Este vídeo pertenece al curso Mainframe - Crucial Role in Modern Enterprise Computing de openHPI. ¿Quiere ver más?
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
- 00:00And another question I think many mainstream people who use Sysplex
- 00:05and Parallel Sysplex know the coupling facility. You contributed to a
- 00:10redbook coupling facility beyond the basics, and my question is
- 00:15what is a coupling facility data table and what for they use it?
- 00:21That's a very good question. Coupling facility data table is a centralized repository
- 00:28for multiple types of purposes. It's
- 00:33primary purpose is to be similar to that of a VSAM file an
- 00:37in memory VSAM file. The definition in CICS is a typical or standard
- 00:43audio definition for a file, and in the file you can specify
- 00:47if it's native VSAM
- 00:49or local VSAM. An LSR or Local Short Resource VSAM or RLS-
- 00:55for a Record Level Sharing VSAM or SDT which is a Shared Data
- 00:59Table which is an in memory for a single audipspace or a CF.
- 01:04The CF type of file is a coupling facility data table.
- 01:08And what it does, it allows you to have a shared repository in memory
- 01:12that all servers in every el part across the plex
- 01:16can use. Okay so instead of having a follow only region that you
- 01:21would have to for a shared file were instead of VSAM RLS which
- 01:24is a persistent VSAM file, you can have an actually in memory.
- 01:27So we have a variety of ways that we use it. We use it as a
- 01:30scratchpad. We use it for passing data back and forth between a
- 01:35parent and child application. Our primary use for is in our
- 01:40caching solution- we call it Z Cache or out on good habits ZECS for enterprise
- 01:44caching services. So we use the
- 01:49CFDT as our in-memory type of cache. Some things are better
- 01:53suited for in-memory instead of persistent or some things are better persistent.
- 01:57So you have to, based upon what your requirements are, you still
- 02:00define it as a file in CICS but based upon the
- 02:05internal setting it's CFDT or local VSAM. Now here's the powerful part about that.
- 02:09So CICS is an app server, a transaction router, a web server, whatever
- 02:15you want it to be. It's multi purpose, supports multiple languages.
- 02:18I believe it's the the most robust app server on the market
- 02:23today, because it supports all these different languages and features.
- 02:26It provides these really simple to use APIs.
- 02:29So in our caching service, we could define for example, Wolfram
- 02:33you create a instance of your ZCache service, Richard creates
- 02:36his own versions of that service. Your requirements need it to be
- 02:40persistent to disk, so when we provision it,
- 02:43we would create is as VSAM RLS.
- 02:46And Richard, he needs is non persistent therefore, we put his in
- 02:50memory as a CFDT. The program does not change. The ZCache code
- 02:55does exact same exe css read file write file based on the
- 02:59audio definition needed right to VSAM RLS, which is
- 03:02the CFDT. So here again, you can have one code base with one API,
- 03:07but yet two or multiple different repositories based
- 03:11upon the needs of each individual use of that service, ok.
- 03:17Do you have any outages during the last years?
- 03:23Well that's a very good question. So we have a
- 03:27two dedicated parallel Sysplexes for our cloud services and
- 03:32for example in our zcache or Zram or our key-value database,
- 03:37you can configure each instance to be either active single
- 03:40in a single parallel sysplex, active active across two parallel
- 03:44Sysplex were active, or standby across two with asynchronous replication.
- 03:49Our first user of Zcache was an application called retail link and
- 03:56it stores state information in our caching service and they required
- 04:02active active across both parallel Sysplexes because their
- 04:05applications lived in two different data centers.
- 04:08They average about fifteen million hits per day.
- 04:10They've been in production since twenty thirteen, so
- 04:14six years it's about thirty two billion hits that they've
- 04:18processed to our z cache. Thirty two billion hits,
- 04:23zero failures. Zero failures, I don't know you can find many
- 04:26products that can do that. But here's one something interesting I
- 04:29remember when it was Rich a couple years, back we did take a
- 04:33brief outage on one of the Parallel Sysplexes. I don't remember what the details were,
- 04:38however because we did active active replication
- 04:41for within our Z Cache service, the retailing customers didn't experience the outage,
- 04:46because it would go ahead and go to the active system,
- 04:49and then when the other system came back active we would go
- 04:52ahead and copy and replicate and then the data would be active on both sides.
- 04:55So here again even if there was an outage in one of the Sysplexes, the
- 04:59way we designed our services to run active active or active standby,
- 05:03none of the applications actually experienced any issues.
- 05:08Great. Really, really impressing your numbers and the last question -
- 05:14did your company think about all the native platforms for processing your transaction workload?
- 05:21Yes. A couple things are first I want to address- a question
- 05:25or something that we hear a lot, that's about cloud first of
- 05:29all I'd like to say that cloud is not a where it's a how. Cloud
- 05:32does not mean a remote service, remote host, CLoud can be local, can be raw,
- 05:36can be x y or z platform. It doesn't matter. Cloud is a delivery
- 05:39model. It's a how not a where. So when you hear about- let's put
- 05:42things in the cloud there is no 'the cloud'.
- 05:46Cloud is a delivery model ok and it's really important for people to understand that.
- 05:50That being the case we have a lot of applications that look
- 05:54for software that were off prim of the Z platform or even
- 05:57off prim of Walmart, because they wanted to be cloud based.
