Brisk: more powerful Hadoop powered by Cassandra

Brisk: More Powerful Hadoop Powered by [email protected]

Monday, July 25, 2011

mailto:[email protected]

mailto:[email protected]

The evolution of Analytics

Analytics + Realtime



Analytics Realtime

replication



ETL


Brisk re-unifies realtime and analytics


The Traditional Hadoop Stack

Master Nodes

Name Node

Secondary Name Node

Job Tracker

ZooKeeper

MetaStore

Slave Nodes

Data Node

Task Tracker

Region Server

Hbase MasterPig

Hive

Region Server

Client Nodes


7


Brisk Architecture


Brisk Highlights

✤ Easy to deploy and operate✤ No single points of failure✤ Scale and change nodes with no downtime✤ Cross-DC, multi-master clusters✤ Allocate resources for OLAP vs OLTP

✤ With no ETL


Cassandra data model

✤ ColumnFamilies contain rows + columns✤ (Not really schemaless for a while now)

password name site* Nate McCall* Brandon Williams

* Jonathan Ellis datastax.com

zznatedriftxjbellis


Sparse

password name* Nate McCall

zznate

password name* Brandon Williams

driftx

password name site* Jonathan Ellis datastax.com

jbellis


Rows as containers / materialized views

driftx thobbs pcmanus jbellis zznatecircle1

xedin mdenniscircle2

xedin pcmanus ymorishitacircle3



CassandraFS

✤ data stored as ByteBuffer internally -- excellent fit for blocks✤ local reads mmap data directly (no rpc)✤ blocks are compressed with google snappy✤ hadoop distcp hdfs:///mydata cfs:///mydata


Hive support

✤ Hive MetaStore in Cassandra✤ Unified schema view from any node, with no external systems

and no SPOF✤ Automatically maps Cassandra column families to Hive tables

✤ Supports static and dynamic column families (and supercolumns)


Hive: CFS and ColumnFamilies

CREATE TABLE users (name STRING, zip INT);

LOAD DATA LOCAL INPATH 'kv2.txt' OVERWRITE INTO TABLE users;

CREATE EXTERNAL TABLE Keyspace1.Users(name STRING, zip INT)

STORED BY 'org.apache.hadoop.hive.cassandra.CassandraStorageHandler';

CREATE EXTERNAL TABLE Keyspace1.Users

(row_key STRING, column_name STRING, value string)

STORED BY 'org.apache.hadoop.hive.cassandra.CassandraStorageHandler';


Pig Support

✤ With standard Cassandra:$ export PIG_HOME=/path/to/pig

$ export PIG_INITIAL_ADDRESS=localhost

$ export PIG_RPC_PORT=9160

$ export PIG_PARTITIONER=org.apache.cassandra.dht.RandomPartitioner

$ contrib/pig/bin/pig_cassandra

grunt>

✤ With Brisk:$ bin/brisk pig

grunt>


Pig: CFS and ColumnFamilies

grunt> data = LOAD 'cfs:///example.txt' using PigStorage() as (name:chararray, value:long);

data = LOAD 'cassandra://Demo1/Scores' using CassandraStorage() AS (key, columns: {T: tuple(name, value)});

data = LOAD 'cassandra://Demo1/Scores&slice_start=M&slice_end=S' using CassandraStorage() AS (key, columns: {T: tuple(name, value)});


19


Data model: Realtime

GOOG LNKD P AMZN AAPLE80 20 40 100 20

Portfolio1

Portfolios

2011-01-01 2011-01-02 2011-01-03$79.85 $75.23 $82.11

GOOG

StockHist

last$95.52

$186.10

$112.98

GOOG

LiveStocks

AAPLAMZN


Data model: Analytics

worst_date loss2011-07-23 -$34.812011-03-11 -$11432.242011-05-21 -$1476.93

Portfolio1

HistLoss

Portfolio2Portfolio3



ticker rdate returnGOOG 2011-07-25 $8.23GOOG 2011-07-24 $6.14GOOG 2011-07-23 $7.78AAPL 2011-07-25 $15.32AAPL 2011-07-24 $12.68

10dayreturns

INSERT OVERWRITE TABLE 10dayreturnsSELECT a.row_key ticker, b.column_name rdate, b.value - a.valueFROM StockHist a JOIN StockHist b ON (a.row_key = b.row_key AND date_add(a.column_name,10) = b.column_name);


2011-01-01 2011-01-02 2011-01-03$79.85 $75.23 $82.11

GOOG

row_key column_name valueGOOG 2011-01-01 $8.23GOOG 2011-01-02 $6.14GOOG 2011-001-03 $7.78



portfolio rdate preturnPortfolio1 2011-07-25 $118.21Portfolio1 2011-07-24 $60.78Portfolio1 2011-07-23 -$34.81Portfolio2 2011-07-25 $2143.92Portfolio3 2011-07-24 -$10.19

portfolio_returns

INSERT OVERWRITE TABLE portfolio_returnsSELECT row_key portfolio, rdate, SUM(b.return)FROM portfolios a JOIN 10dayreturns b ON (a.column_name = b.ticker)GROUP BY row_key, rdate;



INSERT OVERWRITE TABLE HistLossSELECT a.portfolio, rdate, minpFROM ( SELECT portfolio, min(preturn) as minp FROM portfolio_returns GROUP BY portfolio) a JOIN portfolio_returns b ON (a.portfolio = b.portfolio and a.minp = b.preturn);

worst_date loss2011-07-23 -$34.812011-03-11 -$11432.242011-05-21 -$1476.93

Portfolio1

HistLoss

Portfolio2Portfolio3


Portfolio Demo dataflow

Portfolios

Historical Prices

Intermediate Results

Largest loss

Web-based Portfolios

Live Prices for today

Largest loss


OpsCenter



Where to get it

✤ http://www.datastax.com/brisk


http://www.datastax.com/brisk

http://www.datastax.com/brisk


Technology

Brisk: more powerful Hadoop powered by Cassandra