HP NonStop is a series of fault-tolerant server computers designed for online transaction processing (OLTP) and mission-critical applications that require 100% uptime. Originally introduced by Tandem Computers Inc. in 1976, the platform uses a proprietary, integrated hardware and software stack known as NonStop OS (formerly Guardian) to eliminate single points of failure through massive redundancy and “fail-fast” logic.
Historical Timeline by Era
1. The Tandem Founding Era (1974–1981)
- 1974: Tandem Computers Inc. is founded by James (Jimmy) Treybig and a team from Hewlett-Packard’s HP 3000 division.
- 1976: The first system, the Tandem/16 (later NonStop I), is shipped to Citibank.
- 1977: Tandem systems gain early traction as intelligent front-end processors for bank ATM networks.
2. The Stack Machine Expansion (1981–1990)
- 1981: NonStop II is introduced, adding 32-bit addressing capabilities and replacing magnetic core memory with battery-backed DRAM.
- 1983: NonStop TXP (Transaction Processing) launches as the first new implementation of the architecture, featuring cache memory and 2.0 MIPS performance.
- 1986: Introduction of NonStop VLX (Very Large eXpansion) and NonStop SQL, the first fault-tolerant relational database designed for linear scalability.
- 1987: NonStop CLX launches as a lower-cost, compact minicomputer for remote office environments.
- 1989: NonStop Cyclone is released for high-end mainframe markets, featuring superscalar CPUs and fiber optic interconnects.
3. The MIPS RISC Transition & Acquisitions (1991–2003)
- 1991: Cyclone/R (or CLX/R) marks the move to MIPS R3000 RISC processors, using object code translation to maintain backward compatibility.
- 1993: Himalaya K-series is released using MIPS R4400 processors.
- 1995: Open System Services (OSS) is added to the NonStop Kernel to provide a POSIX/Unix-like environment.
- 1997: Compaq acquires Tandem Computers. The Himalaya S-Series introduces ServerNet, which later becomes the InfiniBand industry standard.
- 2002/2003: HP merges with Compaq, bringing the NonStop line under Hewlett-Packard.
4. The HP Integrity & x86 Era (2005–Present)
- 2005: HP Integrity NonStop (TNS/E) is introduced, migrating the platform to Intel Itanium microprocessors.
- 2014: NonStop X (TNS/X) launches, shifting the architecture to Intel x86-64 processors for greater industry-standard alignment.
- 2015: Following the HP corporate split, NonStop becomes part of Hewlett Packard Enterprise (HPE).
- 2020: Sales of Itanium-based systems officially end in July 2020.
- 2024–2025: HPE expands the platform with Virtualized NonStop Software for private clouds and consumption-based models via HPE GreenLake.
HP NonStop Tandem Overview and Timeline History by year
HPE NonStop (formerly Tandem and Compaq NonStop) is a family of fault-tolerant, integrated hardware and software systems designed for mission-critical enterprise workloads. Since its introduction in 1976, it has served as the backbone for high-volume transaction environments like banks, stock exchanges, and payment systems, offering 100% continuous uptime.
Core Architecture and Features
The platform is defined by its “shared-nothing” architecture, where every component is redundant to eliminate single points of failure.
- Continuous Availability: If a hardware or software component fails, a backup takes over immediately without disrupting the application, a process often managed through process pairs (primary and hot backup processes).
- Linear Scalability: You can add capacity (CPUs, memory) seamlessly without downtime. Systems can scale from a few processors to clusters of over 4,000 CPUs while maintaining a single-system image for management.
- Integrated Stack: Unlike standard servers, NonStop includes a fully integrated stack of hardware, the NonStop OS (a proprietary kernel), a relational database (NonStop SQL), and middleware.
- Fail-Fast Design: Modules are self-checking; they stop immediately upon detecting an error to prevent data corruption, allowing the redundant backup to resume processing from the last known good state.
Current Hardware and Deployment
While historically based on proprietary or Itanium processors, modern NonStop systems (NonStop X) utilize industry-standard Intel Xeon processors and high-speed InfiniBand interconnects.
- High-End Systems: Models like the NS9 X5 are built for the most demanding high-volume transaction processing (OLTP).
- Mid-Range/Entry Systems: Models like the NS5 X5 offer fault tolerance for smaller enterprises or development environments.
- Virtualization & Cloud: HPE Virtualized NonStop Software allows the platform to run on standard private cloud infrastructure (e.g., VMware, OpenStack), and it is also available via HPE GreenLake as a consumption-based, pay-as-you-go service.
Software and Security
- Database: Supports NonStop SQL/MX and SQL/MP for multi-tenant, fault-tolerant data management.
- Development: Supports modern languages like Java, C++, Python, COBOL, and the TACL scripting language. Developers can use the Eclipse-based IDE for building and debugging applications.
- Security: Built with Zero Trust principles, including hardware-level vulnerability mitigations (e.g., against Spectre/Meltdown) and real-time threat detection.
Detailed Architecture
HPE NonStop architecture is a fault-tolerant, shared-nothing, massively parallel computing platform designed for 100% operational continuity. Originally developed by Tandem Computers, it is engineered so that no single hardware or software failure can bring down the system.
Core Architectural Pillars
- Shared-Nothing Architecture: Each processor has its own dedicated memory, I/O bus, and copy of the HPE NonStop Operating System (NSK). This eliminates resource contention and single points of failure found in shared-memory systems.
- Massive Scalability: Systems scale linearly by adding more processors. A single node can support up to 16 CPUs, and multiple nodes can be clustered to support over 4,000 CPUs.
- Fault Tolerance (Process Pairs): Software availability is maintained through “process pairs”—a primary process and a passive backup process. If the primary fails, the backup immediately takes over without losing data or state.
- Fail-Fast Design: Hardware and software modules are designed to stop immediately upon detecting an error (“fail-stop”) to prevent data corruption from propagating.
Hardware Components
- Compute Nodes: Modern HPE NonStop X systems use standard Intel Xeon x86-64 processors but implement fault tolerance through specialized system interconnects.
- System Interconnect (Fabric):
- InfiniBand: Used in NonStop X systems for high-speed, low-latency communication between CPUs and I/O devices (up to 56 Gbps).
- ServerNet: The legacy high-speed, point-to-point switched fabric used in older S-series and Integrity i-series systems.
- CLIMs (Cluster I/O Modules): Specialized offload engines for networking (IP CLIM), storage (Storage CLIM), and telco protocols. They handle I/O processing to free up the main host CPUs.
Integrated Software Stack
The NonStop platform is a “tightly coupled” environment where hardware and software are integrated for availability.
- NonStop OS (NSK): A message-based operating system that manages the distributed resources as a single system image.
- HPE NonStop SQL/MX: A distributed, fault-tolerant relational database that provides ANSI SQL compliance and automatic load balancing across the cluster.
- HPE Pathway (TS/MP): An application server and middleware framework that manages workload distribution, load balancing, and automatic process restarts.
- TMF (Transaction Monitoring Facility): Ensures database integrity by managing atomic transactions; if an update fails, TMF automatically rolls back the changes.
Modern Deployment Options
- HPE Virtualized NonStop (vNS): The complete software stack decoupled from proprietary hardware, allowing it to run as a set of virtual machines on industry-standard x86 servers within a private cloud (VMware).
- HPE GreenLake: A consumption-based model providing NonStop capabilities as a cloud-like service.
PROject Templates 