What exactly is Head Node, Master Node, Compute Node ?

So here let’s talk about High Performance Cluster System (HPC) ‘High performance computing, HPC systems are a set of multiple CPUs server and other hardware resources connected through high speed Ethernet/Infini Band interconnect network. Head, Master and compute nodes perform specific task simultaneously and produce result. High performance cluster increase computing performance exponentially by sharing computing workload. 

‘High performance computer (HPC)’ allows research scholars, scientists, engineers to solve complex science, engineering problems using application various application software that require high bandwidth, low latency network, and very high speed compute capabilities.

For example any normal computer normal computer/desktop PC with 03 GHz processor can calculate 03 billion calculation per second as compare to HPC cluster it can perform quadrillion of calculation per second. 

The microprocessor provides the computing power, The node’s peak performance measured in ‘Mips’ (millions of instructions per second) and Mflops (millions of floating-point operations per second).

High performance computing Purpose :-  HPC is a practice of aggregating computing power in a way that it delivers much higher performance than one may get out of a normal desktop computer or workstation in order to solve large scientific, engineering design, virtual reality military training, gaming and weather forecasting related problems.

What is ‘Head Node/Master ?

The Main Primary machine in HPC environment is called Head/Master Node, HPC ‘Master Node’ contain all the major system configuration and HPC management software, application software, scheduler, etc. so basically it plays a very major role to manage multiple set of servers on rack called ‘Compute Nodes’,.

Master node or head node is a kind of manager of ‘HPC’ system scenario which manages all the hardware, software other storage resources connected to High performance computer (HPC) master node such as memory, storage, licenses software etc.  High performance computer (HPCs)’ are also very helpful in research modelling, simulating and analysis. ‘High performance computer (HPC)’ perform large number of operation per seconds it completes time-consuming operating in less time.   

There are different categories of clusters for different purpose.

Storage Cluster – Storage Cluster provides a consistent file system, allow simultaneous read and write operations to a single shared file system. For example PFS, GPFS etc.

Load Balancing Cluster – It sends network service request to multiple clusters nodes to balance the requested load among the cluster nodes.

High Availability Cluster – It provides continuous availability of services by eliminating single point of failure.  

‘High Performance Cluster (HPC)’ – For the purpose of concurrent calculations execution of program simultaneously on different processors attached to the same computer. Parallel processing divide algorithm and software program into small independent parts and execute simultaneously by separate processors.  HPC is the use of parallel processing for running advanced application programs efficiently reliably and quickly.

HPC Hardware :-

HPC Hardware includes following major list of Hardware Component to build HPC setup such as High-end servers, Storage + networking equipment :

1. One Server or master node one can utilized two servers for master node if redundancy required.
2. Compute Nodes
3. Interconnection switch Ethernet or InfiniBand.
4. Storage NAS box PFS preferred.
5. Cables and other peripheral.

HPC Software:-

1. Cluster Management Software (Rocks, xCAT)
2. Job Scheduler (PBS Torque, SLUM)
3. Application software
4. Compliers like armadillo, gcc, gnuplot, gsl, intel, python, qe, vasp, cmake, gcc, gromacs, hdf5, openmpi python etc.