 |
November 2006 Archives
Shoal Dynamic Clustering
Posted by hamada on November 09, 2006 at 11:57 AM | Permalink
| Comments (2)
The Shoal Framework provides a
cluster communication and event mechanisms. Shoal's core service
(GMS), enables an application to dynamically become a member of a
predefined cluster, and as such, is subscribed to cluster events, such
as :
- Member
join, planned shutdown, failures
- Recovery member selection
- Automated delegated recovery initiation
A cluster member is also able to broadcast messages to an individual
member or all members of the cluster(see Messaging).
In addition to messaging, members can share data using GMS's Shared
Cache (see DistributedStateCache).
The main goal for the shoal project to provide a clustering framework
for the GlassFish project, however it does not preclude it from being
used for purposes, such as the foundation for a grid deployment, or a
deployment of a highly available service, etc (I would be interested in
hearing about other use cases).
Shoal/GMS utilizes the JxtaManagement component (a JXTA based group service provider) for
dynamic cluster configuration, formation, and monitoring. This
component is broken out as follows :
- NetworkManager.
Given a instance and group name, NetworkManager uses a SHA-1 hash to
encode the cluster GroupID, and NodeID, in addition it also defines a
set of predefined communication identifiers which are used for
formation, monitoring and messaging. In addition, an application
may pass additional configuration parameters, such as bootstrapping
addresses to facilitate cross sub-net and firewall communication.
- SystemAdvertisement.
An extensible XML document describing system characteristics (HW/SW
configuration. CPU load would be a nice extension). This information is
exchanged during cluster formation, and monitoring. It is
envisioned that such information would serve at the foundation of a
Grid framework.
- MasterNode.
Is a lightweight protocol allowing a set of nodes to discover one
another, and autonomously elect a master for the cluster. The
protocol is resilient to multi node collisions and employs an
autonomous mechanism to avoid network chatter to resolve collisions.
- HealthMonitor.
Is a lightweight protocol allowing a set of nodes to monitor the
health of a cluster. The HealthMonitor relies on a tunable heart beat,
which is acted upon by the MasterNode to notify the group of failures,
and by other members to elect a new master if the master node fails.
- ClusterManger.
Manages lifecycle of this SPI
Figure 1. ClusterManager Software Stack
By using JXTA as the foundation for the ClusterManager, shoal is able
to :
- Use logical name cluster and node addressing (requires unique
naming)
- Achieve near Zero configuration for cross sub-net and firewall
connectivity
- Inherit dynamic transport selection without application
intervention (multicast vs. unicast)
- Inherit dynamic route repair
- Enables mobility, same name different physical addresses
- Automatic rerouting on/to available interfaces on multi-homed
nodes
- Inherit traffic scoping to cluster members
- Establish secure end-to-end channels
for additional blogs on the project, see blog entries by Shreedhar, Bernard, Traversat and Masood Mortazvi
|
java.net RSS Feeds
