1.0 What is a Mainframe?

Este vídeo pertenece al curso Mainframe - Crucial Role in Modern Enterprise Computing de openHPI. ¿Quiere ver más?

1.0 What is a Mainframe?

Duración: aproximada 19 minutes

An error occurred while loading the video player, or it takes a long time to initialize. You can try clearing your browser cache. Please try again later and contact the helpdesk if the problem persists.

Scroll to current position

00:00Welcome to the introduction to the Mainframe Architecture and the Mainframe Paradigm.
00:06My name is Prof. Dr. Philip Brune from the Neu-Ulm University of Applied Sciences.
00:11When we think about the mainframe as an idea, not only as a certain
00:16hardware, then we should think about the difference between scale
00:20up and scale out.
00:23The mainframe paradigm means that we have a classical system,
00:29which has multiple CPUs and the CPUs are
00:33symmetrically accessing the main memory
00:38and have the same rights to access the main memory.
00:40As you can see on the left hand side. So, this is a classical mainframe architecture.
00:43In the mainframe architecture we have one big server system
00:48which has, of course, CPUs and RAM and storage disks, but
00:53all the CPUs share a common main memory, a common ram,
00:58and can access this with equal rights.
01:02This means that in such a system, parallel processes which we need, for
01:07example, for transaction processing are able to know about the status of
01:11other parallel running processes in the same system.
01:14Since they share the RAM, they share a common memory to
01:19exchange information in a very efficient way inside the machine.
01:24On the other hand, when we look in the research, in computer science research and
01:29also of course, on I.T. practice today, the dominating paradigm
01:33is the cloud or grid paradigm.
01:37That means that we have in such an environment we would have a big number
01:42of servers which are, every server itself
01:47is rather simple compared to, for example, a mainframe.
01:52So, I illustrated this here by just giving them one CPU, of course,
01:56in reality they might have also a few CPUs, but the servers
02:01are rather simple, they are commodity hardware and they are connected to each other using
02:06a network. OK, so they communicate over a network and this
02:10could be a high speed network, but it's still a network outside their individual boxes
02:15and they might share as well a common storage with the access of network
02:20components as well.
02:22This paradigm is the typical paradigm that we have in all the cloud and Internet
02:27providers like Google, Amazon and so on.
02:30You can, even as a customer, get rather cheap access
02:35to some of these virtual service that then might be connected to such kind of cluster.
02:39In this case, we have the so-called scale out.
02:43That means if you want to increase performance and you want to increase the capacity of
02:47the system, you can just add further further
02:52servers to the network and
02:57an increase of computing capacity is achieved by adding more servers to
03:01the network. This is called scale-out.
03:05In contradiction or in an opposite is the scale-up in which we would have
03:10the mainframe paradigm. Scale-up means if you want to increase the performance of the
03:14system, you can put in more ram, you could put in more CPUs in this one big box.
03:19And this increasing number of CPUs may communicate internally.
03:25And the question, of course, is now why, for example, Google doesn't need a mainframe
03:30or Amazon or all the other Internet providers?
03:32And the answer to that is Amazon and Google and all the Internet companies,
03:37they have their core applications written in a way that they perform well on such
03:42kind of distributed grid or cloud cluster architecture.
03:47But if you look at businesses where you need transaction security, where they need
03:54the possibility to execute a business transaction in a safe way, then
03:59this requires that parallel running transactions know the state of other transactions
04:03and there is the need to share information and this is, of course, achieved much better
04:08in an architecture that have a shared main memory.
04:11And this is the important concept of the mainframe, for
04:15example. So mainframe like architectures are scale-up based
04:20scalable architectures.
04:23And this is what we want to sort of bit look
04:27deeper into in the following.
04:29When we talk about the mainframe paradigm, we mean by mainframe, we mean
04:34a high end server supporting concurrent operation of
04:38a large number of users, a large number of different parallel
04:43running transactions as well as different operating systems.
04:48This server usually has the capacity to host thousands
04:53of users and applications simultaneously.
04:56So, it's a big centralized machine and it's usually used as
05:00a central data store of an enterprise data center as well.
05:05There are all other applications to access the core data of the company.
05:08For example, in automotive industry, we have the product data, for example, in
05:13this kind of core system. And so it's a server
05:18with a maximum number of hardware
05:22capacity and since it's a central, crucial and critical system, of course,
05:27the server needs to have the maximum possible availability, guaranteed
05:32availability and security. So these are the the core properties, what we classify
05:37as a mainframe system.
05:40Of course, when we talk today about mainframes, we usually refer to
