Minicomputer refers to a high-performance 64-bit computer with 8-32 processors, performance and price between PC server and mainframe. Generally speaking, minicomputers have four characteristics: high computing power, high reliability, high serviceability, and high availability: High computing power (High Performance) 1. Adopt 8-32 processors to realize multi-CPU collaborative processing function; 2. Configure massive memory capacity exceeding 32GB; 3. The system is designed with dedicated high-speed I / O channels. High reliability (Reliability) 1. Continuing the high-standard system and component design technology of mainframes and medium-sized machines; 2. Adopt UNIX operating system with high stability. High serviceability (Serviceability) enables real-time online diagnosis, pinpoints the root problem precisely, and achieves accurate and fast repair. High availability (Availability) multi-redundancy architecture design is the main feature of minicomputers, such as redundant power supply system, redundant I / O system, cooling system, etc. Minicomputer is not equal to RISC server! In the past, people often equated the concept of a special-purpose processor based on RISC architecture with a UNIX-type operating system and a minicomputer. However, as CPU manufacturers such as AMD and Intel adopt 64-bit, multi-channel SMP design, multi-core, integrated memory controller and other advanced technologies, X86 architecture processors have begun to subvert RISC processor manufacturers in terms of technical performance and market acceptance. The monopoly position in the field of minicomputers; on the other hand, as the LINUX operating system matures and is widely used, it further accelerates the demise of the UNIX operating system, which is known as a closed and dedicated system. Today, a new type of minicomputer based on X86 multi-core processor, supporting both UNIX and LINUX-like operating systems, with more open and standard features has appeared. While ringing the death knell of traditional RISC minicomputers, it is sent to the majority of enterprise users The gospel of a new era! About the definition of minicomputer: Minicomputer refers to a high-performance computer whose operating principle is similar to PC (personal computer) and server, but its performance and usage are completely different from them. It is a DCE in the 1970s (Digital Equipment Corporation) A high-performance computing product first developed by the company. Minicomputers have a unique architecture that distinguishes PCs and their servers, as well as their own patented technologies, and some also use special processors for minicomputers. For example, the minicomputers of companies such as Sun in the United States and Fujitsu in Japan are based on SPARC. The processor architecture, while the US HP company is based on the PA-RISC architecture; Compaq company is the Alpha architecture. In addition, the I / O bus is also different, Fujitsu is PCI, Sun is SBUS, and so on. This means that the plug-in cards on the minicomputers of various companies, such as network cards, display cards, SCSI cards, etc. may also be dedicated. In addition, the operating system used by minicomputers is generally based on Unix, like Sun, Fujitsu use Sun Solaris, HP uses HP-Unix, IBM is AIX. Therefore, the minicomputer is a closed dedicated computer system. Users who use minicomputers generally value the security and reliability of the Unix operating system and the high-speed computing power of dedicated servers. At present, manufacturers of minicomputers mainly include IBM, HP, Inspur and Dawning. Typical IBM machines are RS / 6000, AS / 400, etc. Their main feature is that the annual downtime is only a few hours, so they are collectively referred to as the z series (zero). AS / 400 is mainly used in banking and manufacturing, as well as for Domino. The main technology is TIMI (Technology Independent Machine Interface), single-level storage. With TIMI technology, hardware and software can be independent of each other. RS / 6000 is more common and is used for scientific computing and transaction processing. The development trend of minicomputers: RISC processors weaken, X86 processors strengthen With the introduction of 64-bit, multi-core CPU, multi-channel SMP design, integrated memory controller and other innovative technologies, the system performance based on X86 architecture processors is getting more and more Powerful, it poses a great threat to servers based on RISC processors. From the global TOP 500 ranking of high-performance computers, AMD and Intel processors increased from 4% in November 