Cluster Computing
What is a Cluster?
A group of homogeneous computers(nodes) integrated to act as a single computer is called a Cluster. Cluster is a tightly coupled system of computers (commodity or proprietary). Cluster has the following characteristics.
- Cluster computing is called High Performance Computing: Primarily concerned with computational resources. (HPC)
- Mostly uses open source software
- Single System Image (SSI)
- Centralized scheduling, administering and job management
- Highly available (if its setup for high availability) and fault tolerant
- limited number of nodes (resources)
- All the nodes in the cluster are homogeneous
- Physical location of a cluster is static
Cluster setup
You can setup a cluster in two ways:
-
- Gateway machine/ Head Node (2 Network cards) connected to the internet with a Public IP and internally connected to the nodes via Private IP.
- Nodes are accessible only via the head node and they are not visible to the outside world.
- Firewall between the Internet and the Gateway Machine
- Jobs are submitted to the headnode, Job Managements system on the head node distributed the job to the underlying cluster nodes.
Click to enlarge the picture
-
- There is no Head Node or a Gateway machine
- All the nodes are directly connected to the Internet (public IP)
-
- Firewall between the Internet and nodes
Click to enlarge the picture
