Science and education. Computer storiesstories in detail about computer parts
Rumors of the death of mainframes have been greatly exaggerated. Well Fargo Bank's interactive customer service system is entirely built on them. American Express, Charles Schwab and Federal Express interact with their Internet clients using the iron giants, and Chrysler's 25,000 Lotus Notes users are served on mainframes.
Well-known truck manufacturer Vehicle Caterpillar uses mainframe computers to produce 4.5 million technical drawings. The server platform for the airline ticket reservation system SABRE (Semi-Automated Business Research Environment) is again built on mainframes and easily processes over 5 thousand transactions per second during peak hours, which amounts to more than a billion transactions per week. It is therefore not at all surprising that mainframe software vendors such as Oracle, Information Builders, Platinum, SAS and Software AG continue to develop for MVS-compatible systems. There is no doubt that the mainframe business is alive and well.
The main differences between mainframes and PCs and servers are as follows: mainframes provide a high level of reliability due to hardware redundancy; Mainframe operating systems are optimized for batch operation and transaction processing; programs for mainframes are written well known languages third generation Cobol and Fortran, not fourth generation SQL, C++ or Java.
It is obvious that it is too early for mainframes to leave the arena, because only they are able to satisfy the highest requirements for reliability, security and simply computer power that the Internet imposes today.
Dignified old age
Actress Bette Davis once remarked, "Getting old is no girl's business." This statement, the meaning of which is that true elegance is inherent only in adulthood, is quite applicable to computers. At the heart of modern e-business is a scalable three-tier computer architecture. PCs, thin clients, mobile personal assistants and Internet access devices with HTML browsers form the first level of this architecture. The Internet links this layer to Unix and NT servers that support the HTTP protocol and application server software that form the second layer. Second-level systems are connected to third-level databases, where mainframes dominate, since it is at the third level that huge amounts of information have accumulated over the years. So much effort and time has gone into creating databases on mainframes that converting them to another format would require companies to spend more than their entire capital. In addition, these databases solve problems of unprecedented complexity: a billion transactions per week are common for such systems. So why switch from a proven and excellent working system to a solution that may be less reliable and effective?
Fast, reliable and highly available databases require machines of the appropriate class. As shown in Table 1, only mainframes today satisfy the basic criteria: their reliability and availability outweigh the advantages of modern technologies, in particular Unix and Windows 2000. The importance of characteristics such as security, reliability and data integrity is increasing, and in this regard mainframes still outperform other platforms.
| Type | Platform | Downtime per server per year | Availability as a percentage of 24 hours, 365 days a year availability |
| Mainframe | IBM S/390 (sysplex, an overlapping interconnect technology that creates a cluster of processors that behaves like a multiprocessor system) | 10 minutes | 99,998 |
| Compaq Tandem Nonstop | 1.7 hours | 99,98 | |
| IBM AS/400 | 5.2 hours | 99,94 | |
| IBM S/390 (not sysplex) | 8.9 hours | 99,90 | |
| Mini- | Digital VAX | 18.9 hours | 99,78 |
| computer | Unix machines (all Unix variants) | 23.6 hours | 99,73 |
| PC | Windows NT | 224.5 hours | 97,44 |
| Source: K.Fizpattrik, “Platform Availability Data: Can You Spare a Minute?” GartnerAdvisory: Research and Advisory Services, Research Note, Decision Framework, 29 Oct. 1999 |
|||
Three-tier architectures may not be ideal, but they meet the realities of today's computing world. And while Sun Microsystems may be credited with being the first supplier of Internet database machines, there's no denying the fact that mainframe sales are still huge.
Clear signs of life
Unbiased numbers (Fig. 1 - 3) indicate that mainframes are doomed to longevity. Mainframe vendors prefer to measure sales in millions of instructions per second rather than in the number of machines sold: given the scalability of these systems, it makes more sense to use MIPS as a unit of measurement. Between 1995 and 1997, annual shipments expressed in MIPS IBM systems S/390 grew from 320 to 900 thousand. In Fig. 1
shows the price dynamics for mainframes - this curve demonstrates the speed of improvement of these machines. Mainframe prices are falling 32% per year, which is comparable to the rate at which PC prices are falling. Based on the graph, you could pay 32% less for the same amount of MIPS next year than you did this year, or buy 32% more MIPS for the same money. We also note that in terms of scalability, mainframes are ahead of all other types of computers, thanks to their enormous bandwidth |
| Rice. 2. Forecasted mainframe shipments (in MIPS) |
Let me emphasize again: mainframes are the powerful foundation on which all Internet technologies, including the Web, rest. Schwab.com has added two more mainframes to the six it already has in each of its two data centers. Eight redundant mainframes duplicate the main systems. Schwab analysts estimate that the need to comfortably cope with unpredictable surges in data demand is causing peak throughput to triple from time to time. As they say, time is money, and downtime is costly for companies. According to Standish Group research, the cost of downtime ranges from $1,000 per minute for regular email to $13,000 per minute for enterprise applications.