05:44a certain architecture, certain hardware, and this hardware architecture
05:49originates from the original IBM s/360 generation
05:54of computers, which was first presented in 1964.
05:59And the 360 in the name stands for 360
06:03degrees. So, it was considered a general purpose machine for different kinds of
06:07customers, for different kind of tasks and workloads,
06:12scientific computing, business computing and so on, which before that was usually
06:17separated and addressed by differently individually created computers.
06:21And over the years, this original s/360
06:26architecture evolved and it was renamed and it evolved to the
06:30what we call today, the z Systems architecture.
06:33So, this is a long term history of a certain CPU
06:38and hardware architecture, meaning that these machines have a specific instructions
06:43tset and a specific CPU architecture and system architecture.
06:48In the past, different companies offered compatible systems for
06:52these kind of machines since IBM released the specification
06:57of the hardware in a document.
06:59And so it was possible to build custom compatible systems
07:04by other vendors. But during the decades, all the others
07:09sort of went out of the market except Fujitsu.
07:13Fujitsu, formerly Siemens Fujitsu, still is selling
07:18mainframe systems that are originating on the same architecture.
07:22They are based on the s/390 generation of the mainframes.
07:26And we will learn also a little bit later about that.
07:30So if you look at the picture, you see that from the original s/360
07:35to the more modern mainframes.
07:37It's a long way and these systems don't look very similar but still, the
07:41modern mainframe is fully backward compatible to the
07:46original one. So, you could run in principle, a binary program written for the original
07:51s/360 machines on a modern mainframe as well, if you need to
07:55do that. So, this is the hardware architecture, it has a long standing,
08:00backward compatibility of all existing architectures today.
08:08So, what is typical for mainframe workloads, so mainframes are
08:13a high end technology and high end technology means that it's also a
08:18very costly technology
08:23and this means that it's, of course, used for only the important
08:27applications. So, all or many large mission critical applications,
08:32traditional enterprises run on mainframes.
08:36The mainframe guarantees the highest availability of very high or the possibility
08:41to process large data volumes, high data volumes, since it is an
08:45architecture that was optimized for our input output operations - IO operations
08:50from the beginning, it enables, it
08:55features a very high transaction throughput.
08:58So, the mainframe itself is not necessarily mainly
09:03a CPU centric architecture, which is very focusing on quick
09:08calculations, for example, but it's an architecture that is balanced in a way
09:12to achieve the highest possible throughput of parallel running transactions and this
09:17involves more than CPU, it's the CPU, the memory, and IO as well.
09:21And of course, it supports the special security
09:26requirements, special security features like encryption and other things in hardware.
09:31Examples for typical mainframe workloads are, of course, the traditional
09:36mainframe applications are bookkeeping systems, as we know from aircraft
09:41industry, planes, bookings, travel bookings, financial
09:46bookings, financial industry and all these industries, you have
09:51the need to do a very fast and consistent bookings.
09:56On the other hand, we have the mainframe as a centralized data
10:01store, a centralized database server for all important enterprise
10:05data. As I said, for example, in the car manufacturers,
10:10the mainframe can also run SAP, ERP or CRM application servers,
10:15which then could be or is is a business critical
10:20central application for many companies as well and here typically
10:24for the application servers, Linux or the z systems, for example, would be used.
10:29Data warehousing, data mining, artificial intelligence, big data analytics are typical
10:34workloads that could benefit from the mainframe architecture.
10:38More modern workloads in recent years, cloud computing, cloud hosting,
10:43the possibility to run the backends for mobile services, highly scalable
10:48backends for mobile services, for example, in retail business are
10:53workloads that could run on the mainframe.
10:55And of course, blockchain is also an interesting hot topic at the moment because block
11:00chain applications benefit a lot from cryptographic features which the mainframe
11:04architecture supports in hardware.
11:08If you look at the typical architecture of a modern mainframe
11:12system, then you can see that it is designed for the highest possible availability,
11:17which is crucial for critical, mission critical applications.
11:22And so the hardware is highly redundant, everything's redundant in the mainframe box -
11:26the CPU, the RAM, and it allows to configure
11:30the resources at runtime to assign resources and move
11:35resources around between different parallel running applications
11:40at runtime without, for example, shutting down the system.
11:44And by coupling multiple mainframe boxes, which
11:48are called CEC, to a so-called Parallel
11:53Sysplex, which is kind of a very sophisticated cluster mechanism.
11:57It allows even to have systems that achieve the highest possible performance
12:03availability of
12:0799.999 Percent. So the "five nines" and this is the maximum you could get
