Architecting Immediacy: The Design of a High-Performance Portable Wrangling Engine

Talk at Strata San Jose, 2016, joint with Seshadri Mahalingam

Transcript

1. Architecting Immediacy Joe Hellerstein and Seshadri Mahalingam The Design of a High-Performance Portable Wrangling Engine PHOTON
2. Wrangling and Immediacy Technical Challenges Architectural Context JavaScript and its Discontents Photon: A Clean Slate Design Outline 1 2 3 4 5
3. Turn of the Century Data Transformation - Schemas and annotations - Box-and-arrow programming - Batch execution
4. Research Roots: Potter’s Wheel 2001 + Real data, sampled on the fly + Menu-driven transforms + Immediate execution and feedback [Raman & Hellerstein, VLDB01]
5. Research Roots: Open Source Data Wrangler, 2011 – Browser-sized data sets + Predictive Transformation + Immediate feedback on
 multiple choices [Kandel, Heer & Hellerstein, CHI 11]
6. 2016: Data Wrangling and Immediacy DEMO Grown from research roots. + Deeper and broader Predictive 
 Transformation + Sample-to-Scale with Intelligent 
 Execution + Interactive Visual Profiling Multi-faceted immediacy!
7. The Value of Immediacy Immediate, step-by-step feedback during data transformation Understand options Confirm intent Assess
 progress
8. The Value of Immediacy Miller ’68, 
 Card ’91, 
 Nielsen ‘93 Elapsed Time (secs) 1 2 3 4 5 6 7 8 9 10 0 User attention User flow “Instantaneous” Batch processing is in conflict with these goals!
9. Without immediacy, what’s left? suggestion previews suggestion previews SUGGESTION PREVIEWS
10. Without immediacy, what’s left? suggestion previews suggestion previews SUGGESTION PREVIEWS
11. Without immediacy, what’s left? COLUMN PROFILES
12. Without immediacy, what’s left? COLUMN PROFILES
13. Without immediacy, what’s left? DATA QUALITY BARS
14. Without immediacy, what’s left? DATA QUALITY BARS
15. Without immediacy, what’s left? SUMMARY STATS
16. Without immediacy, what’s left? SUMMARY STATS
17. Without immediacy, what’s left? DATA SAMPLES
18. Without immediacy, what’s left? Programming. DATA SAMPLES
19. Without immediacy, what’s left? Programming. SCHEMA TRANSFORM SPEC
20. Without immediacy, what’s left? Programming. TRANSFORM SPEC GO! SCHEMA
21. Wrangling and Immediacy Technical Challenges Architectural Context JavaScript and its Discontents Photon: A Clean Slate Design Outline 1 2 3 4 5
22. Three Technical Challenges Performance Scale The Dirty Details
 of Dirty Data
23. Technical Challenges Performance: Stay in the Flow Miller ’68, 
 Card ’91, 
 Nielsen ‘93 Elapsed Time (secs) 1 2 3 4 5 6 7 8 9 10 0 User attention User flow Instantaneous
24. Technical Challenges Scale Scaling the Unknown P NO PRECOMPUTING NO SCHEMA
 NO INDEXES NO SUMMARIES
25. Technical Challenges The Dirty Details of Dirty Data Expect the Unexpected
26. The Dirty Details of Dirty Data 1 2 3 4 5 Technical Challenges Ambiguous Schemas 6 Heavy String Processing Ambiguous Types Noise & Exceptions Limited Filtering Rich Transforms
27. Ambiguous Schemas Technical Challenges
28. Heavy String Processing Technical Challenges
29. Ambiguous Types Technical Challenges
30. Noise & Exceptions Technical Challenges
31. Technical Challenges Limited Filtering
32. Rich Transforms Technical Challenges
33. This is the New World of Data Wrangling Stay in 
 the Flow Scale the Unknown Expect the Unexpected
34. Wrangling and Immediacy Technical Challenges Architectural Context JavaScript and its Discontents Photon: A Clean Slate Design Outline 1 2 3 4 5
35. Architecture: UX, DSL Compiler and Runtime DSL COMPILER
36. Scaling Up Via New Compiler Targets Server User-interface DSL COMPILER
37. Sample-to-Scale Server User-interface DSL COMPILER 1 Extract Sample
38. Sample-to-Scale Server User-interface DSL COMPILER 2 Wrangle Sample
39. Sample-to-Scale Server User-interface DSL COMPILER 3 Execute final
 wrangling at scale