Renaissance
Software developers responded immediately to the rise in the mainframe business. IBM's S/390 Partners program included 100 suppliers in 1995, and by 1997 the number of participants exceeded a thousand. The OS/390 operating system includes all standard Unix APIs, and the API for Windows developers from Microsoft can be purchased from Bristol Technology (http://www.bristol.com/). Bristol's Win/U interface allows you to recompile NT applications for the MVS Open Edition OS, and thus get a mainframe version. There is not much interest in porting by other means - why stoop to the Windows level if you can take advantage of the technical and economic advantages of the MVS system.
Representative of the sixth generation of IBM mainframes, computer system The 390, which was released in June 1999, is a good example of the computing power available on modern mainframes. The sixth-generation system can run up to 12 processors, which collectively provides 1600 MIPS on a single machine. The System 390 delivers 50% more performance than the fifth-generation model, introduced just nine months earlier. According to George Walsh, director of production for the S/390 line, the 1600 MIPS machine uses less electricity than a typical household hair dryer.
As follows from the figures, the price dynamics for mainframes are comparable to those for PCs and workstations. But the mainframe's resurgence in popularity has a deeper reason: its inherent superior bandwidth. $350,000 may seem like a lot for a sixth-generation single-processor mainframe, but keep in mind that this system has 32 GB of memory and can be expanded to hundreds of terabytes. Compared to Unix or NT machines, the mainframe has enormous bandwidth, for example, the System 390 supports 24 fiber channels, parallel sysplex, HiPerLinks (IBM's processor interconnect technology) and Gigabit Ethernet. All together this amounts to 256 channels. Need peripherals? This input/output system allows you to connect up to 16 thousand external devices. As the sidebar "A Note About the PC Farm" illustrates, PCs may have given up any hope of matching this unique machine's performance and scalability.
Mainframes are no slouch in the price war. Take, for example, Microsoft's Web server rack, which consists of a thousand 4-processor Pentium servers, which in MIPS is equivalent to about 20 mainframes. Let us assume that each such system costs
10 thousand dollars, then Microsoft will pay 10 million dollars for its configuration. For this amount I can buy from 10 to
20 mainframes, so the total investment is comparable.
However real problem- support and reliability. Microsoft will actually have to pay more because it needs to keep its network of servers up and running. That's why large corporations like Charles Schwab are returning to mainframes - the "iron monsters" provide huge advantages in cost of ownership and availability.
In the 70s, when the scale of IBM's monopoly exceeded the power of today's Microsoft, S/360 ruled the world information technologies. Back then the data was stored on mainframes, which did everything. Today, mainframes still store all the data and do much of the work. Stupid green-screen terminals were replaced by personal computers, but the Internet again returned mainframes to their former popularity. How more changes, the more consistency.
Amdahl, Bull, Unisys and other manufacturers of "iron giants" expect to make big profits from the emerging return to the style of the 70s. 25 years ago we talked about commerce, now we talk about e-commerce, but the requirements for security and reliability of enterprise-level systems have not changed. Only mainframe hardware and operating systems meet these requirements. True, IBM's competitors intend to transfer much less reliable Windows N.T. Are they smart enough not to do this?
About the author
Ted Lewis - Professor at the US Navy Graduate School, head of the Technology Assessment Group. Ted is the author of Microsoft Rising... and Other Tales of Silicon Valley, published by Computer Society Press.
T. Lewis. Mainframes Are Dead, Long Live Mainframes, - IEEE Computer, August, 1999, pp. 102-104, Reprinted with Permission, Copyright IEEE CS, 1999. All rights reserved.A note about the PC farm
Using several mainframes instead of several thousand PCs makes more sense in every way. It will be cheaper, easier to maintain and safer. All the shortcomings of a PC-based solution are demonstrated by Microsoft's Web server racks.
This whole machine receives 190 million requests per day and downloads 18 GB of data daily in the form of 324 thousand files and 307 thousand graphic images. About a thousand four-processor Pentium-based systems, each with 512 MB of RAM, are connected to each other by six local networks with a throughput of 600 Mbit/s. Two 1.2 Gbit/s OS-12 lines provide communication with the outside world. You can imagine the size of this enterprise and the hundreds of system administrators who rush around trying to keep individual servers in working order. It's not hard to guess how often one of the machines breaks down.
However, the server racks process fewer Web page hits each week than the small group of mainframes that support the SABER backup system. All of Microsoft's monster RAM can fit on one mainframe, and the bandwidth of six Microsoft local networks is equivalent to the bandwidth of just three mainframes. If Microsoft specialists remembered mainframes, the company would save money a large amount money and got a more reliable, manageable and scalable Web site. However, from a marketing point of view, using mainframes can make a bad impression on potential Windows NT customers.
IBM Corporation introduced a new family of mainframe servers, the IBM z Systems family, the first representative of which was the IBM z13 server.