- 06:00So while there have been a lot of challenges running
- 06:05off prim on Z, or running outside of our data center walls,
- 06:09when there have been failures or have been challenges
- 06:12that did not meet our needs, people have come to our team to
- 06:16me and Rich and to the other to find technicians on our team
- 06:19to use the current service we have or even develop new services, for example
- 06:24there's an application that needed to have an event
- 06:27control process where they send in millions upon millions of
- 06:32events to log and then do searches on those events looking
- 06:36for a needle in a haystack, and they tried some other products
- 06:40off- prim. They tried some "in the cloud", wherever the heck that is,
- 06:44and it didn't work. In fact, it failed miserably and so they
- 06:48approached our team and we developed a product
- 06:51called ECP- Event Control Process that
- 06:55initially, day one production is running about five million
- 06:58transactions a day, and by this time next year,
- 07:01one of the service types will be running five hundred million
- 07:04transactions a day, and another service type we're in the process developing today,
- 07:08will be another five hundred million. That's a billion transactions
- 07:11a day that we will be running on z/OS. So to answer your question, yes
- 07:16developers have looked for other solutions
- 07:18off of z/OS. However when they look for the RAS, capabilities that Z provides,
- 07:24the reliability, availability scalability, they
- 07:28typically come back to us and look what the z platform provides.
- 07:32And something to, to think about.
- 07:35Going back to my little rant on cloud
- 07:38there's five essential characteristics cloud- there's self service,
- 07:41provisioning, there's metered measured service,
- 07:43there's broad network access,
- 07:46there's pooled shared resources. It's very important
- 07:50understand that z/OS out of the box has four of these five.
- 07:54The fifth one being the self service provisioning.
- 07:56IBM's been addressing with SMF and some other projects as well.
- 08:01So when you look at what z/OS provides out of the box, it's
- 08:04already cloud-like to begin with, and with a little bit of work,
- 08:06a little bit of creative coding, you can create a natural cloud
- 08:09delivery model and a cloud service provider on z/OS.
- 08:13I like to add to that though that even though IBM is working on
- 08:19things like Randy mentioned, COSMF and the
- 08:23like, you know a lot of us have already developed a ton
- 08:28of automation over the years
- 08:31to satisfy our immediate consumers.
- 08:36And you know I think
- 08:39that shouldn't be ignored. Randy and I have taken that, maybe a step further
- 08:46to make sure we satisfy every need of a consumer,
- 08:50but a lot of shops already do a lot of automation, and it's
- 08:53just a matter of taking that maybe a step further
- 08:57and making sure that your consumers are taken care of immediately,
- 09:03automatically, buy the things they need. So I just wanted
- 09:07to point that out.
- 09:10Wolfram, if you don't mind, can I add more thing to that?
- 09:12You okay with that? Okay, no worry. Okay Rich just talks about automation
- 09:17and here again, cloud is a how not a where and one of the
- 09:21aspects of cloud is the self service provisioning, which happens to be
- 09:25automation, the the automation that Rich was talking about.
- 09:27And so many Z shops already have so many things automated they're
- 09:31just like really, really close to being able to provide this
- 09:34cloud delivery model with the tools they've already developed.
- 09:37What we did to take this a step further to make sure that we
- 09:40meet the needs of our developers, is we made this self-service provisioning
- 09:44a portal. In other words, a something from a browser or
- 09:48something that's web accessible.
- 09:51So we developed our own portal called Zpac, which is our provisioning portal
- 09:55and happens to run in CICS and happens to be cobol and assembler language.
- 09:59It happens to use just a standard stuff that's out of the box
- 10:03that everybody else can use. It uses JSON.
- 10:06Well I have a script in there, right? Now that's true.
- 10:13The front end is javascript the entire back end is cobol and assembler.
- 10:18Let's make that clear. Right, right, so now the original
- 10:24version did not have the javascript, but we did add that to
- 10:26it. So that that is right. What we did is we took our automation
- 10:30and we put it in a web based it ok through the JSON services,
- 10:35and so now our developers can go to our portal.
- 10:38They can see a list of our services. They can select which service they want,
- 10:43click it we get a drop-down box of all the different things.
- 10:46You need to fill in and by doing so we do our own API management.
- 10:51We create the url, we create the security
- 10:54based upon the information they give us.
- 10:56And then, in about ten seconds time we've provisioned the entire service forum.
- 11:01We give an endpoint for data, input for Q,A, ten seconds through up and running.
- 11:05That's what it means about automation and giving to the developers
- 11:08what they need and when they need it. The sooner you give it
- 11:11to them, the more apt they're going to use this platform instead
- 11:16of going off platform to look for a service. So it speeds the
- 11:19market, it's also the capabilities that that they actually need something that's useful.
- 11:23So anyway, hopefully that helps.
- 11:26Okay thank you very much guys for sharing this information
- 11:31with us. It was a great discussion and yeah thank you very much
- 11:36for contributing to mainstream MOOC in Germany.
- 11:39well thank you for having us. we're honored. It
- 11:42was an honor, privilege to be there last year at your presentation
- 11:46to be the keynotes and to have that workshop and
- 11:49Wolfram, really appreciate the work that you do to further this platform,
- 11:53you know not only in Germany but also in central
- 11:56Europe and in the United States. So thank you for the work that
- 11:58you do as well.
To enable the transcript, please select a language in the video player settings menu.