2001 to 77.6% in November 2005, while dedicated processors such as RISC dropped from 96% to 22.4%. Weak UNIX operating system and strong Linux operating system On the one hand, the UNIX operating system faces three major dilemmas, such as the closeness brought by the inconsistency of standards, the high cost of expensive investment, and the uncertainty caused by the unclear development strategies of various UNIX vendors. On the other hand, as the LINUX operating system has matured, its performance has been comparable to that of the UNIX operating system, and has eroded the inherent position of UNIX a lot. From the global TOP 500 ranking of high-performance computers, Linux rose from 7.8% in November 2001 to 74.2% in November 2005; on the contrary, UNIX dropped from 88.6% to 20%. The development of the minicomputer industry ecosystem towards openness, standardization, openness, and standardization has always been one of the laws of the global IT industry. It can bring higher cost performance, better scalability, better compatibility, and more Return on investment. It is this rule that is pushing the entire minicomputer ecosystem including CPU, operating system, complete machine, application software manufacturers and users from closed to open. Many SI and software developers have launched more and more application software or solutions based on Linux or Windows. For example, enterprise-level database software such as Oracle and Sybase originally used under UNIX has been successfully applied under LINUX, and Get a test of time. More and more users of traditional RISC minicomputers are migrating to the system to get rid of the shackles brought by closure. Migration is a trend IDC research results show that since 2001, the global Unix server market has been shrinking in shipments and sales. In 2001, Unix servers accounted for 21 billion of the 47 billion US dollars in sales in the entire server market. In 2004, Unix had only 16.2 billion US dollars, and the average annual recession rate exceeded 5%. It is expected that by 2008, X86 servers will not only far outperform non-industrial standard servers in sales volume, but will even catch up in sales. Processor IBM uses the latest Power6 of the Power series; Sun uses the latest UltraSparc T2 of the UltraSparc series; HP uses the Intel Itanium series processor Note: Since SUN was acquired by oracle, how does Sun's successor develop (unknown)? So the following focuses on the comparison: IBM small machine and HP small machine. Some friends asked HP and IBM models to compare, just recently the unit is in the system Selection work (mainly based on Unix-based systems), I have done several technical exchanges with HP and IBM, I will talk about some of the content I know, I hope you will have a preliminary understanding of HP and IBM systems through my introduction understanding. Since the technical exchange is not over, some content may be changed. I use as much information as possible on public sites. First introduce the tpc organization and tpc value: 1. A third-party evaluation organization composed of IBM, Dell, etc., is written by Transaction Processing Performance HP and TPC Council. It can be known literally that this is a "transaction processing performance council." In short, it is an organization that evaluates the performance of each machine. The TPC value is generally the optimized score of each system under full configuration. 2. Several indicators of TPC (only introduce the meaning, please go to the tpc website for details-in fact, I am not good at writing and lazy) 2.1, TPC-C: OLTP (Online Transaction Processing)-The factory's MIS system pays more attention to this indicator, and our unit's selection is basically based on this indicator 2.2, TPC-H: Decision Support for Ad Hoc Queries (decision support based on specific query) 2.3, TPC-W: Web e-Commerce (Internet and e-commerce) 2.4, TPC-R: Decision Support for Business Reporting (decision support based on business report) 3. Overview of TPC-C Since the selection of our unit is basically based on this indicator, I feel that it is necessary to explain further. These contents are based on technical exchange lectures or tpc site materials. The TPC-C indicator (benchmark) is the score of a benchmark test based on a trading model. This transaction model is the simplest transaction system consisting of 9 tables. It has several basic functions: enter and deliver orders, payment records, query order status, and inventory monitoring (These transactions include entering and delivering orders, recording payments , checking the status of orders, and monitoring the level of stock at the warehouses.) Because most of the MIS systems of general enterprises have to carry out the above operations, this indicator has a more important reference for the selection of MIS or ERP systems. effect. 