Z13 is one of the most powerful servers of this type, while z13 is distinguished not only by high performance, but also by the ability to encrypt and analyze transactions in real time. It took five years and about a billion US dollars to create this mainframe. In the process of creating the z13, innovative technologies from more than 500 new patents were used.
The results were quite impressive. For example, the system can process approximately 2.5 billion transactions per day, and the technologies used in z13 allow fraud detection in real time for 100% of business transactions, instantly providing the user with analytical data for assessing the operation.
Nowadays, an increasing number of transactions are carried out from mobile devices. And any such operation uses data about past purchases, initiates the procedure for encrypting and decrypting information, checking the user’s form and other processes. According to experts, a basic user operation involves, on average, about 100 different system interactions. And every such interaction can be exploited by scammers. IBM MobileFirst technology running on the z13 helps solve this problem.
The advantage of the z13 mainframe is not only the ability to detect fraudulent transactions, but also a powerful processor. The developers claim that this is the fastest microprocessor that bypasses everything existing solutions, exceeding them by approximately twice. In this case, a vector type of calculation is used to conduct mobile transactions. As for the advantage in memory capacity, the estimate is 300%, bandwidth is 100% higher than other solutions. The processor has 8 cores, manufacturing technology is 22 nm. The clock frequency is 5 GHz. The maximum configuration of the z13 includes 141 processors and 10 TB of RAM.
When using a mainframe, you can run up to 8,000 virtual servers, which is about 50 servers per processor. A preliminary estimate of the performance of one processor on “traditional” mainframe applications is 1695 MIPS (millions of operations per second). Maximum performance for 141 processors is more than 111,000 MIPS. Not long ago, IBM conducted a survey among directors and managers of enterprise IT departments. The topic of the survey is the priority of various factors for the sphere mobile technologies. 71% of respondents noted that safety is the most important factor in this area. And now new IBM technologies provide built-in encryption and real-time analytics.
According to Mike Gilfix, director of corporate mobile solutions at IBM, the z13 mainframe allows companies to process great amount customer requests, and ensure fast order processing. The capabilities of z13 allow companies to increase the level of personalization of their services - in real time, the system evaluates behavioral factors and provides information about customer behavior patterns (habits).
“Today, businesses do not have the ability to analyze 100% of consumer transactions. Thanks to the z13 system, enterprises will be able to use IBM technologies for intelligent analytical modeling, software statistical processing data, and quickly personalize operations at the time of their implementation,” said the IBM press service.
It is worth noting that the mainframe is open platform, with support for Linux and OpenStack. The corporation has also released a new OS release - z/OS, which uses a new type of analytics, has support for Hadoop and support for data storage systems. Currently, IBM Corporation is largest producer mainframes in the world.
Reference:
- The IBM z13 server is available in five models: 2964-N30, N63, N96, NC9, NE1
- Construct - two-frame
- Cooling - hybrid (internal circuit - liquid, external - air or water)
- Power supply - 3 phase
- Processor: 8 core, 5.0 GHz, 22nm technology.
- Customers have access to from 1 to 141 cores, such as CP, IFL, ICF, zIIP, SAP.
- zAAP - not used, zAAP functionality is fully supported by zIIP
- For zIIP and IFL, up to 2 streams are supported (SMT)
- The permissible ratio of the number of zIIPs to CPs is 2:1
- Single core performance - 1,695 MIPS (max - 11,556 MIPS)
- Maximum RAM - 10 TB (RAIM)
- Maximum number of logical partitions (LPARs) - 85
- Maximum memory capacity per LPAR is 10 TB
- Maximum number of logical input/output subsystems (LCSS) - 6
- Maximum number of I/O adapters - 160
- Maximum number of FICON/FC channels - 320 (FICON Express16S)
- Maximum number of LAN ports - 96 (OSA-Express5S)
Story
History mainframes It is customary to count from the appearance in the year of the universal computer system System/360, on the development of which IBM Corporation spent $5 billion. The term “mainframe” itself comes from the name of the typical processor racks of this system. In the 1960s and early 1980s, System/360 was the undisputed market leader. Its clones were produced in many countries, including the USSR (ES series of computers).
IBM mainframes are used in more than 25,000 organizations around the world (not including clones); in Russia, according to various estimates, there are from 1,500 to 7,000 (including clones). About 70% of all critical business data is stored on mainframes.
Sources
see also
Links
| Computer dimensions | |||||||
|---|---|---|---|---|---|---|---|
| Large | Supercomputer · Mini-supercomputer · Mainframe· Minicomputer · Supermini | ||||||
| Desktop | Microcomputer · Home computer · Personal computer · All-in-One · Workstation · Server | ||||||
| Mobile |
|
||||||
| Other | Embedded system · Wireless sensor network · Smartdust · Nanocomputer | ||||||
Wikimedia Foundation. 2010.
