Fiberstack's cloud infrastructure consists of efficient server farms working closely together to build a highly scalable, reliable and self-healing platform.
Unlike other cloud hosting companies, Fiberstack does not use traditional centralized storage. A major caveat of centralized storage is the lack of guaranteed disk I/O performance and the inability to properly scale as more virtual servers are added to the cloud. Additionally, all I/O requests have to go through a network where latency can take its toll on performance.
In designing our infrastructure, Fiberstack implemented a decentralized and distributed storage system. You might wonder what that means and it's very simple. Rather than having a single central storage system acting as a single point of failure, we use each cloud hypervisor and group them into a virtual storage system. This method decentralizes and distributes the storage system across multiple servers allowing it to easily scale as we add more computing power to our cloud. A major advantage of a distributed system is that virtual servers no longer have to read data via the network. Since the data now resides on all cloud hypervisors, the virtual machine can read from the local disk on the hypervisor. In addition, in the event of a hypervisor failure, the workload is automatically shifted to a neighboring server and your data remains intact. This implementation greatly improves resiliency and performance versus traditional centralized storage.