12:13because these systems allow, for example, to move around workloads at runtime without
12:17shutting them down if one of the cluster partners breaks down.
12:21The CPU architecture of the modern z14
12:26is the most recent generation of the mainframes is a 64
12:31bit architecture CPU, one CPU has 10 cores and
12:35these cores are called PU - processing units and it's
12:40possible to have about 170 PUs available to the customer.
12:44There are some more for internal administration things.
12:48This generation runs at 5.2GHz
12:53clock speed and it still is a complex
12:58instruction set computer. So, the mainframe architecture of the
13:02binary instructions are complex instructions.
13:07So, the mainframe processor has a very, very
13:11large set of machine instructions with very special instructions for encryption,
13:16for binary coded decimal arithmetics and for many other special things that we will
13:22go bit deeper into in the following units.
13:26One important feature of the mainframe is that you can partition the system and you can
13:30partition it in different sort of virtual machines
13:35and these different partitions then can run different operating systems.
13:39So every partition can run another operating system and serve as a
13:44sort of virtual server. And resources may be allocated dynamically between these
13:49partitions at runtime. This is a virtualization mechanism that's built in the hardware
13:53and in the firmware of the system and allow us to
13:57divide a physical machine into multiple servers, even before
14:02you install an operating system.
14:06One important thing is that how we use a mainframe, traditional mainframe
14:10customers typically think of mainframe using a terminal,
14:15a terminal, which is called the 3270
14:19terminal, which you can see on the left hand side.
14:21This is a classical terminal, physical device, sort of a dumb
14:26input output device that is connected to the machine, to the central machine
14:31where the user can just enter data and get data.
14:34And this in the past used to be a dedicated hardware.
14:38So, it's really the device that you can see here.
14:41But of course, today we don't have these devices anymore, but we use software
14:45emulation of this classical text-based screens that
14:50are used to interact with the system.
14:52So, that was the classical approach and for many, many, many decades
14:57sort of defined for the users, what would a mainframe look like.
15:02But, of course, today, users are not mainly interacting with these kind of terminals with
15:06the system anymore. But since the high end server, of course, we
15:11have other features, usually from our end users perspective
15:15the mainframe is hidden behind web frontends, behind mobile apps
15:20that call the mainframe transactions, for example, using web services.
15:24And for the developers, typically we have modern tools like Eclipse to program and
15:30of course, all the other tools that a developer developers are used, thw FTP, SSH login,
15:34and so on. And so today, this classical
15:39term and a user interface is mainly used by administrators and people who are directly
15:44configuring, for example, the system.
15:48Finally, a short overview about operating systems, when we talk about the mainframe and
15:52thinking, OK, what kind of operating systems are running there, you will see that there
15:56are other operating systems than Linux and Windows that you might know from the
16:01your daily life. So, on the on the mainframe, we have operating systems that
16:06have a long time history and these are IBM specific operating systems
16:10but also, of course, we have Linux on the mainframe, which is very important nowadays
16:15and of course, there is another operating system for the Fujitsu, formerly
16:19Siemens line of mainframes, which is called year BS2000
16:25and you will also get a short introduction to that during the course.
16:28The traditional IBM mainframe operating systems
16:32start with so-called MVS, the MVS family of systems,
16:37which today it is called z/Os and then, of course, we have
16:42something that used to be DOS/VSE, which today is called the z/VSE,
16:47and we have the z/VM.
16:49The z/VM is a hypervisor virtualization environment that we
16:54also have a deeper look on later.
16:56OK, so these are the three families of systems, over the years the
17:01features and also the name has changed, but it's more or less the same line of
17:05operating system until today.
17:07The most important one is z/Os, this is flagship operating system for the mainframe,
17:12which features all the highly specialized features
17:16like Parallel Sysplex and all the other things.
17:20And besides this, of course, today, mainly Linux is a very important.
17:24VSE declined over the years, nevertheless, there are still some installations
17:29running, but it's not as important.
17:31It used to be the sort of the small brother of MVS for, let's say, smaller companies,
17:36for example. And VM has become became
17:40more important in recent years, as well as the virtualization environment to host
17:45multiple Linux instances. So this is the family and here you can see also
17:50the evolution of the hardware starting from 16 or 24 bit for the original
17:54s/360 until the 64 bit today.
17:58And this would be a short overview on the platform and the architecture and the history.
18:03And in the following chapters of this course, we will now look at all these different
18:08technologies in more detail.
18:11Thank you very much.