Synonyms:The highest reliability is due to the double or triple redundancy of nodes and blocks incorporated during the design, as well as unique architecture and structure designs in combination with solutions for internal mathematical software and external operating environment. One of the main advantages is the organization of the input/output system, in which simultaneous parallel performing operations on many dozens of input/output channels without stopping the calculation processes. The presence of 10 to 20 processors and parallel execution of input/output provide high performance and computing power and allow you to simultaneously solve many problems.
One of the most important advantages of the mainframe platform is the fact that the mainframe architecture, combined with the capabilities of IBM operating systems, allows you to dynamically reallocate computing resources between tasks in accordance with current needs and priorities. This means that the highest priority tasks at certain critical moments can be transferred to all or the largest part mainframe computing power. Thus, a fairly uniform load of computing resources over time is achieved.
Modern models of IBM mainframes, which are the development of the line S/390, under the general name eServer zSeries, based on architecture z/Architecture, which is an extension of the architecture ESA. This architecture allows you to provide full support for 64-bit real and virtual memory, supports clustering (up to 640 processors) and virtual machines that allow you to run hundreds and thousands of instances of other operating systems, in particular Linux, allows you to eliminate problems associated with a lack of addressable memory, and with the help of an Intelligent Resource Director (IRD) can automatically direct available resources to solve the highest priority tasks.
Reliability and fault tolerance of IBM z mainframes Series provided through a number of functions: dynamic memory reservation , extended dynamic system reconfiguration, opportunity rebooting part of the memory - this is only a small part of an extensive list unique opportunities, embedded both in the architecture and in the system software.
Due to the expanding useLinux on mainframe z/Series and S servers /390 (models 9672 generation G 6) gain new life. On one platform with high reliability, performance and scalability, it is possible to install a large number of different applications using different operating environments.
When using Linux technologies running the z/VM operating environment, you can have hundreds or thousands of Linux servers on one mainframe platform. A separate Linux server is used for each application and each database. Linux servers are effectively used to create file servers, mail servers, web servers, etc. Thus, there is a consolidation of Linux servers on one platform.
When using z/VM, you can create new Linux servers online as needed. When creating Linux servers running z/VM, there is no need to strictly assign a certain amount of RAM to each server. In this case, the memory for the servers is virtual. To connect Linux servers with local user networks via the Internet, a special node available in the mainframe architecture is used - the Open System Adapter (OSA), which allows you to exchange data at speeds of 1 or 2 Gbit per second. for the S/390 platform (models 9672 generation G6) and 1, 2 or 4, and in the latest developments and 10 Gbit per second. for the z/Series platform. At the same time, one OSA adapter can be used by many or all Linux servers on the mainframe, dynamically switching between them depending on current needs and priorities.
The zSeries mainframe family offers powerful data security features in both software and hardware. To ensure data security, zSeries family mainframes include a built-in hardware programmable cryptographic adapter that enables SSL and public key encryption operations. The experience of many clients and IBM itself proves that “big hardware” opens up great opportunities, including in the area of reducing infrastructure costs.
Publications
Oleg Takovitsky
In the relatively short history of computing, mainframes have typically been the main protagonists. Indeed, these computers, otherwise also called mainframe computers, in the era of the 60-80s. of the past century almost completely dominated the information technology market. By the early 80s, the division of computers into large, mini and micro machines was simple and clear. It was determined by price physical dimensions, performance, the scale of the tasks being solved, the system software used (primarily the operating system), as well as the architecture. The very concept of "mainframe" is inextricably linked with the name of their first manufacturer, IBM Corporation (http://www.ibm.com).
Yet mainframes are more than just powerful and expensive cars. They have been and remain the basis for ensuring reliability, safety and integrity information systems. And most importantly, for several decades now, these computers have served as an outpost for the centralization of functions and data, and have not fallen under the onslaught of distributed computing. Nowadays, centrifugal forces of control and management in architectural solutions begin to change direction, turning centripetal. It becomes clear that without centralizing resources (in other words, consolidation), it is almost impossible to solve many serious business problems.
At the end of the last century, thanks to someone (they say it was one of the top managers of Sun Microsystems), mainframes were called dinosaurs. In addition, the press and leading research agencies, wittingly or unwittingly, contributed to the creation of their negative image. Many began to perceive mainframes as yesterday's computing technology, considering Unix and PC servers to be more modern and promising.
Generally speaking, one of the reasons for the sharp decline in interest in mainframes in the 80s was the rapid development of PC and Unix-based machines, in which, thanks to the use of new technology for creating CMOS chips, it was possible to significantly reduce power consumption, and their sizes reached the size of desktop stations. At the same time, the installation of mainframes required huge areas, and the use of outdated semiconductor technologies entailed the need for water cooling. So, despite their computing power, due to the high cost and complexity of maintenance, mainframes were less and less in demand in the computing market.
The main argument against mainframes was that they do not respect the basic principle of open systems, namely compatibility with other platforms.
