M7 and Apache Drill, Micheal Hausenblas


Published on

Published in: Technology
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • http://solr-vs-elasticsearch.com/
  • (This is a ASR-35 at DEC mainframe–other console terminals used were Teletype model 35 Teletypes)Allowing the user to issue ad-hoc queries is essential: often, the user might not necessarily know ahead of time what queries to issue. Also, one may need to react to changing circumstances. The lack of tools to perform interactive ad-hoc analysis at scale is a gap that Apache Drill fills.
  • Hive: compile to MR, Aster: external tables in MPP, Oracle/MySQL: export MR results to RDBMSDrill, Impala, CitusDB: real-time
  • Suppose a marketing analyst trying to experiment with ways to do targeting of user segments for next campaign. Needs access to web logs stored in Hadoop, and also needs to access user profiles stored in MongoDB as well as access to transaction data stored in a conventional database.
  • Geo-spatial + time series data with highly discriminative queries (timeframe, region, etc.)
  • Re ad-hoc:You might not know ahead of time what queries you will want to make. You may need to react to changing circumstances.
  • Two innovations: handle nested-data column style (column-striped representation) and multi-level execution trees
  • repetition levels (r) — at what repeated field in the field’s path the value has repeated.definition levels (d) — how many fields in path thatcould be undefined (because they are optional or repeated) are actually presentOnly repeated fields increment the repetition level, only non-required fields increment the definition level. Required fields are always defined and do not need a definition level. Non repeated fields do not need a repetition level.An optional field requires one extra bit to store zero if it is NULL and one if it is defined. NULL values do not need to be stored as the definition level captures this information.
  • Source query - Human (eg DSL) or tool written(eg SQL/ANSI compliant) query Source query is parsed and transformed to produce the logical planLogical plan: dataflow of what should logically be doneTypically, the logical plan lives in memory in the form of Java objects, but also has a textual formThe logical query is then transformed and optimized into the physical plan.Optimizer introduces of parallel computation, taking topology into accountOptimizer handles columnar data to improve processing speedThe physical plan represents the actual structure of computation as it is done by the systemHow physical and exchange operators should be appliedAssignment to particular nodes and cores + actual query execution per node
  • Drillbits per node, maximize data localityCo-ordination, query planning, optimization, scheduling, execution are distributedBy default, Drillbits hold all roles, modules can optionally be disabled.Any node/Drillbit can act as endpoint for particular query.
  • Zookeeper maintains ephemeral cluster membership information onlySmall distributed cache utilizing embedded Hazelcast maintains information about individual queue depth, cached query plans, metadata, locality information, etc.
  • Originating Drillbit acts as foreman, manages all execution for their particular query, scheduling based on priority, queue depth and locality information.Drillbit data communication is streaming and avoids any serialization/deserialization
  • Red: originating drillbit, is the root of the multi-level execution tree, per query/jobLeafs use their storage engine interface to scan respective data source (DB, file, etc.)
  • Relation of Drill to HadoopHadoop = HDFS + MapReduceDrill for:Finding particular records with specified conditions. For example, to findrequest logs with specified account ID.Quick aggregation of statistics with dynamically-changing conditions. For example, getting a summary of request traffic volume from the previous night for a web application and draw a graph from it.Trial-and-error data analysis. For example, identifying the cause of trouble and aggregating values by various conditions, including by hour, day and etc...MapReduce: Executing a complex data mining on Big Data which requires multiple iterations and paths of data processing with programmed algorithms.Executing large join operations across huge datasets.Exporting large amount of data after processing.
  • Designed to be as easy as possible for language implementers to utilizeDon’t constrain ourselves to SQL specific paradigm – support complex data type operators such as collapse and expand as wellAllow late typing
  • Insert points of parallelization where optimizer thinks they are necessaryPick the right version of each operatorApply projection and other push-down rules into capable operators
  • M7 and Apache Drill, Micheal Hausenblas

    1. 1. The MapR Big Data platformM7 and Apache DrillMichael Hausenblas, Chief Data Engineer EMEA, MapRHUG France, Paris, 2013-06-05
    2. 2. Whichworkloads doyouencounter inyourenvironment?http://www.flickr.com/photos/kevinomara/2866648330/licensedunderCCBY-NC-ND2.0
    3. 3. Batch processing… for recurring tasks such as large-scale data mining, ETLoffloading/data-warehousing  for the batch layer in Lambdaarchitecture
    4. 4. OLTP… user-facing eCommerce transactions, real-time messaging atscale (FB), time-series processing, etc.  for the serving layer inLambda architecture
    5. 5. Stream processing… in order to handle stream sources such as social media feedsor sensor data (mobile phones, RFID, weather stations, etc.) for the speed layer in Lambda architecture
    6. 6. Search/Information Retrieval… retrieval of items from unstructured documents (plaintext, etc.), semi-structured data formats (JSON, etc.), aswell as data stores (MongoDB, CouchDB, etc.)
