Clustering

Definition

Clustering is a technique for grouping several independent computers as "nodes" of a single entity called "cluster". There are several types, the most famous are those of High Availability (abbreviated as "HA") allowing a group of master and slave servers to relay in case of failure, and those of calculation (often called "computer cluster") or storage consisting in a single supercomputer (the cluster) pooling computing power or storage space of all of its nodes.

Advantage

Note: We will detail here only high availability solutions.

High Availability

More and more companies, even in Madagascar depend on the IT or are starting to have an obsolute need for it. Availability is therefore a crucial issue in the IT infrastructure. Whether redundancy in case of hardware failure, backup, restore, or recovery in case of error, the need of availability can easily reach the "24 hours a days, 7 days a week" even for organizations having no activities or real-time needs during such period.

Data replication and mode "Master / Slave"

This is the most conventional method in which we cumulate the following techniques:

  • Server redundancy
    • A slave server takes over the master server in case of failure
  • Securing Data
    • Real-time replication, passive or active
  • Outsourcing backups
    • External support and centralization on third party sites

In addition to these techniques, depending on the material used, it is possible to replace parts of a server (hard disk, RAM, power supply) in "hotswap" mode, that is to say, without rebooting, and choose between continuous mode (master / slave with the same power) for constant comfort or degraded mode (less powerful slave) to reduce costs.

Our skills

Thanks to our knowledge of these techniques and tools widely tested, we can announce to our customers an availability rate of 99.9% (called "three nines"), less than nine hours of downtime per year. Our capacity for remote administration, according to the network infrastructure of our customers, even allow us to further increase this availability rate.

The ultimate solution: Clustering + Virtualization

By combining this technique with virtualization, we are proud to offer the ultimate infrastructure in terms of performance, availability and cost reduction. By enabling the virtualization of Master / Slave clusters, you combine the best features of each of these techniques:

  • Less than 1 minute of downtime per year
  • Full redundancy of your data
  • Resource optimization
  • Simplified evolution
  • Optimal security
  • Cost reduction

Shortcuts