1.1 The Mainframe - A Server OS Perspective

This video belongs to the openHPI course Mainframe - Crucial Role in Modern Enterprise Computing. Do you want to see more?

Enroll yourself for free

1.1 The Mainframe - A Server OS Perspective

Time effort: approx. 25 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

00:00Yeah, welcome to the mainframe server OS perspective.
00:06As an introduction to the mainframe course, we just want to talk
00:10a little bit about where mainframe computing comes from,
00:15how server computing is different from traditional desktop computing.
00:19What are the key optimization goals and the criteria for server systems?
00:24And there are attributes like capacity and scalability,
00:29integrity, security, the availability server
00:34computers typically have access to large amounts of data.
00:39They cannot be run without sophisticated systems management facilities.
00:44And oftentimes people talk about autonomic capabilities and mean that
00:49the systems maintain themselves to some extent.
00:56So just the definition with capacity.
00:59We talk about the potential for holding and storing
01:05huge amounts of data disk, for instance,
01:10or also long term storage like tape
01:15with capacity. People also address the power to produce and perform
01:22tasks, sufficient computing capacity, for instance, to run the programs that
01:26process the data or also sufficient capacity to store the
01:31data on a single server and sometimes
01:35also on distributed and dispersed systems.
01:41With a service level agreement, we address the fact that computer systems
01:46cannot themselves solve all the problems on the planet, but
01:51we have agreements between service providers and recipients, business
01:55units, server owners.
01:58These are different terms for the same fact.
02:01And typically it is that SLA describes a baseline
02:06for capacity demands. For instance, we want to say that 99 percent
02:11or 95 percent of ATM transactions are to be completed in less
02:16than a second, or that 90 percent of the daily reports are completed
02:20by 6:00 a.m.
02:23It is beyond the scope of the technical solution to define what
02:28happens if SLAs are not met.
02:31Oftentimes the lawyers take over in that moment.
02:36Another aspect that is typical for server systems is that we are
02:41planning for downtime, we need to address the fact that systems
02:45are out and that there are outages.
02:49These may be planned or unplanned, for instance, for software maintenance
02:54to implement new levels of features or to fix problems.
03:00And oftentimes the continuity of a service can be provided
03:05by sharing resources for critical workloads across several machines.
03:11Planning for downtime also means that it makes a
03:16big difference whether a system comes down planned or unplanned, that you
03:20need to plan ahead, that you need to schedule maintenance and to force.
03:26In case we want to provide continuity of service in event
03:31of failures, we need to have two or more physical machines.
03:34Obviously the scalability, we want to address the ability of a system
03:38to continue to function if it's changed in
03:43size or volume. So scalability means that we can add
03:48additional hardware, that we can add additional software, or that we can change
03:52from a single system to a distributed system.
03:56It also addresses the ability to retain performance levels when adding
04:01additional resources such as processors or memory or storage.
04:05With a scalable system, we can adapt to work within
04:10smaller networks to perform tasks of varying complexity.
04:14So we basically can adapt ourselves to changing requirements.
04:21Integrity and security addressed the question to not only
04:26be operational, but operate on valid data,
04:30which means we have to implement data security as protection against unauthorized
04:35access, against transfer or denial of service, against
04:40modification or destruction.
04:41And these attempts can be done accidentally or intentionally.
04:46Oftentimes, the accidental data security
04:50problems are due to the most difficult ones to address.
04:57In order to talk about terms like integrity and security, you have to have
05:02well-defined security objectives, which means you have to address the
05:06job of a security administrator.
05:08You need to talk about security policies, the security measures and so forth.
05:13And also the system needs to provide interfaces to
05:17enforce authority rules.
05:22RAS, or reliability, availability, serviceability,
05:27those three terms address the stability
05:32of a system and can be seen as key features for data
05:37processing, which means that in the system
05:41architecture, we place a high priority on systems that
05:46are available and in service all the time.
