SDN Controller Programming Guide

Persistence

Distributed Persistence Overview

The SDN Controller provides a distributed persistence for applications in form of a Cassandra [11]

database node running on each controller instance. A team of controllers serves as a Cassandra

cluster. Cassandra provides the following benefit as a distributed database:

 A distributed, peer-to-peer datastore with no single point of failure.

 Automatic replication of data for improved reliability and availability.

 An eventually-consistent view of the database from any node in the cluster.

 Incremental, scale-out growth model.

 Flexible schemas (column oriented keyspaces).

 Hadoop integration for large-scale data processing.

 SQL-like query support via Cassandra Query Language (CQL).

Distributed Persistence Use Case

The distributed persistence architecture is targeted at applications that have distributed active-

active requirements. Specifically, applications should use the distributed persistence framework if

they have one or more of following requirements:

 Consumer application have high scalability requirements i.e. there are generally multiple

instances of the app running on different controller nodes that need access to a common

distributed database store.

 The distributed database should be available independent of whether individual nodes are

present or not e.g if there are controller node crashes.

 The applications have high throughput requirements: large number of I/O operations.

Further, they have requirements wherein as the number of controller nodes increases,

performance needs to scale linearly.

For addressing applications with such requirements, a distributed persistence layer that uses

Cassandra is exported as the underlying distributed database. The HP VAN SDN Controller

provides a Data Access Object (DAO) layer on top of Cassandra for performing distributed

persistence operations.

Persistence Data Model

Introduction to DAO Pattern

A data access object (DAO) is an object that provides an abstract interface to some type of

database or persistence mechanism, providing some specific operations without exposing details

of the database. It provides a mapping from application calls to the persistence layer. This

isolation separates the concerns of what data accesses the application needs, in terms of domain-

specific objects and data types (the public interface of the DAO), and how these needs can be

satisfied with a specific DBMS, database schema, and so on. Figure 33 and Figure 34 show Data

Access Object pattern [31].