4. Estimation method of TPC value: Having said so much, how should TPC value be estimated? According to the information obtained at the technical exchange meeting, the estimation method of TPC value is roughly as follows (for example): Take a bank as an example, according to According to the analysis of historical transaction records, the bank has 1 million transactions per day, of which the two hours in the morning and afternoon are the peak period, accounting for 75% of the real transactions, then we can get 3125 transactions per minute, and then according to A 1: 10 ～ 20 empirical ratio formula to guess the actual number of transactions to be completed in the computer system to complete these transactions. If we choose the ratio coefficient of 15, then the TPC-C value of this banking system can be roughly determined as 46875, Round up to 50,000. In this way, the TPC requirements of this banking system are determined. This time the TPC value of our system is estimated to be no less than 120,000 and can be expanded to 240,000. This article mainly talks about Unix-based systems, only involving two series of department-level servers and minicomputers. The first part: IBM series IBM claims to have a market share of more than 47% in China's Unix servers. It has not been verified in the second half of the year. However, IBM has a long history of entering the Chinese market. It is not impossible to have such a market share. Yes, I believe IBM will not talk nonsense. 1. Overview of IBM servers 1.1. Classification of IBM server families After integration, IBM servers are collectively referred to as eServer. Under eServer, they are divided into the following sub-series, namely: xSeries-servers based on Intel architecture; servers based on AMD architecture; BladeCenter Blade server; pSeries—UNIX server, the follow-up series of RS / 6000 (the focus of this article); iSeries—medium-sized enterprise server, the follow-up series of AS / 400; zSeries—mainframe, the follow-up series of S390; 1.2, P Series server introduction The p series server is the Unix server series promoted by IBM in the market. The whole P series server has the following models according to the performance from high to low: p610: 1-2 way high scalability entry-level server p630: 1- 4-way POWER4 + UNIX server, support dynamic logical partition p650: 1-8 way UNIX server, using POWER4 chip, support dynamic logical partition p670: new generation UNIX server with 4-16 way processor p690: 8-32 way POWER4 dual CPU chip , Single machine provides 16 logical partitions 1.3. Further introduction of the IBM server (the IBM server mentioned in this article refers to the p series) 1.3.1. Classification of servers: The three models p650, p670 and p690 recommended by IBM this time are all based on Power4 chips. The difference is that the number of supported CPUs and memory capacity are different. If it is not big, it seems unimportant. These three models of servers can be divided into application servers represented by p650 and enterprise servers represented by p670 / 690. Since the current system mostly adopts a three-layer or multi-layer structure, application servers with middleware as the main application generally adopt a horizontal expansion method, that is, a server method to improve performance. The enterprise server represented by Oracle database and SAP system generally adopts the method of vertical expansion, that is, adding CPU and memory to improve performance. Therefore, when choosing a model, pay attention to the difference between the two and plan the system architecture of the enterprise. When selecting the system, the half-matched data is generally taken, which means that the 120,000 TPC value we need is the performance in the case of half-matched. Generally speaking, the TPC value can reach about 70% in the full-matched case in the half-matched case. 1.3.2. The main configuration CPU of the server (unit: way) Maximum memory (unit: G) p650 1-8, 1.2 GHz or 1.45 GHz POWER4 + 64 p670 4-16, 1.1GHz, 1.5GHz or 1.7GHz POWER4 + or 1.1GHz Or 1.3GHz POWER4 128 p690 8-32, 1.5GHz or 1.7GHz POWER4 + or 1.1GHz or 1.3GHz POWER4 512 operating system AIX 5L version 5.1 / 5.2, SuSE Linux Enterprise Server Version 8 (only run in logical partition) 1.3.3. IBM's partitioning technology The partitioning technology is a big difference between IBM and HP series. HP is