05:50So RAS is the central design feature for many computer systems,
05:55and RAS talks about systems, which means it doesn't talk about hardware alone.
05:59It doesn't talk about the OS only or the application, but it talks
06:04about the entire stack.
06:06And this reliability will oftentimes mean that
06:10systems have to have extensive self-checking and self-recovery.
06:15Availability means that a system can recover from failed components without
06:21impacting the rest of the warning system. That means hardware recovery,
06:25for instance, by replacing components with spares and software recovery,
06:31one has to understand that there's a big difference between hardware and software because
06:34hardware units that are replicated can be seen as independent units,
06:39whereas replicating the same software bug just creates the
06:44same problem.
06:46People talk about mean time between failures as a measure that describes
06:51the availability of a computer system, and with serviceability,
06:57we mean that a system can basically
07:01postmortem determine why a failure occurred.
07:05And that way we can replace elements like hardware
07:10and software elements, and we have as little impact on the operational system
07:15as possible. Serviceability also means that we have well-defined units of
07:19replacement that might be hardware or software components.
07:26We already mentioned that for subway systems it's
07:30typical that we have access to large amounts of data, which means we need to store
07:35the data and we need to process the data.
07:37And the good old tape is still available
07:42and denserthan ever before, basically to store
07:47data, huge amounts of data.
07:48People oftentimes talk about structured and unstructured data
07:53and talk about the massive growth of
07:58unstructured data in server computing.
08:02We often talk about systems of record
08:07and systems that do transaction processing and systems that typically operate
08:12on structured data rather than unstructured data.
08:18When processing data, we need to move them from disk to
08:23memory, and for that we are going to have
08:28many discs, because the large discs don't give us a good i/o
08:33bandwidth. However, a huge number of smaller discs, this is typical
08:37for server systems and the bandwidth is crucial.
08:42When it comes to systems management, we talk about a collection of tools
08:47and disciplines to monitor and control systems behavior, which means performance
08:52monitoring, workload, configuration operations, also problem management, looking
08:56at the network, the storage, security and also change management.
09:03All these operations have to be defined by clear processes
09:07and they are either performed by the operating system
09:12themselves or appropriate subsystems.
09:16This is also the market for specialized tools from various software
09:21companies in order to achieve high availability,
09:26not only the hardware infrastructure and the software components are
09:30critical, but also the operational procedures and good systems management.
09:34And in the end plays a vital role for achieving high availability.
09:42I already mentioned that systems management is a complex task and
09:49the desire is high to have systems manage themselves.
09:53Autonomic capabilities mean exactly that.
09:57So in analogy to the autonomic
10:02central nervous system in the human body.
10:06We want to have the systems adjust to many situations automatically without
10:11external help.
10:13And this means fewer people, fewer human maintenance
10:18and basically quicker fixes and quicker debugging
10:23of problems in computer systems.
10:29So these are targets and design goals basically
10:34for server systems.
10:36And now if we look at the time axis and ask ourselves
10:41how did systems evolve and what
10:46is the status, where are we and what can we learn from history, then
10:51one comes up with business picture similar to that one here.
10:55We started out a long time ago with big computers, expensive
11:00computers. People were not supposed to touch the computers and mainframes fall
11:05exactly in that class. The hardware was king and it was from one big
11:10vendor or typically a vendor
11:15like IBM or a couple of others.
11:17But in all these cases, the hardware was closed and proprietary.
11:23In the nineteen-eighties there was the advent of the PC.
11:28With PCs, software became crucial and software was king basically.
11:33And Microsoft for over ruled the world, if you will,
11:38because Microsoft created the understanding of stanard of a software
11:43and people would send word documents and email something
11:47that was completely unacceptable a couple of years before.
11:51So. There was the move from expensive centralized
11:56hardware to personal computers, however, that move came
12:00at a price of reliability.
12:04People were assumed to reboot the computer very often.
