Architecture

RadiantOne’s Big Data Directory, HDAP, is versatile and designed to meet your identity infrastructure needs, now and well into the future. The directory can be leveraged in a variety of deployment scenarios in conjunction with RadiantOne FID or as a stand-alone LDAP server.

The Big Data Directory on a Cluster

When deployed as a cluster, HDAP is highly available and extremely scalable—in fact, you can bring new nodes online in minutes. Cluster sites require a local area network with good connectivity, running at speeds of one gigabit or better. Each site can have one or more clusters deployed.
There are two main types of deployment architectures within the cluster framework.

Single Cluster: Add Follower-Only Nodes for Better Throughput
For high availability within a cluster, a minimum of three nodes—each with a replica of the HDAP store—is required, and a load balancer directs client traffic across all cluster nodes. Of the three core nodes, there will always be a single leader node and the rest will be either follower nodes (which could take over if the leader fails) or follower-only nodes.

The status of each node is handled by a distributed configuration manager, which maintains configuration consistency for the cluster. If the leader node fails, a new one will be automatically elected to ensure the integrity of the cluster. The consistency of the directory image across all cluster nodes is addressed with block replication, and any changes to the virtualized data are done on the leader node, then replicated out to the followers.

RadiantOne Big Data Directory

In a cluster deployment, changes to the virtualized data are done on the leader node, then replicated out to the followers.

If you have the need for multiple sites—whether that means several data centers or a disaster recovery site—you can deploy multiple clusters. HDAP stores deployed across more than one site can participate in inter-cluster logical replication.

Multi-Cluster

HDAP stores deployed across more than one site can participate in inter-cluster logical replication, so that the data is always up to date.

This deployment scenario supports multi-master replication based on the same model that traditional LDAP directories use today, where the current leader node in each cluster is responsible for accepting changes from other clusters. These changes will then be pushed out to other nodes within the cluster through block replication, as shown in the diagram above.

Big Data Directory Deployed on One Node as a Stand-Alone LDAP Server

Big Data Directory can also be deployed in the smallest environments as a single-node, classical LDAP server. In this case, HDAP offers the same quick performance profile as classical LDAP for information lookup (reads), while delivering up to ten times faster throughput for writes. When used in this way, it can add flexibility and scalability to many deployments, including Web Access Management, federation and more. It’s an excellent solution for companies that want to consolidate and migrate away from aging Sun/Oracle/Novell Directory architectures.

Resources:

  • Download our HDAP, RadiantOne Big Data Directory, overview paper.