Gowtham Rajappan

HDFS – Hadoop Distributed File System modeled on Google GFS.

Hadoop MapReduce – Similar to Google MapReduce Hbase – Similar to Google Bigtable

Master: hadoop01.cselabs.umn.edu Slaves: hadoop02 – hadoop05.cselabs.umn.edu You will require cselabs account to access this cluster. You

can login to any of these machines from any cs/cselabs machine.

Data is divided into various tables Table is composed of columns, columns are grouped into

column-families

Partitioning A table is horizontally partitioned into regions, each

region is composed of sequential range of keys Each region is managed by a RegionServer, a single

RegionServer may hold multiple regions Persistence and data availability

HBase stores its data in HDFS, it doesn't replicate RegionServers and relies on HDFS replication for data availability.

Region data is cached in-memory Updates and reads are served from in-memory cache

(MemStore) MemStore is flushed periodically to HDFS Write Ahead Log (stored in HDFS) is used for

durability of updates

HBase shell provides interactive commands for manipulating database

Create/delete tables Insert/update/read from tables Manage regions

Hbase provides single row atomic operations CheckAndPut – Similar to test-and-set CheckAndDelete All row operations are atomic no matter how many

columns are involved.

Hbase also provides row level exclusive locks You can use these locks to implement single row level

transactions

HBase stores multiple versions of a column in a row. Each version is identified by a integer timestamp

By default system time is used as version timestamps. However user can specify a logical timestamp for versioning

Each update to a row creates a new version, for the specified column.

A version can be accessed or deleted using its timestamp. HBase allows to obtain list of all the versions.

Hadoop Home - http://hadoop.apache.org/ Hbase - http://hbase.apache.org/ API

http://hbase.apache.org/apidocs/ http://hadoop.apache.org/