12:07In fact, there was a time when logging onto computers meant to just
12:12reboot the computer. And now we are trying
12:16to mix both worlds, the expensive but reliable mainframe
12:21world, and the cheap but unreliable PC world
12:26into something that is best of breed, basically.
12:30And when we look at the cloud and we see today a mixture, we have clusters
12:35of workstations, we have people participating
12:41at home in distributed applications, helping out, giving
12:45away resources. But we also have the centralized systems
12:49like the Linux one is the latest addition in the mainframe scene
12:55that contribute to cloud computing,
13:00The big difference now via hardware had
13:04its time as the central component and software was most important
13:09on a single PC in the 80s. Now we have to middleware the software layer or an
13:13abstract layer of abstraction that sits between systems and distributed
13:18systems
13:20technology basically, that is most important.
13:22And we rely on open standards like Web services.
13:26It's not clear who is going to rule the world.
13:30Looks like neither Microsoft nor IBM will be in that role,
13:35but they will still play an important role.
13:39And for us, when looking at these systems, the question is what can
13:44we take away or can we reinvent reuse
13:49as ideas that came from the old times like Vista
13:54or 360 and the mainframe operating systems up to the US?
13:59We had things like the resource access control framework or the job entry subsystem.
14:05We have licensing models that might be appropriate to cloud as well.
14:09With systems like VMS, the system from Digital Equipment
14:14Corporation. There were solutions like clustering and failover
14:18and the versioning file system.
14:20So this is something that has completely disappeared from most systems nowadays.
14:25And if you look at IBM's AIX Unix, then
14:30they really introduced journalling filesystems.
14:32They have high availability, availability solutions.
14:36They have logical partitioning.
14:38These are all features that need to be carried over to the world of cloud computing.
14:42And even if one is mostly interested in,
14:47say, Web services, it's still worthwhile looking at these
14:52older systems because they introduce the technologies that are sought after
14:56today.
14:59Another story is interesting, and this is the story of looking at history
15:04of operating systems.
15:06And in that graph, there is a number of
15:11lineups of systems like the MULTICS and UNIX line
15:16in the very center where systems
15:20created the prototypes for generations of operating
15:25systems, like the MULTICS with the process model led to
15:30UNIX and then later on to different variants like the AT&T and the
15:34Berkeley distribution.
15:36The same with the OS 360 from IBM, which was
15:42the initial desired operating system for the
15:46mainframe.
15:47And in fact the OS 360 was
15:52also the key for this
15:56software engineering story.
15:59The mythical Man Month by Fred Brooks.
16:04And this is the story about creating a software system that was
16:09late, that was too big, that would not fit the
16:13computer system when it was released and that was
16:18difficult to maintain. There was written with inappropriate tools
16:23that had so many box that and showed
16:28the behavior that fixing box would introduce more new box than
16:34old box were removed.
16:37So from a software engineering standpoint, that system was a big challenge and
16:42it succeeded in the market.
16:44However, it triggered research in a field which is termed software
16:49engineering nowadays and which in the very end led to
16:54institutes like the Hasso Plattner Institute, where we do research in the field of
16:58how software engineering for big systems may
17:03work. And we all know processes like agile development
17:08and we know the DevOp.
17:10This all originated in one system, which was the OS 360.
17:16And of course, all 360 technology
17:21has been carried over several to several generations, enterprise systems,
17:25architecture and the 390 up to the system that we know
17:30is the OS nowadays.
17:33This is also interesting with mainframe,
17:38the IBM introduced not only the term of computer architecture.
17:45This is the original flier from April
17:507, 1964, when people talked about
17:56computers that can fulfill all computing needs, the 360
18:01degree picture, the windows.
18:03But also when people talked about different
18:08models of computers that were able to run the same program.
18:12So something that is taken for given nowadays that
18:17we have upward compatability that the next
18:21new computer will be able to run the same programs like today's computer
18:26that was not given before 64.
18:29So this is the start of a long
18:34story.