Taking the criticism constructively, management at IBM, a major manufacturer of mainframe hardware and software, developed a radically new strategy for the platform to dramatically improve performance, reduce cost of ownership, and achieve high reliability and system availability. The achievement of these plans was facilitated by important changes in the technological field: the bipolar technology for manufacturing mainframe processors was finally replaced by CMOS technology. Transition to a new one element base allowed to significantly reduce the level of power consumption of mainframes and simplify the requirements for the power supply and cooling systems ( water cooling was replaced by air). Mainframes based on CMOS chips quickly gained in performance and lost in size. The most dramatic event was the transition to the 64-bit z/Architecture architecture. Modern mainframes are no longer a closed platform: they are capable of supporting hundreds of servers running various operating systems, including Linux, on one machine.
Among the basic differences between mainframes and conventional servers, it is usually noted that mainframes maintain a high level of reliability due to hardware redundancy, and the operating systems for them are optimized mainly for batch operation and transaction processing. However, IBM believes that one of the most important reasons for the market acceptance of zSeries systems is their support for such non-mainframe tasks as applications for Linux and the Web.
Mainframes are highly regarded for their resilience to problems such as failure central processor. They are equipped with special software, and their memory and data transfer subsystems differ from those used in most servers. Mainframes' ability to process enormous numbers of transactions per second has made them widely used in tasks such as managing distribution networks or maintaining bank accounts.
Rumors of the death of mainframes have been greatly exaggerated. Total zSeries server shipments, measured in MIPS (millions of instructions per second) processing power, grew 3% in the first quarter of this year compared to the same period last year. Mainframe sales accounted for $3 billion of IBM's total 2002 turnover of $81 billion. Despite the pessimism of some analysts, these systems are still popular in the 21st century. Thus, according to one of the forecasts of the Gartner Group, the last mainframe was supposed to be turned off back in 1993. This forecast has long since expired, but the mainframe market remains stable.
It is worth recalling here that, according to one theory, dinosaurs did not become extinct, but turned into birds.
| IBM S/390 mainframe. |
The first "swallows"
At the heart of the information system there are usually computers of the highest level of reliability and performance, designed to withstand any conceivable load levels and be prepared for various failures and accidents. This is a series of powerful, scalable, easy-to-manage and reliable systems. These are more than systems - rather, they are solutions that meet the requirements of the corporate objectives of the different scales: from workgroups to large data centers.
These systems are suitable for solving almost any problem - from scientific and engineering to business, requiring large computing power. They have a well-balanced multiprocessor architecture, with the ability to boot multiple independent copies of the OS. The scalability of the architecture allows you to obtain a calculated, controllable performance increase when increasing the number of processors and memory. The large amount of RAM in such systems creates new, previously unavailable capabilities in many application areas - from maintaining large resident databases to complex scientific computing, in areas such as human genome research or offshore oil exploration.
When IBM began rebranding its server systems in October 2000, the changes were presented as a response to the growing demands of Internet businesses. The corporation's management announced its intention to use open standards and products (TCP/IP, HTML, Java, XML, Apache) on all platforms and a desire to support the rapidly gaining popularity of the Linux OS. It was then that mainframes received the name eServer zSeries - it indicates zero down time of these systems.
The z/Architecture on which zSeries systems are based provides new standard performance and integration, serving as a continuation of the balanced system concept in the S/390 architecture. These systems can eliminate addressable memory bottlenecks, providing virtually unlimited 64-bit addressing capabilities and providing enormous headroom for unexpected workloads and growing enterprise applications.
Mainframe ProcessorsOver the past few years, IBM has been borrowing technology from its mainframes for other server families. Now it is often said that this process has gone into reverse. The company is considering a major overhaul of the zSeries mainframe architecture to use the same Power processors and other technologies as the less expensive pSeries and iSeries servers. By using the same processors across all product families, IBM is able to significantly reduce the cost, time to market, and maintenance costs of mainframe computers. However, to achieve this, new processors must have improvements that bring them closer to mainframe processors. For example, a mainframe processor has two processor cores that perform the same operations simultaneously. If the results are different, the calculation is automatically repeated, perhaps several times, and if the difference still persists, the task is transferred to another processor. Switching to new processors is not an easy task, but IBM already has similar experience. The company managed to transition the AS/400 family to the Power architecture without losing its customers. Blue Flame dies, packaged into multi-chip processor modules, are specifically designed for high-end zSeries systems. 20% of the area on Blue Flame is allocated directly to support RAS (Reliability, Availability, Serviceability) functions to provide computer systems the ability to respond to errors and perform maintenance without shutting down. According to representatives of the corporation, the presence of the RAS area is one of the main advantages of Blue Flame, which gives a significant advantage over alternative processors. Such hardware solutions will allow IBM to focus not only on finding errors in systems, but also on collecting information that will protect against future failures.
