What is Federation in the cloud control environment?
Federation is a key architectural component of senseTM. The overall system is able to automatically establish real-time relationships with all the auto-discoverable sense nodes in a network and domain. This auto-discovery is not dependent on the physical location and runtime infrastructure of the nodes. A sense node is a server instance that is able to manage and conduct all sense services. The node logically collects sense services with other services of the same family in other areas of the network and in the cloud. All services from the same family are visible to each other no matter where each service is deployed and running.
A member of the same service family is able to provide the same set of services, functionality, and responses to a requester. If needed, this ability will be fulfilled by using a different implementation or even a completely different service provider. After joining the federation, sense will monitor the ‘health’ status of the federation and the status of all the services by:
- monitoring each node’s perspective from all other nodes – which means no centralized controller and single-point-of-failure
- acquiring information about all protocols that services on the nodes have made available – increasing system resilience by enabling sense to instruct via intra-sense communication multiplexing at the time the communication is required via the most efficient available protocol.
A federation snapshot is always available through the sense web console. The web console displays the nodes and services running in the network and cloud. The console also indicates the performance according to the assigned SLAs. The federation is the key sense technology that provides individual and collective ‘health’ status for services, feelings, business flows and sense nodes. The sense federation enables services to compute on the same virtual space of resources and to be extended to other spaces.
sense is able to define the federation in a local network of a server farm(s) and establish a federation on other networks (clouds) and external virtual server farms. Federation via the internet is accomplished by configuring sense’s nodes in a secure way. sense then begins managing availability of these nodes and measuring network latency. sense informs all other services in the federation of the most proficient service instance (of the family) to be called upon to fulfill work. The federation is not continuously monitored by a federation daemon to prevent the creation of additional overheard. The federation is only refreshed at each ‘life sign’ of the system, when something is changing for the better or worse. For instance, an event based refresh on any SLA status change of any managed asset. Federation cache is kept in memory for highest performance. All or a portion of federation cache can be directed to a permanent XML and relational databases and/or to a data warehouse for business intelligence analysis. sense comes standard with a default persistent store for using a MySQL database. BI analysis and reporting capabilities are also out-of-the-box using an embedded implementation of the open source Pentaho platform – other Bi tools integration is easy to achieve.
A cache for the cloud
sense federation establishes a distributed caching mechanism for the cloud using a federated cloud messaging model (see example below).
This enables sharing of information and business objects throughout the cloud to provide information near the computing platforms. This replaces the practice of concentrating information in a controller server and sparse distribution on the cloud. Each sense instance shares the cache and implements logic for managing information in the cache. An example is a feeling that caches the calls of the last 100 unique users and the mock strategy resurrects this cache because something went wrong. Federating information in the cloud with this approach enables each sense node to be transparently aware of all other sense nodes and services via a local cache. This method is beneficial for performance, and also stays current of the last interesting event change.