18:34And by the time people talked about IBM and the Seven Dwarfs because IBM was producing
18:39approximately 70 per cent of our computers in 64.
18:43However, the major technological
18:50development in the 60s, there was the IBM system 360.
18:53And what they introduced early on was.
18:57Not only the concept of running an operating system by that time,
19:02not yet called the OS. But still operating system on a partition
19:07and a system. But they also introduce the idea of using virtualization,
19:12running an operating system in their cases, even as in
19:16the virtualization layer. And in fact, there are two levels of socialization.
19:21This partitioning of is logical partitions policy and the introduction
19:26of the hypervisor. There was a basic feature of the system.
19:31Later on, the system was able to run Linux and Linux
19:36unspotted, basically. And this Eiffel's we talk about integrated facilities
19:40for Linux, basically CPUs that have special
19:45special rules attached to it.
19:50Why is virtualization important and what does it really
19:54mean from a historical standpoint this year, hardware virtualization,
19:59where we first virtualize all the interfaces, which means
20:03we have a virtual machine monitor, we have our virtual lifestyle interfaces,
20:08we have social memory, we have processes as models for
20:13virtualizing computer hardware.
20:17Virtualization is also great means for isolation and for running different
20:22operating systems, different personalities on the same metal, on the same
20:27computer. There are other means of virtualization besides hardware
20:31virtualization. There's also processor virtualization.
20:35Early example, that was DePasquale compiler that would translate
20:40not into machine code but into peacoat, a pseudo pseudocode
20:44for virtual machine.
20:47And if you look at Java nowadays, it still translates into bytecode, which is
20:51the machine code of an.
20:55Non-existent computer, basically, and
21:00for efficiency reasons, we oftentimes now see instructions said racialization,
21:05like the binary translation that we find and transmit cruiser for energy efficiency,
21:10or that we find an Rosero where, for instance, Apple made
21:15a move from power architecture to intel architecture and was able to run power
21:19binaries on on the new systems.
21:25And another concept that was first coined on
21:29the mainframe and introduced in the market and became widespread use, nowadays,
21:34that's the concept of core processors. So if you look at a computer at home, then you
21:38find the GPU and there's the concept of
21:43the General-Purpose Graphical Processing Unit, which is rather a simply computer
21:48computer in the box, which is an add on to the main
21:53system. And this is a
21:57story that started a long time ago with introducing cryptographic
22:02units into the mainframe processor and introducing specialized
22:08up codes into the into the machine language of the system.
22:14So this open s.L and Vesicular, that's the attempt to unify
22:19both worlds, the you and to GP world, and we will see
22:24more specialized compute units in the future.
22:27And the systems, the most recent, the fourteeners,
22:32again making a strong move in that direction.
22:36And additional examples, these are, for instance, the eyehole channels, the question
22:41of lockstep execution for high availability,
22:45the question of redundancy or how you do monitoring or billing
22:50so many fields. Fair mainframes introduce concepts
22:55that are now widespread use in the server marketplace.
23:00During the course, we are going to address additional aspects
23:05and we start with architecture, the instructions said
23:10the virtualization approach.
23:13If you talk about the enterprise hardware and give an overview of the
23:18architecture and the subsystems there, talk about the availability
23:22and wrasse features in general monitoring.
23:28There will be discussion of operating systems and the evolution
23:32we talk about Linux on systems, the programing interfaces
23:37like the transaction management, like kicks the DB to database
23:42and the like, and in the end, the mainframe.
23:47That started many movies and many trends in our industry.
23:52In the end, it's not much different from the programing
23:57standpoint than a regular system.
24:00You can have your Linux running, you can have your eclipse programing environment
24:05running, and to the programmer should be able to
24:10make steps, first steps on a mainframe, even even
24:14relatively quickly, even if it's a new platform to you.
24:19So this is my introductory message to the mainframe, Mücke.
24:23And I just want to say welcome to openHPI and to
24:28our corsia. Thank you.