|
The new flagship of mainframes is the IBM eServer zSeries 900 series of computers, optimized for e-business tasks. It includes 64-bit multiprocessor systems with 64 GB of RAM and I/O and network adapter throughput of 24 and 3 GB/s, respectively. zSeries 900 performance exceeds 2500 MIPS. Their most important component is the 20-processor MCM (Multi-Chip Module) module. Its 16 processors are designed to execute application tasks in SMP mode, and the rest perform such system functions, such as input/output control, error recovery, cryptographic protection.
 |
zSeries 900 system. |
Each system can operate independently or as part of a Parallel Sysplex cluster with other zSeries computers and a number of IBM S/390 systems. The cluster provides high scalability and exceptional availability. Up to 32 zSeries 900 systems can be clustered using Parallel Sysplex technology.
The zSeries 900's 15 logical partitions can run multiple operating systems (z/OS, z/VM and Linux for zSeries) independently of each other, accessing shared system resources.
Fully integrated with zSeries 900 hardware and firmware, 64-bit z/OS with MVS kernel enables Intelligent Resource Director (IRD) technology to automatically prioritize system resources among applications. Broadly speaking, the zSeries IRD combines the benefits of three core technologies: Workload Manager, Logical Partitioning, and Parallel Sysplex.
The system is capable of dynamically allocating processor time, channels, and channel-controller interactions among multiple virtual servers to intelligently serve unpredictable e-business workloads based on business priorities. Sysplex Distributor Function operating system z/OS is another option for intelligently balancing TCP/IP traffic through the Parallel Sysplex. Dynamic Virtual Internet Protocol Addressing (VIPA) support extends the TCP/IP network workload balancing and availability capabilities of Parallel Sysplex.
Hyper Sockets technology allows you to form IP connections between logical partitions and transfer information between them with the speed of RAM. Linux for zSeries supports original applications for this OS. With z/VM, you can run hundreds of applications on a single logical partition, creating multiple virtual servers. Thus, on one system you can run up to 1000 virtual Linux servers in parallel, thus creating an effective “network in one case.”
Readiness
The zSeries family of systems provides the high level of application availability required in a global network environment. Even a single zSeries server can avoid failures and recover from failures, minimizing business process disruptions.
High component reliability and design features help prevent failures and provide fault tolerance, as well as hot maintenance and repair. Enhanced Dynamic Memory Sparing, ESCON port redundancy, hot service/upgrade of Coupling Links, hot service of I/O cards (Concurrent Service for I/O Cards) and automatic switching of service elements (Auto-Switch over for Service) Elements are just some of the built-in features that help reduce both planned and unplanned downtime.
For systems that require higher levels of availability, zSeries servers feature IBM Parallel Sysplex clustering technology. More fast connections provide balanced performance of a powerful zSeries 900 server in a Sysplex cluster. ISC communications provide transfer rates of up to 2 GB/s, and ICB communications provide transfer rates of up to 1 GB/s. At the same time, backward compatibility with similar connections in S/390 systems is maintained.
Another aspect of availability is the continued growth of zSeries, enabled by on-demand storage capacity expansion. zSeries 900 servers can expand storage capacity and create virtual servers without interruption, as well as install FICON, ESCON, OSA-Express ATM and Gigabit and Fast Ethernet and PCI cryptographic coprocessor cards; In addition, memory initiation without system shutdown is supported.
Safety
The built-in security features of z/Architecture give IBM a leadership position in e-business security. Cryptographic CMOS coprocessors comply with the US government information processing standard - FIPS 140-1 level four (the most high level certification of a commercial safety product ever awarded by the US government). These coprocessors are structurally located in a module with a single chip. Moreover, each of them can be serviced independently, which eliminates downtime when repairing a cryptographic crystal.
zSeries 900 servers can support up to 16 on-demand PCICC coprocessors, allowing them to perform more than two thousand SSL operations per second. By combining two types of coprocessors, applications can use industry-leading cryptographic DES standards, Triple DES and RSA to provide scalable security and flexibility while quickly migrating to new standards.
Self-control of computing systems
eServer zSeries computers are designed and built with extensive use of core technologies from the eLiza project aimed at creating self-managing systems. The goal of the project is to create an intelligent, self-managing IT infrastructure that will minimize the complexity of managing it.
The IBM eServer series offers a broad range of new self-managed computing solutions based on four core principles.
Self-configuring
The most important elements of the automatic configuration system are the automatic detection of new hardware resources and the mechanism for dynamic allocation of OS resources. Hardware subsystems and resources can automatically change their own operating parameters both at the system boot stage and during task execution. Reasons for redistributing resources may be hardware failures, errors in firmware, or changes in current optimization parameters. It is also possible to change the configuration of hardware resources at the request of administrators, maintenance personnel, or the program that manages hardware resources.
Automatic recovery (Self-healing)
The self-healing mechanism allows you to immediately detect and localize problems in hardware or firmware, minimizing the possible consequences of failures that can negatively affect the operation of the OS and applications.