mainly hardware partitioning and IBM is mainly software partitioning-dynamic logical partitioning. The essence of IBM's dynamic logical partition is a software partition. On the p670 / 690 series, up to 16 partitions can be achieved, and the granularity can reach 1 cpu, and each partition can run a different operating system. The advantage of software partitioning is that it can flexibly configure resources for each partition. For example, if IBM runs a and b2 applications at the same time, and the main load of the 2 systems is day and night, then the advantages of logical partitioning can be reflected. We can The resources of the b application that are lightly loaded during the day are allocated to the partition where the a application is located, and at night, the resources of the partition where the a application is located are allocated to the partition where the b application is located. However, the weakness of IBM's partitioning technology is that it needs to be tested in terms of performance and stability. Since dynamic logical partitions are essentially software partitions, there is definitely a loss in performance. According to other vendors, IBM's software partitions cause performance losses of more than 10%. Moreover, it is generally believed that the system reliability based on software partitioning technology is not as high as that of hardware partitioning. 1.3.4. IBM's TPC scores According to the latest TPC ranking published by TPC, the highest score is currently the IBM p690Turbo model: IBM p690 Turbo 7040-681 CPU: 32, Power4 1.7GHz memory: 512G total system cost : 6,349,223 US $ TPC-C: 763,898 TPC-C version: 5.1 Unit cost: 8.31 US $ Database system: IBM DB2 UDB 8.1 Operating system: IBM AIX 5L V5.2 Middleware: BEA Tuxedo 8.0 Whether cluster: no client Information: Number: 30 CPUs per client: 4 CPU Model: IBM Power 1GHz and another model ’s score may be more valuable: Model: IBM p690 CPU: 32, Power4 1.3GHz Memory: 512G total system cost : 7,591,038 US $ TPC-C: 427,760 Unit cost: 17.75 US $ Database system: Oracle 9i Enterprise Database Server 9.2.0.1 Operating system: IBM AIX 5L V5.2 Middleware: Websphere App. Server Ent. Edition V 3.0 cluster : No Client Information: Number: 14 Number of CPUs per client: 4 CPU Model: IBM RS64-IV 600 MHz 1.4. IBM system software IBM can be said to be relatively stronger in software than HP. The operating system has AIX and is gradually compatible with Linux. The database has DB2 and the middleware has Websphere. And, buy some models of machines are also bundled with unlimited user version of DB2, I think it's worth it. 1.5. IBM's CPU development plan According to IBM's technical communication personnel, IBM currently walks on three legs. One is IBM's own Power series CPUs. Currently, Power5 is basically mature; the second is the traditional IA64 chip; the other is IBM AMD chips are now supported to combat Itanium2. I am deeply impressed by IBM's CPU. The latest Power5 chip integrates a cache of about 1.92MB (3 × 640KB Cache, 10-way joint), and integrates a 32M L3 cache on the MCM, which can run at a speed of 1GHz. This is undoubtedly Will greatly improve processing performance. 1.6. Comprehensive analysis of IBM servers Comprehensive p-series products, we can see that the positioning of IBM servers is relatively clear, and it is excellent in overall performance indicators. However, due to its software partitioning technology, users have to question the stability and reliability of the system. After all, an enterprise like ours needs 24 * 365 operation, and there is no way to remedy the loss of a lot of real-time data. In fact, IBM also has hardware partitioning technology, which is only used on mainframes, like it is on the z series. It is worth mentioning that IBM is indeed ahead of the chip technology. By comparing the TPC scores of IBM and HP, we found that IBM's CPU can basically achieve 1 way to reach 2 ~ 3 HP's CPU, which shows its power. However, IBM's chip integration is very high, like the Power4 + chip used on p690, each chip integrates up to 8-channel CPU, that is to say, 32-channel CPU only needs 4 chips to achieve. Some people may ask, is there no way if I want to add 2 CPUs? Of course there is, IBM has an on-demand purchase method, which is to buy a chip that integrates 4 CPUs and pay the price of 2 CPUs (of course than 2 way more)), then IBM will activate the 2 way CPU for you and the other 2 way CPU can be temporarily not used, wait until you need to pay and then activate it. Part 2: HP series 2. Overview of HP Server 2.1. Classification of HP server family Since HP and Compaq