    7. 7. MapR Big Data platform
    9. 9. HBase Adoption Used by 45% of Hadoop users Database Operations:Large scale key-value storeBlob storeLightweight OLTP Real-time Analytics:Logistics Shopping cartBilling Auction engineLog analysis
    10. 10. HBase IssuesReliability•Compactions disrupt operations•Very slow crash recovery•Unreliable splittingBusiness continuity•Common hardware/software issues cause downtime•Administration requires downtime•No point-in-time recovery•Complex backup processPerformance•Many bottlenecks result in low throughput•Limited data locality•Limited # of tablesManageability•Compactions, splits and merges must be done manually (in reality)•Basic operations like backup or table rename are complex
    11. 11. M7: Enterprise-grade HBaseEASY DEPENDABLE FASTM7 Enterprise GradeNo RegionServers NoCompactionsConsistent Low LatencySnapshotsMirroringInstant RecoveryNo Manual Processes
    12. 12. Unified Namespace
    13. 13. Unified Namespace$ pwd/mapr/default/user/dave$ lsfile1 file2 table1 table2$ hbase shellhbase(main):003:0> create /user/dave/table3, cf1, cf2, cf30 row(s) in 0.1570 seconds$ lsfile1 file2 table1 table2 table3$ hadoop fs -ls /user/daveFound 5 items-rw-r--r-- 3 mapr mapr 16 2012-09-28 08:34 /user/dave/file1-rw-r--r-- 3 mapr mapr 22 2012-09-28 08:34 /user/dave/file2trwxr-xr-x 3 mapr mapr 2 2012-09-28 08:32 /user/dave/table1trwxr-xr-x 3 mapr mapr 2 2012-09-28 08:33 /user/dave/table2trwxr-xr-x 3 mapr mapr 2 2012-09-28 08:38 /user/dave/table3
    14. 14. Fewer LayersMapR%M7
    15. 15. Eliminating CompactionsHBase-style LevelDB-style M7Examples BigTable, HBase,Cassandra, RiakCassandra, RiakWAF Low High LowRAF High Low LowI/O storms Yes No NoDisk space overhead High (2x) Low LowSkewed data handling Bad Good GoodRewrite large values Yes Yes NoWrite-amplification factor (WAF): The ratio between writes to disk and application writes. Note that data mustbe rewritten in every indexed structure.Read-amplification factor (RAF): The ratio between reads from disk and application reads.Skewed data handling: When inserting values with similar keys (eg, increasing keys, trending topic), do othervalues also need to be rewritten?
    16. 16. Portability (In Both Ways)• HBase applications work as is with M7– No need to recompile– No vendor lock-in• You can also run Apache HBase on an M7 cluster– Recommended during a migration– Table names with a slash (/) are in M7, table names without a slashare in Apache HBase (this can be overridden to allow table-by-tablemigration)• Use standard CopyTable tool to copy a table from HBase to M7 andvice versa– hbase org.apache.hadoop.hbase.mapreduce.CopyTable --new.name=/user/tshiran/mytable mytable
    17. 17. MapR Big Data platform
    18. 18. Apache Drillinteractive, ad-hoc query at scale
    19. 19. http://www.flickr.com/photos/9479603@N02/4144121838/ licensed under CC BY-NC-ND 2.0How to dointeractivead-hoc queryat scale?
