PFSOFT has developed a so-called «high-availability» cluster solution option for servers hosting the company’s Protrader trading platform.
A cluster is a group of computers which work together in order to:
- Combine the processing power of the individual computers into one powerful server to increase speed of processing for large amounts of transaction data (High Performance (HP) cluster).
- Provide a more reliable «fail-safe» environment (High Availability (HA) cluster).
- Balance loads between individual computers in the cluster (Load Balancing (LB) cluster).
A high-availability (HA) cluster is a type of cluster designed primarily to ensure continuity of the services which it provides, but also allows «scalability». This solution requires at least two nodes (computers) which replicate each other. If either node fails, then the other (stand-by) node takes over operation of service requests immediately.
Protrader’s offers two HA cluster solutions (which also facilitate scalability): simple and extended. The simple scheme provides for stand-by of both the main servers for the Protrader platform to function on: one server for processing client requests and one database server. This is achieved by deploying one cluster for Tomcat (Nodes 1 and 2 for Tomcat) and another cluster for the database (Nodes 1 and 2 for the database).
The extended model (see diagram) allows elimination of several drawbacks of the simple scheme, giving the following advantages:
- Improvement of the degree of high-availability by utilizing standby switches and communications channels. On failure of one or more communication channels or of one of the Back-End or Front-End switches, the system will continue to function.
- Use of shared storage for database files provides for a significant decrease in replication traffic between Nodes 1 and 2 for the database.
- Deployment of additional communication channels significantly improves the degree of high-availability of clusters with relatively low additional financial investment.