40. Sample-to-Scale Server User-interface DSL COMPILER 4 Return results, profiles, interesting rows
41. Architecture: Multiple Runtimes, Intelligent Execution Server User-interface DSL Compiler Intelligent Execution
42. Architecture: Spectrum of Execution Server User-interface DSL Compiler Intelligent Execution
43. Flexible Deployments Trifacta Wrangler App On Premises DSL Compiler Intelligent Execution Hosted DSL Compiler Intelligent Execution AWS DSL Compiler Intelligent Execution
44. Wrangling and Immediacy Technical Challenges Architectural Context JavaScript and its Discontents Photon: A Clean Slate Design Outline 1 2 3 4 5
45. Why a JS Wrangling Engine? Stanford Wrangler team was deep in JS JS runs in the browser Dynamically typed JS is the way people-centric data apps are built today! Easy to prototype & add new features Continued to attract top JS talent …and on the desktop too JS deals well with ambiguous types and structures found while in wrangling
46. Why a JS Wrangling Engine? Stanford Wrangler team was deep in JS JS runs in the browser Dynamically typed JS is the way people-centric data apps are built today! Easy to prototype & add new features Continued to attract top JS talent …and on the desktop too JS deals well with ambiguous types and structures found while in wrangling But does it perform? Does it scale?
47. Jeff Heer’s Datavore Prototype Datavore can complete queries over million-element data tables at interactive (sub-100ms) rates. in-memory column-oriented database
48. Hard to extend, performantly You will have to modify the guts of the engine to add new aggregate operators add new logic to the inner loop of the query processor (for both dense and sparse queries)
49. Type ambiguity Inconsistencies with strongly typed execution engines Core engine slows down switch (typeof inputValue) { case 'string': return inputValue; case 'number': return inputValue + '';
50. Code Generation is deceptively easy Function bodies can be analyzed: Function.prototype.toString() Functions can be generated: new Function(functionBody) Inlining function calls brought ~15-20% speed up Difficult to maintain & debug. Modest wins don’t justify the maintenance cost.
51. JavaScript runtime: fast, but not easily tamed Difficult to exploit parallelism via multi-core processors Garbage-collected language pain points: Value copies, object allocation and garbage collection 
 Small inputs easily blow up memory usage
52. Escape from JS: A Study of Alternatives DSL COMPILER INTELLIGENT
 EXECUTION Interaction via roundtrip to a server User-interface Server
53. A Clean Slate Go for performance? Maybe write our own C++/LLVM- based query engine! Java? Not viable in the browser Portability a primary concern Browser + 
 Single-node Server But … portability? LLVM
54. Surprise: What can’t a browser do these days? Native
 code execution frameworks Chrome’s Portable Native Client
 (PNaCl) asm.js Upcoming: WebAssembly Run compiled
 C and C++ code LLVM No JIT overhead Dense data representations No unpredictable garbage collection
55. Portable Native Client (PNaCl) LLVM-based, cross-compilation style toolchain https://developer.chrome.com/native-client We went with Chrome’s PNaCl Plenty of ports for popular vendor libraries (webports)
56. Wrangling and Immediacy Technical Challenges Architectural Context JavaScript and its Discontents Photon: A Clean Slate Design Outline 1 2 3 4 5
57. Performance Requirements Low memory footprint % of dataset size Immediate Feedback < 1 second response time
58. Engineering Process Start by establishing “Speed of Light” Step 1 Then add features for functionality and extensibility Step 2 Compare to speed of light; compromise judiciously. Step 3 Iterate, back to step 1 Step 4
59. Inspiration & Reading ➔ Serializable & human-readable description of data flow ➔ Partitioned (sharded) data flow ➔ Strategies for complex transformations ➔ Data locality maximization Impala HyPer Tupleware Spark
