Why choose a minicomputer as the server? Then you must know minicomputers.
Minicomputer refers to a kind of high-performance computer, whose working principle is similar to PC and server, but its performance and use are completely different from them. It is a high-performance computing product first developed by DEC (Digital Equipment Company) in 1970s.
Minicomputers have a unique architecture that distinguishes PCs from their servers, as well as proprietary technologies of various manufacturers, and some also use minicomputer-specific processors. For example, the minicomputers of Sun Company of the United States and Fujitsu Company of Japan are based on SPARC processor architecture, while the minicomputers of Hewlett-Packard Company of the United States are based on PA-RISC architecture. Compaq is an Alpha architecture. In addition, I/O buses are different, Fujitsu is PCI, Sun is, and so on. This means that plug-in cards on small computers of various companies, such as network cards, graphics cards and SCSI cards, may also be dedicated. In addition, minicomputers generally use Unix-based operating systems, such as Sun Solaris for Sun and Fujitsu, HP-UNIX for HP, and AIX for IBM. Therefore, minicomputer is a closed special computer system. Users who use minicomputers generally value the security and reliability of Unix operating system and the high-speed computing ability of dedicated servers.
At present, the main manufacturers of minicomputers are IBM and HP, Inspur and Dawning. Typical IBM machines are RS/6000, AS/400, etc. Their main feature is that the downtime is only a few hours a year, so they are collectively called zero zero. AS/400 is mainly used in banking and manufacturing, as well as Domino. The main technologies are TIMI (Technology Independent Machine Interface) and single-level storage. With TIMI technology, hardware and software can be independent of each other. RS/6000 is relatively common and is used for scientific calculation and transaction processing.
Minicomputers are just low-cost, small-scale large computers. Typical minicomputers run UNIX or special operating systems, such as MPE and VEM. They are cheaper than mainframes, but they have almost the same processing power. HP's 9000 series minicomputers can almost compete with IBM's traditional mainframe computers.
The common technologies of high-end minicomputers include: RISC-based multiprocessor architecture, megabyte cache, gigabyte RAM, hundreds of GB of disk storage on dedicated I/O channels using I/O processors, and dedicated management processors. They are small and air-cooled, so there is no special requirement for cooling pipes at the customer's place. At present, there is no absolute clear boundary between minicomputers and medium-sized computers and mainframes, because IBM has realized many technologies originally used only in mainframes and medium-sized computers on minicomputers.
Minicomputers are very different from ordinary servers (that is, PC-SERVER). The most important point is the high RAS (Reliability, Availability and Serviceability) characteristics of minicomputers.
RAS is the abbreviation of three English words: reliability, availability and service. They reflect three famous characteristics of computers: high reliability, high availability and high service. Their specific meanings are as follows:
High reliability: the computer can run continuously and never stop.
Availability: important resources are backup; Be able to find potential problems and transfer the tasks running on them to other resources, thus reducing downtime and maintaining continuous operation of production; It has real-time online maintenance and delayed maintenance functions.
High maintainability: real-time online diagnosis, accurate location of root causes, accurate and rapid repair.
Regarding the RAS characteristics of IBM minicomputers, I have to mention the IBM eLiza (Lizard) project.
Eliza originated from a plan of IBM in the mid-1960s, that is, using artificial intelligence technology to design a program to realize the communication between people and computers. In the 1990s, the deep blue supercomputer designed by IBM defeated Kasparov, a chess master, and its comprehensive processing ability is comparable to that of lizards, that is, it has the abilities of prediction, prevention, judgment and self-healing.
In today's IT world, many enterprises are facing a huge challenge: in order to form a powerful and flexible e-commerce operating environment, a large number of servers, network equipment and complex application software are needed. The IT technology involved in these devices is changing with each passing day, and enterprises are seriously short of experienced engineers to maintain and manage the whole software and hardware system.
IBM integrated its own IT experience for many years and put forward eLiza plan, which was endowed with forward-looking and important contents: that is, if modern enterprises want to be invincible in the competition, the software and hardware systems used in their e-commerce environment should have the following four principles:
Self-configuration capability: the system can dynamically self-configure related resources.
Self-protection ability: The system has the ability to protect itself from illegal access and attacks.
Self-healing ability: the system can automatically predict errors, avoid errors, repair errors and replace related error components.
Self-optimization ability: the system can automatically monitor and manage related resources and adjust the system performance to the best state.
In order to achieve the above objectives, the system also needs to have the following seven elements:
Load management (workload management)
Security mechanism (security)
Clustering technology (clustering)
Virtual server hosting
End to end automation.
Disaster recovery mechanism (disaster recovery)
End-to-end system management.
The following describes the RAS features of IBM p690 server.
Automatic first error data capture and diagnostic error isolation function
Self-repairing internal POWER4 processor group redundancy
The industry's first PCI bus parity error recovery
Avoid checking for aborted uncollectible error handling.
Dynamic error recovery
Error checking and correcting (ECC) or equivalent protection on main memory, all 1 level, level 2 and level 3 caches, and internal processor groups.
Continuous redundant bit operation makes main memory have self-healing function.
Chipkill correction function on main memory.
N+ 1 online maintenance of redundant power supply, dual power lines, power supply and fan.
Predictive Error Analysis for Processor, Cache, Memory, I/O and DASD
Based on the runtime error, it is decided whether to separate the running and starting of the processor (dynamic processor offline and permanent processor offline).
Off-line technology of cache and main memory
Reduce failures by selecting high-quality components.
Parallel and cross-running detection technology based on the first fault data capture technology of power supply, fan and I/O subsystem