are merged, the product line of HP server seems a little complicated, including both HP's own server series and Compaq server series. Here are two series of servers encountered in this selection: HP rack servers and HP high-expansion servers. HP rack servers are divided into three categories according to the CPU used: IA-32 servers Itanium processor family servers pa-risc servers-to be introduced in this article, pa-risc is divided into many products, I know the following: rp7410 (2-8 CPUs) rp8400 (2-16 CPUs) As for the HP high-expansion server, that is, the hp 9000 Superdom server, generally referred to as Superdome 2.2, HP series server introduction HP is known as the world's No. 1 Unix server market share, according to research. The following describes the three servers encountered in this selection: rp7410: 2 ～ 8 CPUs rp8400: 2 ～ 16 CPUs Superdome: 2 ～ 64 CPUs 2.3. Further introduction of HP server (the IBM server mentioned in this article refers to p series) 2.3.1. Classification of servers: The three models of servers rp7410, rp8400 and hp9000 recommended by HP this time are based on PA-RISC chips. The difference is that the number of supported CPUs and memory capacity are different. It doesn't matter if the performance impact is small. 2.3.2. The main configuration of the server CPU (unit: way) memory (unit: G) rp7410 2 ～ 8 PA-RISC8700 700MHz or PA-RISC8800 875MHz 16G rp8400 2 ～ 16 PA-RISC8700 700MHz or PA-RISC8800 875MHz 16G HP9000Superdom 2 ~ 64 PA-RISC8700 700MHz or PA-RISC8800 875MHz 256G operating system HP-UX 11i 2.3.3. HP's partitioning technology HP and IBM are quite different in partitioning technology. IBM is a soft partition and HP is a hard partition. The two adopt different partitioning technologies related to the system structure. HP is based on the architecture of the Cell board. Each Cell board has 4 CPU sockets. The smallest unit of each hardware partition is a Cell board. This partitioning method is not very flexible, but because it is a hard partition, it is generally considered to be more stable and reliable than a soft partition. However, the disadvantage of hardware partition is that it is not flexible. If you want to change the partition system, you must restart it. HP also has software partitions, but Oracle ’s license mechanism does not recognize HP ’s soft partitions and only recognizes the number of CPUs in hard partitions. Users who purchase Oracle licenses by cpu should pay attention, but according to IBM, Oracle can correctly identify IBM ’s software partitions. Partition. Regarding the Cell board, HP's machine is based on the Cell board, which has CPU and memory. According to HP's introduction, I think that the Cell board is somewhat similar to the motherboard in our PC, but in the HP machine, there are multiple Cell boards, such as Superdome has 16 Cell boards. It is precisely because HP's machine is based on the structure of the Cell board, so HP uses hard partitioning technology. In the hard partition, different partitions can be divided by the combination of Cell board resources. Each partition is similar to an independent machine, so from this perspective, the hardware partition is more reliable than the software partition. 2.3.4. HP's TPC score According to the latest TPC ranking published by TPC, HP also has a model with more than 700,000 Model: hp superdome Client / Server CPU: 64 channels, Intel Itanium 2 Processor 6M 1.5GHz memory: 256G total system cost: 5967033 US $ TPC-C: 707102 TPC-C version: 5.1 unit cost: 8.44 system shipping date : 2003-10-23 Database system: Microsoft SQL Server 2000 Enterprise Ed. 64-bit Operating system: Microsoft Windows Server 2003 Datacenter Edition Middleware: Whether cluster: No Client information: Number: 64 CPU number per client: 2 CPU model: Intel Pentium III 800 Mhz Because the above system is based on WinTel, in order to be comparable with IBM's system, we look at the score of the following HP models: Model: HP 9000 Model Superdome Enterprise Server CPU: 64 Road, HP PA-RISC 8700 875MHz memory: 256G system total cost: 6621072 TPC-C: 423144.41 TPC-C version: 5 unit cost: 15.64 US $ system shipping date: database system: Oracle 9i Enterprise Database Server 9.2.0.1 Operating system: HP UX 11.i 64-bit Middleware: BEA Tuxedo 8.0 Whether it is clustered: No Client information: Number: 1 CPU per client : 28 CPU Type: HP PA-RISC 8700 875MHz 2.4. HP system software HP is not as powerful as IBM in terms of software. HP has no other software other than its own operating system HP-UX11i, but this helped in the TPC test. All the systems running on HP It is a third-party software, which makes people think its score is more objective. And IBM can all run tests with its own things. At this point, I do not know whether to say that IBM is not open enough or strong enough. 