    20. 20. ImpalaInteractive Query (?)low-latency
    21. 21. Use Case: Marketing Campaign• Jane, a marketing analyst• Determine target segments• Data from different sources
    22. 22. Use Case: Logistics• Supplier tracking and performance• Queries– Shipments from supplier ‘ACM’ in last 24h– Shipments in region ‘US’ not from ‘ACM’SUPPLIER_ID NAME REGIONACM ACME Corp USGAL GotALot Inc USBAP Bits and Pieces Ltd EuropeZUP Zu Pli Asia{"shipment": 100123,"supplier": "ACM",“timestamp": "2013-02-01","description": ”first delivery today”},{"shipment": 100124,"supplier": "BAP","timestamp": "2013-02-02","description": "hope you enjoy it”}…
    23. 23. Use Case: Crime Detection• Online purchases• Fraud, bilking, etc.• Batch-generated overview• Modes– Explorative– Alerts
    24. 24. Requirements• Support for different data sources• Support for different query interfaces• Low-latency/real-time• Ad-hoc queries• Scalable, reliable
    25. 25. And now for something completely different …
    26. 26. Google’s Dremelhttp://research.google.com/pubs/pub36632.htmlSergey Melnik, Andrey Gubarev, Jing Jing Long, Geoffrey Romer, Shiva Shivakumar, Matt Tolton,Theo Vassilakis, Proc. of the 36th Intl Conf on Very Large Data Bases (2010), pp. 330-339Dremel is a scalable, interactive ad-hocquery system for analysis of read-onlynested data. By combining multi-levelexecution trees and columnar data layout,it is capable of running aggregationqueries over trillion-row tables inseconds. The system scales to thousands ofCPUs and petabytes of data, and hasthousands of users at Google.…““Dremel is a scalable, interactive ad-hocquery system for analysis of read-onlynested data. By combining multi-levelexecution trees and columnar data layout,it is capable of running aggregationqueries over trillion-row tables inseconds. The system scales to thousands ofCPUs and petabytes of data, and hasthousands of users at Google.…
    27. 27. Google’s Dremelmulti-level execution treescolumnar data layout
    28. 28. Google’s Dremelnested data + schema column-striped representationmap nested data to tables
    29. 29. Google’s Dremelexperiments:datasets & query performance
    30. 30. Back to Apache Drill …
    31. 31. Apache Drill–key facts• Inspired by Google’s Dremel• Standard SQL 2003 support• Plug-able data sources• Nested data is a first-class citizen• Schema is optional• Community driven, open, 100’s involved
    32. 32. High-level Architecture
    33. 33. Principled Query Execution• Source query—what we want to do (analystfriendly)• Logical Plan— what we want to do (languageagnostic, computer friendly)• Physical Plan—how we want to do it (the bestway we can tell)• Execution Plan—where we want to do it
    34. 34. Principled Query ExecutionSourceQuery ParserLogicalPlan OptimizerPhysicalPlan ExecutionSQL 2003DrQLMongoQLDSLscanner APITopologyCFetc.query: [{@id: "log",op: "sequence",do: [{op: "scan",source: “logs”},{op: "filter",condition:"x > 3”},parser API
    35. 35. Wire-level Architecture• Each node: Drillbit - maximize data locality• Co-ordination, query planning, execution, etc, are distributedStorageProcessDrillbitnodeStorageProcessDrillbitnodeStorageProcessDrillbitnodeStorageProcessDrillbitnode
    36. 36. Wire-level Architecture• Curator/Zookeeper for ephemeral cluster membership info• Distributed cache (Hazelcast) for metadata, localityinformation, etc.Curator/ZkDistributed CacheStorageProcessDrillbitnodeStorageProcessDrillbitnodeStorageProcessDrillbitnodeStorageProcessDrillbitnodeDistributed Cache Distributed Cache Distributed Cache
    37. 37. Wire-level Architecture• Originating Drillbit acts as foreman: manages query execution,scheduling, locality information, etc.• Streaming data communication avoiding SerDeCurator/ZkDistributed CacheStorageProcessDrillbitnodeStorageProcessDrillbitnodeStorageProcessDrillbitnodeStorageProcessDrillbitnodeDistributed Cache Distributed Cache Distributed Cache
    38. 38. Wire-level ArchitectureForeman turns intoroot of the multi-levelexecution tree, leafsactivate their storageengine interface.nodenode nodeCurator/Zk
