: automatic optimization of declarative queries, PLDI’11. MadLINQ, largescale matrix computation. EuroSys’12 Best Paper. Optimus: A Dynamic Rewriting Framework for Data-Parallel Execution Plans. EuroSys’13. More: http://research.microsoft.com/en-us/projects/dryadlinq/http://research.microsoft.com/en-us/projects/dryadlinq/