A Tandem Server (historically built by Tandem Computers and known today as HPE NonStop) is a specialized, fault-tolerant computer system engineered to guarantee 99.999% to 99.9999% uptime and eliminate any single point of failure. Introduced in 1976, these servers were designed specifically for high-volume Online Transaction Processing (OLTP) where even a few seconds of downtime or data corruption could cause catastrophic financial loss.
The technology behind Tandem servers relies on a specialized architecture that separates them from standard server clusters: 1. The Core Architecture
“Shared-Nothing” Design: Tandem servers utilize a Loosely Coupled Multiprocessing environment. Every processor module runs its own dedicated memory, I/O logic, and separate copy of the operating system. Memory corruption or a crash in one CPU cannot physically leak into or affect another.
Active-Active Redundancy: The name “Tandem” comes from the fact that almost all hardware components operate in pairs. Power supplies, cooling fans, and storage devices are dual-ported and cross-connected. If one component fails, its twin seamlessly absorbs the workload without a system restart.
Interconnect Fabric: Because processors share no main memory, they communicate by passing rapid messages across a dual-path, ultra-low-latency network fabric. This technology evolved from Tandem’s original ServerNet into the industry-standard InfiniBand protocol used widely across modern data centers. 2. Software-Driven Fault Tolerance
Leave a Reply