Hyperconverged Infrastructure (HCI)
Hyperconverged infrastructure (HCI) combines compute, storage and network virtualisation in a single server type: instead of separate SAN arrays and server farms, identical nodes form a cluster whose local drives are merged in software into one shared, redundant storage pool. You expand by adding nodes rather than new storage systems. Microsoft's HCI stack consists of Hyper-V, Failover Clustering and Storage Spaces Direct and scales from 2 to 16 servers per cluster.
Hyperconverged Infrastructure (HCI) in practice
HCI solves the problem that classic three-tier architectures of servers, Fibre Channel fabric and SAN create three separate lifecycles, three support contracts and three fault domains. In a hyperconverged cluster the drives sit inside the servers; a software-defined storage layer mirrors or erasure-codes data across nodes so that the loss of a disk or an entire server does not interrupt operations.
Storage Spaces Direct supports two to sixteen servers per cluster and, from Windows Server 2019 onwards, over 400 drives and a storage pool of up to four petabytes. The chosen resiliency determines capacity efficiency: two-way mirror yields 50 percent usable capacity, three-way mirror 33.3 percent, and dual parity from four nodes 50 to 80 percent depending on cluster size.
Commercially the main gain is simplification: one hardware type, one update path, one support contact, predictable growth in node increments. HCI reaches its limits where compute and storage demand diverge sharply — scaling then means paying for the resource you do not need. Robust performance depends on the network: inter-node storage traffic runs over SMB Direct, for which Microsoft requires at least 10 GbE on two- to three-node clusters and recommends 25 GbE with RDMA from four nodes upwards. Alendris has delivered such a modernisation for a manufacturing company, among others — a hyperconverged cluster with NVMe storage alongside a network upgrade to 10G.