Architectures forMassive Data Management

Imagemap
Architectures for Massive Data Manageme ...extend and reuse techniques fromcharacterized bySample systems (mostly uncomparable, som ...distributed databaseskey-value storessuch asVoldemortRedisDynamoDB from AmazonOracle NoSQLMemCachedgraph storessuch asNeo4JPregelVirtuosoIBM DB2 RDF StoreOracle RDF Semantic GraphMapReduce systemsextendedByFlinkSparkrun as back-end forHive from ApachePig from ApacheHBase from Apachebased onMapReduce modelextendsBulk-synchronous parallelism (BSP) such asHadoop from ApacheHortonworksMapRP2P systemsmediator systemsdata warehouses"Big Table" systemssuch asCassandra from ApacheSpanner from GoogleF1 from GoogleBSP systemssuch asPregel from GoogleGiraph from ApacheDocument storessuch asMongoDBCouchDB from ApacheOracle NoSQLIBM PureXMLIBM DB2 JSON
hide

Architectures for
Massive Data Management

hide

extend and reuse
techniques from

hide
Distributed Systems
hide
are examples of
leaf

Distributed file
systems (DFS)

hide
are examples of
leaf

Distributed operating
systems

hidebutton_ok
Distributed Databases User Link
hide
extend techniques from
leafbutton_ok
(Centralized) databases
leaf
Stream Data Management Systems
hide
characterized by
hidebutton_ok
dimensions User Link
hide
distribution scale
leaf
centralized
leaf
moderate
leaf
large
hide
data model
leaf
relational
leaf
document (XML, JSON)
leaf
key-value
leaf
graph
hide
data heterogeneity
leaf
none
leaf
some
leaf
a lot
hide
data manipulation language
leaf
put/get
leaf
declarative query language
hide

distribution of
control and work

leaf
independent (symetric) nodes
leaf
master-slaves
leaf
set of masters, set of slaves
hide
concurrency control model
leaf
mostly relevant if updates!
leaf
atomicity
leaf
partition-tolerance
leaf
consistency
hide
replication
hide
impacts
leaf
consistency
hide
durability
leaf
none (in-memory)
leaf
some (periodic or on-demand back-up)
leaf
systematic (write-to-disk)
hide

Sample systems (mostly uncomparable,
some built on each other!)

These are just names,

check out the dimensions

characterizing each of them.

Many of these are also called "NoSQL"

leafbutton_ok
distributed databases
hide
key-value stores
hide
such as
leaf
Voldemort
leaf
Redis
leaf
DynamoDB from Amazon
leaf
Oracle NoSQL
leaf
MemCached
hide
graph stores
hide
such as
leaf
Neo4J
leaf
Pregel
leaf
Virtuoso
leaf
IBM DB2 RDF Store
leaf
Oracle RDF Semantic Graph
hidebutton_ok
MapReduce systems User Link
hide
extendedBy
leafbutton_ok
Flink User Link
leaf
Spark
hide
run as back-end for
leaf
Hive from Apache
leaf
Pig from Apache
leaf
HBase from Apache
hide
based on
hidebutton_ok
MapReduce model
hide
extends
leaf
Bulk-synchronous parallelism (BSP)
hide
such as
leafbutton_ok
Hadoop from Apache
leaf
Hortonworks
leaf
MapR
leaf
P2P systems
leaf
mediator systems
leaf
data warehouses
hide
"Big Table" systems
hide
such as
leaf
Cassandra from Apache
leaf
Spanner from Google
leaf
F1 from Google
hide

BSP systems

hide
such as
leaf
Pregel from Google
leaf
Giraph from Apache
hide
Document stores
hide
such as
leaf
MongoDB
leaf
CouchDB from Apache
leaf
Oracle NoSQL
leaf
IBM PureXML
leaf
IBM DB2 JSON