60. PHOTON Architecture DSL COMPILER INTELLIGENT
 EXECUTION Router Cache Engine Execution Nodes Scalar & Aggregate Functions
61. In-memory data layout Table Metadata Row Metadata Column 0 Column 1 Column 2 Chunk 0 Chunk 1 Column Metadata Chunk 0 Chunk 1 Chunk 0 Chunk 1 Column Metadata Column Metadata
62. In-memory data layout Row Batch Chunk 0 Row Metadata Row Batch http://arrow.apache.org/ Chunk 0 Chunk 0 Chunk 1 Row Metadata Chunk 1 Chunk 1
63. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T0 T1
64. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer Row Batch 0 Row Batch 1 Row Batch 2 T0 T1
65. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer Row Batch 2 T1 Row Batch 1 T0 Row Batch 0
66. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer Row Batch 2 T1 Row Batch 1 T0 Row Batch 0
67. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer Row Batch 2 T1 Row Batch 1 T0 Row Batch 0
68. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer Row Batch 2 T1 Row Batch 1 T0 Row Batch 0
69. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer Row Batch 2 Row Batch 0 T0 T1 Row Batch 3 Row Batch 1
70. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T1 Row Batch 2 T0 Row Batch 3
71. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T1 Row Batch 2 T0 Row Batch 3
72. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T1 Row Batch 2 T0 Row Batch 3
73. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T1 Row Batch 2 T0 Row Batch 3
74. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T1 Row Batch 2 T0 Row Batch 3 Row Batch 3
75. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T0 T1 Row Batch 3
76. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T0 T1 Row Batch 3
77. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T0 T1 Row Batch 3
78. Execution Thread Pool Source Filter Map Agg Map Sink Producer Pipe Pipe Pipe Barrier Consumer T0 Row Batch 3 T1
79. Development Toolchain CMake Handles
 cross-compilation toolchain for PNaCl & Desktop Google C++ libraries Modern C++11/14 std::unique_ptr & std::shared_ptr simplify memory management Standard
 threading library New
 container classes LLVM Clang libraries ASAN Address Sanitizer TSAN Thread Sanitizer Useful for catching large classes of bugs googletest googlemock benchmark
80. Was it all worth it?
81. Experimental Setup 86 MB dataset
 ~250k rows MacBook Pro
 (mid 2014) Experimental Setup Use Cases
82. Experimental Setup Experimental Setup COLUMN PROFILES suggestion previews suggestion previews SUGGESTION PREVIEWS
83. Spark performance problematic for populating grid Spark performance problematic for populating grid
84. Memory pressure a problem Memory pressure a problem
85. Aggregation: Computing histograms
86. Memory pressure is still an issue Memory pressure is still an issue
87. Design Requirements What is your desired workload? Low memory footprint % of dataset size Immediate Feedback < 1 second response time
88. ➔ The right engine for the right job ➔ Photon is yet another payoff of (DSL + compiler + intelligent execution) ➔ Alongside Spark and Hadoop ➔ Photon specialized for UX: immediacy and scale ➔ Start by establishing speed-of-light ➔ Design to ensure it remains achievable as you grow ➔ Memory management with portability ➔ LLVM + toolchain form a powerful portability platform. ➔ Explicit memory management is critical for immediacy, even more so in-browser Lessons learned
89. Part of Trifacta’s Intelligent Execution Alongside the distributed processing of Spark and Hadoop Portability
 
 Browser, Desktop, Single-Node Server Lean usage of
 client memory Efficient and predictable. Photon performance preserves user flow No loss of context
 Immediate data wrangling Fluid UX with predictions, profiles, previews
90. Questions? @joe_hellerstein @seshness