GRID ARCHITECTHRE

There are different approaches towards designing a grid computing architecture.  Some of the design principles that Clustreon accomplishes contain the following:

1. Simplicity – The design is simple in that there are few mechanisms to the system.  In a complex system, if there are many distinct components that are needy upon each other to operate, it is more likely that it will break.  

Therefore in this case, less is better.

2. Ease of Administration – With fewer pieces to the puzzle, the system becomes easier to administer.  The users of every node are empowered to have control over various functions of the grid and therefore there is less overall administration.

3. Dynamic – There is the ability to add and remove nodes to the Grid without having to cause disruption to the system as well.

4. Cost Effective – The software of each piece is simple enough to be less expensive.  This is coupled with ease of administration in lower operating costs.

One of the architectural plan approaches which accomplish this is to add intelligence to each node and have the central control be a central database.  In this case, the central controller is just SAS tables rather than complex pieces of software.  This functions as a form of traffic police while the intelligence and execution of the SAS jobs remain on the client nodes of the grid. The central tables are efficient and managed by the nodes which determine the optimal presentation of each job.  The computing processing will therefore stay spread throughout each node within the grid.  The machine that manages the grid controlling tables is tuned for this function and is frequently a file server.  In this design, the file server functions as together a traditional file server and the main grid controller. 

SAS and Non‐SAS Nodes

You can add your machine to the node as an active SAS node which will execute SAS programs.  on the other hand, you can be a node on the grid without SAS yet still have the ability to assign SAS programs and appoint the processing of SAS to another machine.  The configuration of the two types is somewhat different.  However, they share one thing in common in that they communicate to one central server which manages the processing of SAS programs.