1. L/O/G/O 云端的小飞象系列报告之二 Cloud 组

2. L/O/G/O Hadoop in SIGMOD 2011 www.themegallery.com

3. Outline Introduction Nova: Continuous Pig/Hadoop Workflows Apache Hadoop Goes Realtime at Facebook Emerging Trends in the Enterprise Data Analytics A Hadoop Based Distributed Loading Approach to Parallel Data Warehouses

4. Industrial Session in Sigmod 2011 Data Management for Feeds and Streams(2) Dynamic Optimization and Unstructured Content (4) BusinessAnalytics(2) Support for Business Analytics and Warehousing (4) Applying Hadoop (4) Industrial session

6. Nova: Continuous Pig/Hadoop Workflows By Yahoo ！

7. Nova Overview  Scenarios  Ingesting and analyzing user behavior logs  Building and updating a search index from a stream of crawled web pages  Processing semi-structured data  Two-layer programming model (Nova over Pig)  Continuous processing  Independent scheduling  Cross-module optimization  Manageability features

8. Workflow Model  Workflow  Two kinds of vertices: tasks (processing steps) and channels (data containers)  Edges connect tasks to channels and channels to tasks  Four common patterns of processing  Non-incremental (template detection)  Stateless incremental (shingling)  Stateless incremental with lookup table (template tagging)  Stateful incremental (de-duping)

9. Workflow Model (Cont.)  Data and Update Model  Blocks: A channel’s data is divided into blocks Contains a complete snapshot of data on a channel as of some point in time Base blocks are assigned increasing sequence numbers(B 0, B 1, B 2…… B n ) Base block Used in conjunction with incremental processing Contains instructions for transforming a base block into a new base block( ) Delta block

10. Workflow Model (Cont.)  Task/Data Interface  Consumption mode: all or new  Production mode: B or Δ

11. Workflow Model (Cont.)  Workflow Programming and Scheduling  Data-based trigger.  Time-based trigger  Cascade trigger.  Data Compaction and Garbage Collection  If a channel has blocks B 0 ， ，, ， the compaction operation computes and adds B 3 to the channel  After compaction is used to add B 3 to the channel ， and current cursor is at sequence number 2 ， then B 0 ， ， can be garbage-collected.

12. Nova System Architecture

13. Apache Hadoop Goes Realtime at Facebook By Facebook

14. Workload Types  Facebook Messaging  High Write Throughput  Large Tables  Data Migration  Facebook Insights  Realtime Analytics  High Throughput Increments  Facebook Metrics System (ODS)  Automatic Sharding  Fast Reads of Recent Data and Table Scans

15. Why Hadoop & HBase  Elasticity  High write throughput  Efficient and low-latency strong consistency semantics within a data center  Efficient random reads from disk  High Availability and Disaster Recovery  Fault Isolation  Atomic read-modify-write primitives  Range Scans  Tolerance of network partitions within a single data center  Zero Downtime in case of individual data center failure  Active-active serving capability across different data centers

16. Realtime HDFS  High Availability - AvatarNode

17. Realtime HDFS (Cont.)  Hadoop RPC compatibility  Block Availability: Placement Policy  a pluggable block placement policy

18. Realtime HDFS (Cont.)  Performance Improvements for a Realtime Workload  RPC Timeout  Reads from Local Replicas  New Features  HDFS sync  Concurrent Readers

19. Production HBase  ACID Compliance (RWCC: Read Write Consistency Control)  Atomicity (WALEdit)  Consistency  Availability Improvements  HBase Master Rewrite ， Region assignment in memory -> ZooKeeper  Online Upgrades  Distributed Log Splitting  Performance Improvements  Compaction （ minor and major ）  Read Optimizations

20. Emerging Trends in the Enterprise Data Analytics: Connecting Hadoop and DB2 Warehouse By IBM

21. Motivation 1.Increasing volumes of data 2. Hadoop-based solutions in conjunction with data warehouses

23. A Hadoop Based Distributed Loading Approach to Parallel Data Warehouses By Teradata

24. Motivation  ETL(Extraction Transformation Loading) is a critical part of data warehouse  While data are partitioned and replicated across all nodes in a parallel data warehouse, load utilities reside on a single node(bottleneck)

25. Why Hadoop for Teradata EDW （ Enterprise Data Warehouse ） ?  More disk space can be easily added  Use as a intermediate storage  MapReduce for transformation  Load data in parallel

26. Block Assignment Problem –HDFS file F on a cluster of P nodes (each node is uniquely identified with an integer i where 1 ≤ i ≤ P) – The problem is defined by: assignment(X, Y, n,m, k, r)  X is the set of n blocks (X = {1,..., n}) of F  Y is the set of m nodes running PDBMS (called PDBMS nodes) (Y ⊆ {1,..., P })  k copies, m nodes  r is the mapping recording the replicated block locations of each block. r(i) returns the set of nodes which has a copy of the block i.

27. Block Assignment Problem （ Cont. ） An assignment g from the blocks in X to the nodes in Y is denoted by a mapping from X = {1,..., n} to Y where g(i) = j (i ∈ X, j ∈ Y ) means that the block i is assigned to the node j. An even assignment g is an assignment such that ∀ i ∈ Y ∀ j ∈ Y | |{ x | ∀ 1 ≤ x ≤ n&&g(x) = i}| - |{y | ∀ 1 ≤ y ≤ n&&g(y) = j}| | ≤ 1. The cost of an assignment g is defined to be cost(g) = |{i | g(i) r(i) ∀ 1 ≤ i ≤ n}|, which is the number of blocks assigned to remote nodes.

28. L/O/G/O Thank You!