2.5. HP's CPU development plan HP now has four types of CPUs, namely Alpha, PA-RISC, IA-32 / 64 and Itanium2. However, according to HP's own introduction, the Alpha series is no longer being developed and will be discontinued in 2005. The PA-RISC series has the highest frequency up to the current 875MHz. The next-generation PA-RISC will use IBM's multi-core technology to integrate 2 CPUs in a chip. HP will fully switch to Itanium2 series CPU in the future, and guarantee binary compatibility from PA-RISC to Itanium2. 2.6. HP server comprehensive analysis The several HP servers we contacted this time are very consistent in architecture, and HP promises that the Superdome can be upgraded to the next-generation PA-RISC chip. HP's system is basically tied with IBM in TPC testing. Part 3: Comparison of HP and IBM machines 3.1. Through several technical exchanges and some information about CPU and memory, we found that the number of CPUs used by HP in the TPC test is generally double that of IBM. According to IBM ’s own tests, the latest Power4 + 1.7GHz chips are all the way. Equivalent to 3.6 channels of HP's PA-RISC8700 875MHz. We can see that IBM is much ahead of HP in central processor technology. However, we also found that the amount of memory used by IBM is double that of HP. Superdome and p690 are also under full configuration. Superdome is equipped with 256G memory and p690 is equipped with 512G memory. When choosing a system, a CPU-to-memory ratio is generally established, which is generally 1: 1 to 2. If it reaches 1: 4, it is very good. If we only look at the CPU-to-memory ratio of HP and IBM, we find that HP is 1: 4 and IBM is 1:16. At this point, HP has always ridiculed IBM that IBM's system needs to consume a lot of memory and thus increase the cost of the system. However, we can make an analogy in this way. In the PC world, the memory capacity increases as the chip frequency increases. I personally think that in a memory environment with the same capacity, a fast CPU must be able to handle more tasks than a slow CPU. 3.2. About partition technology When introducing IBM and HP separately, the differences and similarities of the partitioning technologies of IBM and HP have been introduced. In general, IBM's partitioning technology is more flexible, while HP's partitioning technology is more robust. IBM plans to further enhance the dynamic virtual partition technology in the future, so that it can be partitioned on one CPU, that is, the CPU is divided into several CPUs to use. IBM's dynamic virtual partition technology has been developed and used for more than 20 years. I believe that it is also tested in performance and stability, but it is not stable enough compared to HP's hardware partition. However, large enterprises often need stability over performance, hoping that IBM can perceive this demand and improve partitioning technology. From the practical application of partitions, HP's multi-CPU system is conducive to partitioning, but HP's system is a hard partition, so it is a bit troublesome to do. 3.3. Redundancy IBM and HP have done a good job in the minicomputer. IBM uses its ChipKill technology to isolate faulty components (including various PCI cards, CPUs, memory, etc.) so as not to affect the normal operation of the system. According to the introduction, ChipKill technology monitors the system components through a system monitoring chip and related software, and the isolated faulty components can be hot replaced. After replacing the good components, the system can automatically replace them. of Components are added to system resources. Based on ChipKill technology, IBM has further developed a system called "Electronic Lizard" for self-repair and self-regulation of the system, thereby ensuring the normal and stable operation of the system. HP's redundancy is basically carried out through redundant equipment. For example, HP provides two PCI buses (pci bus) for the same PCI interface. HP products also support hot replacement.
We are manufacturer and supplier of Coffee Bean Canister,Tea Canister,Tea Container,Tea Storage Container,Stainless Steel Coffee Canister. And we located in Jiangmen, Guangdong, China. If any interested, please contact us for free.
Coffee Bean Canister,Tea Canister,Tea Container,Tea Storage Container,Stainless Steel Coffee Canister Jiangmen Xinhui Zhancha Metal Products Co,. Ltd. , https://www.zchardware.com