At Clouding we've developed a 100% redundant platform. This means that the entire platform has been designed so that the failure of a component does not affect normal operation.
Here are some examples of redundancy:
Electrical power supply
- 3 x medium voltage lines
- The data center is powered by 3 medium voltage connections from different geographical points
- 4 x generators
- 4 generators available (2 x 1 Megawatt and 2 x 750 Kilowatts)
- The data center can remain active with only 2 of the 4 generators
- 2 x power lines per rack
- The data center is divided between power lines A and B/li>
- Each power line is independent from the entrance to the data center, to the power supply of the equipment
Equipment
- 2 x power supplies per device
- Each device has two power supplies, each connected to a different electrical segment of the data center (A or B)
- 2 x network connections
- All the devices that make up the platform have at least two network connections in high availability (LAG).
- Each network connection goes to a different switch, so that a failure in a switch does not lead to any break in the service.
- N x 2 fans
- The equipment can withstand the failure of several fans before overheating.
- ECC RAM
- We only use ECC memory in all of our equipment. This protects the RAM against data corruption, which could give errors in calculations or even cause the failure of an entire computer.
Hypervisors
- Hypervisors in N x 1.25
- We have enough space to support even the failure of 25% of hypervisors
- In the case of failure of a hypervisor, servers hosted on it are automatically started on other hypervisors
Storage
- 3 x synchronous data replicas in different cabins
- Each time a cloud server writes to disk, writing is done the same time in 3 different NVMs, housed in different storage booths.
- The system can be kept online with only one replica of the 3 available
- replacement disks
- In the event of a disk failure, the data is automatically replicated to other discs.
Net
- 3 x switches per rack
- 2 x 40G ethernet switches for equipment connection
- 1 x switch for out-of-band management
- 4 x different connection paths per rack
- Each rack connects to the central Racks through 4 40G connections with a total bandwidth of 160G per rack.
- 2 x switch power supplies
- Each switch has two different electrical connections, one connected to the A socket and the other to the Rack B socket.
- Multiple fiber optic operators
- We have several 10G circuits with local data center providers or with links to other data centers, so that a problem in a provider never affects connectivity.
Refrigeration
- Cooling N x 2
- The data center has two cooling systems.
- During normal operation, both systems alternate.
- In the case of failure, only one of the two systems is able to maintain cooling.