Automatic optimization (Self-optimizing)
The built-in optimization engine detects the current load on various hardware resources and automatically changes the configuration of hardware resources to achieve maximum performance.
Self-protecting
This mechanism protects the computing system from internal and external attacks that threaten the integrity and confidentiality of applications and data.
Project eLizaThis project implements the concept of autonomous computing. It serves as the foundation for creating information systems with reduced complexity and cost of operation, use, and administration. The goals of the eLiza project are formulated as self-optimization, self-configuration, self-healing and self-defense. The eLiza project is believed to overcome challenges to on-demand e-business such as management complexity and resource efficiency. According to expert firms, the percentage of efficient use of resources for mainframes averages 40%. This figure may seem low, but for Unix machines it is only 10%, and for Windows machines it is 5%. Mainframes pioneered the challenge of efficient resource utilization, and zSeries and z/OS servers are at the forefront of the eLiza project. However, its results are immediately transferred to other IBM platforms. Thus, today all four IBM server platforms provide logical separation - the ability to present one powerful server as several separate virtual computers, perhaps with different operating environments and different performance and functionality, but with single center management. The eLiza project's computing autonomy goals are combined with Grid computing goals, so these initiatives are being developed in concert. The evolution of the concept of self-managing information systems has led to the emergence of a new strategic initiative of IBM - Autonomic Computing. It is noteworthy that this happened at a time when IBM announced its intention to lead the movement to create a new generation of e-business solutions, bringing together the efforts of various solution providers, the open standards community and users. Besides intelligent functions self-protection and self-healing capabilities being developed through the eLiza project, a new IBM initiative aims to enable dynamic adaptation computing systems to the conditions of business activity of the enterprise. The new name reflects the more universal and profound nature of the concept. The ideas of self-government that form the basis of the eLiza project and which have many loyal supporters receive further development, opening up the prospect of improving business without disrupting established processes and business mechanisms.
|
Little "predator"
Last year, IBM announced the IBM zSeries 800 (previously codenamed Raptor), a new inexpensive system entry-level, which managed to radically change the price characteristics of the mainframe market. The new system is available in several versions: eight models general purpose and the only mainframe of its kind running full Linux. They differ primarily in the number of processors (from one to four) and the amount of RAM (from 8 to 32 GB).
With the release of the zSeries 800, IBM was able to offer the reliability and performance of zSeries technology to customers who previously found mainframes unaffordable. In addition, IBM was the first to implement modern technology Parallel Sysplex clustering on entry-level mainframes. Let us recall that this technology provides virtually zero downtime, high application availability and business reliability by combining multiple mainframes into a network cluster.
Mainframe customers are increasingly adding new Web applications to existing infrastructures to save energy, space and management costs. The zSeries 800 is designed for IBM business partners who require server consolidation options for mid-market customers financial opportunities. The new system eliminates expensive and underutilized server pools made up of Web servers, file servers, print servers and email servers by moving the entire load onto a single mainframe, thereby simplifying administration and reducing costs. Thanks to technology virtual machines The IBM z/VM zSeries 800 system can integrate from 20 to several hundred Sun or Intel servers on one physical platform.
Over the past few years, Linux has gained a strong position in areas such as e-commerce, Web services, email services, application development and testing, and much more. One of the main advantages of Linux is its ability to easily adapt to a variety of hardware platforms and run a variety of programs in distributed computing environments. Thus, users and developers dealing with Linux do not necessarily need to learn different operating systems. Linux provides complete openness and hardware independence. For these reasons, it is widely offered as a reliable, open standards-based, cost-effective alternative to other operating systems such as Unix and Windows. But Linux typically runs on multiple separate servers, which doesn't always provide the performance, flexibility, functionality, integration, and manageability that data centers need. Moreover, the hardware Linux independence may in some cases result in an increase in the cost of technical support.
The cutting-edge zSeries 800 technology provides a cost-effective and flexible environment for application development, testing and operation, application migration from 32-bit to 64-bit platforms, and new e-business workloads.
The zSeries 800 system uses self-healing and self-management technologies implemented in IBM computers, including redundant capacity, Parallel Sysplex clusters, concurrent I/O, and automatic referral to IBM when a system failure is detected. At the same time, IBM announced a special version of the 64-bit z/OS.e operating system, which is intended for running e-business applications, including WebSphere application server, DB2 databases and MQSeries applications.
Florida State University (USA) purchased a zSeries 800 mainframe this year to conduct research in the field of network computing, which turned out to be the thousandth Raptor system sold by IBM since March 2002.
"Tyrannosaurus" on the computer market
In May, IBM unveiled its most powerful enterprise-scale system to date - latest model zSeries family. The new mainframe is officially called IBM eServer zSeries 990, and its code name is T-Rex, "Tyrannosaurus Rex". Well, the executives of IBM Corporation cannot be denied humor.