    39. 39. On the shoulders of giants …• Jackson for JSON SerDe for metadata• Typesafe HOCON for configuration and module management• Netty4 as core RPC engine, protobuf for communication• Vanilla Java, Larray and Netty ByteBuf for off-heap large data structures• Hazelcast for distributed cache• Netflix Curator on top of Zookeeper for service registry• Optiq for SQL parsing and cost optimization• Parquet (http://parquet.io) as native columnar format• Janino for expression compilation• ASM for ByteCode manipulation• Yammer Metrics for metrics• Guava extensively• Carrot HPC for primitive collections
    40. 40. Key features• Full SQL – ANSI SQL 2003• Nested Data as first class citizen• Optional Schema• Extensibility Points …
    41. 41. Extensibility Points• Source query  parser API• Custom operators, UDF  logical plan• Serving tree, CF, topology  physical plan/optimizer• Data sources &formats  scanner APISourceQuery ParserLogicalPlan OptimizerPhysicalPlan Execution
    42. 42. … and Hadoop?• How is it different to Hive, Cascading, etc.?• Complementary use cases*• … use Apache Drill– Find record with specified condition– Aggregation under dynamic conditions• … use MapReduce– Data mining with multiple iterations– ETL*)https://cloud.google.com/files/BigQueryTechnicalWP.pdf
    44. 44. Basic Demohttps://cwiki.apache.org/confluence/display/DRILL/Demo+HowTo{"id": "0001","type": "donut",”ppu": 0.55,"batters":{"batter”:[{ "id": "1001", "type": "Regular" },{ "id": "1002", "type": "Chocolate" },…data source: donuts.jsonquery:[ {op:"sequence",do:[{op: "scan",ref: "donuts",source: "local-logs",selection: {data: "activity"}},{op: "filter",expr: "donuts.ppu < 2.00"},…logical plan: simple_plan.jsonresult: out.json{"sales" : 700.0,"typeCount" : 1,"quantity" : 700,"ppu" : 1.0}{"sales" : 109.71,"typeCount" : 2,"quantity" : 159,"ppu" : 0.69}{"sales" : 184.25,"typeCount" : 2,"quantity" : 335,"ppu" : 0.55}
    45. 45. SELECTt.cf1.name as name,SUM(t.cf1.sales) as total_salesFROM m7://cluster1/sales tGROUP BY nameORDER BY by total_sales desc
    46. 46. sequence: [{ op: scan, storageengine: m7,selection: {table: sales}}{ op: project, projections: [{ref: name, expr: cf1.name},{ref: sales, expr: cf1.sales}]}{ op: segment, ref: by_name, exprs: [name]}{ op: collapsingaggregate, target: by_name,carryovers: [name],aggregations: [{ref: total_sales, expr:sum(name)}]}{ op: order, ordering: [{order: desc, expr:total_sales}]}{ op: store, storageengine: screen}]
    47. 47. {@id: 1, pop: m7scan, cluster: def,table: sales, cols: [cf1.name, cf2.name]}{@id: 2, op: hash-random-exchange,input: 1, expr: 1}{@id: 3, op: sorting-hash-aggregate, input: 2,grouping: 1, aggr:[sum(2)], carry: [1], sort:~agrr[0]}{@id: 4, op: screen, input: 4}
    48. 48. Execution Plan• Break physical plan into fragments• Determine quantity of parallelization for eachtask based on estimated costs• Assign particular nodes based on affinity, loadand topology
    49. 49. BE A PART OF IT!
    50. 50. Status• Heavy development by multiple organizations• Available– Logical plan (ADSP)– Reference interpreter– Basic SQL parser– Basic demo
    51. 51. StatusJune 2013• Full SQL support (+JDBC)• Physical plan• In-memory compressed data interfaces• Distributed execution
    52. 52. StatusJune 2013• HBase and MySQL storage engine• User Interfaces
    53. 53. User Interfaces
    54. 54. User Interfaces• API—DrillClient– Encapsulates endpoint discovery– Supports logical and physical plan submission,query cancellation, query status– Supports streaming return results• JDBC driver, converting JDBC into DrillClientcommunication.• REST proxy for DrillClient
    55. 55. User Interfaces
    56. 56. ContributingContributions appreciated (not only code drops) …• Test data & test queries• Use case scenarios (textual/SQL queries)• Documentation• Further schedule– Alpha Q2– Beta Q3
    57. 57. Kudos to …• Julian Hyde, Pentaho• Lisen Mu, XingCloud• Tim Chen, Microsoft• Chris Merrick, RJMetrics• David Alves, UT Austin• Sree Vaadi, SSS• Jacques Nadeau, MapR• Ted Dunning, MapR
    58. 58. Engage!• Follow @ApacheDrill on Twitter• Sign up at mailing lists (user | dev)http://incubator.apache.org/drill/mailing-lists.html• Standing G+ hangouts every Tuesday at 18:00 CEThttp://j.mp/apache-drill-hangouts• Keep an eye on http://drill-user.org/