Such systems are intended for companies in the financial sector and other industries that require maximum fault tolerance, information security and good computing capabilities. The cost of the new IBM eServer zSeries 990 starts at $1 million. The new system is the result of four years of work by 1.2 thousand IBM developers. Investments in the development of Tyrannosaurus, according to IBM representatives, amounted to about $1 billion. However, the system is worth it.
 |
"Tyrannosaurus" - zSeries 990. |
The zSeries 990 is considered IBM's most powerful and scalable mainframe in their 40-year history. This server has twice the virtualization capabilities and the ability to handle approximately three times more work than the zSeries 900. New design allows customers to increase power without shutting down the system, and a significantly simplified product structure reduces the number of mainframe models from 42 to 4.
The system sets a new standard for enterprise computing in the following areas.
Virtualization. The zSeries 990 architecture can support hundreds or even thousands of virtual Linux servers in a single chassis. This is roughly equivalent to an entire data center housed in a single server the size of a refrigerator.
Automation. The zSeries 990 features IRD technology that dynamically allocates system resources to the workloads that need them, based on the priorities and goals set by the user.
Scalability. zSeries 990 offers ample opportunities both vertical and horizontal scaling. It can scale to handle 450 million e-business transactions per day or manage hundreds of virtual Linux servers. The zSeries 990 cluster system can handle 13 billion transactions per day, which exceeds the average weekly volume of the New York Stock Exchange. In addition, the zSeries 990 has the ability to add or disable additional computing power, allowing customers to temporarily activate server processors during peak periods.
Safety. The IBM mainframe makes it virtually impossible to break into a system, making it one of the most secure servers on the market. The zSeries 990 system with 16 processors can securely process up to 11 thousand transactions per second.
Reliability. Mainframe reliability is measured not in days or weeks, but in decades, with zSeries cluster system availability reaching 99.999%, or less than 5 minutes of downtime per year.
Companies that only occasionally need such powerful computing resources do not necessarily need to buy an entire machine. The new mainframe will serve as the foundation for multi-platform centers that enable customers of all sizes to deliver on-demand computing power in new ways. IBM Global Services will offer mainframe capabilities using Utility Management Infrastructure (UMI) technology, providing customers with a fully integrated infrastructure that includes processors, storage, networking and mainframe middleware.
For the first time, with On/Off Capacity on Demand technology, mainframe users can activate additional capacity within seconds when needed and turn it off when no longer needed. This provides high flexibility, helping to respond to daily or seasonal surges in activity.
The 32-processor zSeries 990 system has a speed of 9000 MIPS. This model contains two times more processors, and delivers triple the performance of the zSeries 900, so customers can scale a server from one to 32 processors without taking the system down.
The zSeries 990 supports up to 30 logical partitions (LPARs), twice the capacity of the zSeries 900. With z/VM 4.4, you can quickly create and efficiently manage hundreds of virtual Linux servers in a single physical enclosure. IBM's advanced virtualization technologies make the zSeries 990 a good platform for consolidation when you need to reduce the cost of server farms and management costs.
Breaking its own security record, the new 16-processor zSeries 990 can process 11,000 transactions per second over Secure Sockets Layer (SSL) (a 57% increase over the 16-processor zSeries 900). Handshake handshake via SSL protocol(what causes the locked padlock icon to appear in the bottom bar of browsers) is very important for e-business transactions and allows companies to more securely process orders online. The ability to process more SSL transactions means organizations can serve more customers and achieve more sales in less time.
For those customers who need great ability Connectivity for new e-business workloads running on the mainframe, the new zSeries 990 system provides up to 512 I/O channels, doubling the capabilities of its predecessor. In addition, up to 16 HiperSockets are now available, providing high-speed TCP/IP connectivity between virtual servers within a single zSeries 990 system, quadruple the capabilities of the zSeries 900. IBM has also introduced a new technology called Logical Channel Subsystem that will make it easier to customers to consolidate multiple mainframes into a single zSeries 990 system.
In addition, the zSeries 990 has four times the memory capacity of the zSeries 900 - 256 GB versus 64 GB.
The heart of the zSeries 990 is the multi-chip MCM module. The newly designed 3.7 x 3.7 x 0.75 inch module, which fits in the palm of your hand, contains 16 dies mounted on 101 layers of ceramic glass and connected to more than 5,000 pins via 500 m of conductors. The new MCM is 50% smaller and provides the zSeries 990 with nearly triple the processing power of the zSeries 900 in the same compact size. The module uses copper interconnect and silicon-on-insulator technologies. It contains over 3.2 billion transistors. This technology provides significant benefits in performance, power consumption and reliability. Besides, new design The superscalar microprocessor in the zSeries 990 helps achieve up to 60% performance gains for Linux, e-business applications and traditional workloads.
On/Off Capacity on Demand features will be available on the zSeries 990 this September. Secure key cryptography, support for 30 logical partitions, and 512 I/O on z/OS will be available in October.
