The Main in the Frame Stays Mainly In The IB … Aim?

[00:00:00.10]
Ned: This episode of Chaos Lever is brought to you by Integro One and their OneCon conference, coming this October to Bethlehem PA. More on that later. You think for a second, Hey, I'm never going to need that mini USB, then you're wrong.

[00:00:14.13]
Chris: We're not in the world of 75,000 different kinds of Scuzzie cables anymore, but it's not that far off.

[00:00:21.28]
Ned: I am certain I still have a Scuzzie Terminator somewhere in a box, and someday I'm going to need it.

[00:00:32.08]
Chris: Sounds like the lamest sci-fi movie ever. The Scuzzie Terminator.

[00:00:50.21]
Ned: Hello, Alleged Human, and welcome to the Chaos Lever podcast. My name is Ned, and I'm definitely not a robot. I'm a real human person who is not filled with cables, transistors and dreams of one day flying up in the stratosphere with the rest of my brethen. I mean, I like my feet on the ground, just like Chris, who's also here. Hi, Chris.

[00:01:12.28]
Chris: What's wrong with the ground? This is where all my stuff is.

[00:01:19.23]
Ned: It feels icky on my ambulatory appendages.

[00:01:25.11]
Chris: Well, if you're being on the ground, why do you fall down so much?

[00:01:31.15]
Ned: Because the world wants to punish me.

[00:01:35.08]
Chris: That makes sense.

[00:01:36.28]
Ned: Does that all line up for you?

[00:01:39.25]
Chris: Although I will have to admit that I absolutely 100% ingloriously ate shit on my last long run.

[00:01:47.27]
Ned: I mean, that's terrible to hear. I'm so sorry.

[00:01:51.06]
Chris: It was in the trails, so that was a bit of a mercy.

[00:01:55.14]
Ned: That's where I usually eat shit, too.

[00:01:57.29]
Chris: I caught a root, full I was, of course, running downhill when it happened, because why wouldn't I be? Landed on the water bottle. I have a bruise on my thigh that resembles One of the countries in the Caucasus Mountains.

[00:02:18.11]
Ned: You're not going to tell us which one?

[00:02:19.20]
Chris: I don't actually know.

[00:02:21.02]
Ned: Okay.

[00:02:22.02]
Chris: Let's go with Druidia?

[00:02:27.01]
Ned: Sounds right. Sounds familiar. I think I've heard that before. If it makes you feel any better, I also ate shit at the end of my long run last weekend, and I was on a trail.

[00:02:39.24]
Chris: I don't even ask you anymore. I just assume that's what happened.

[00:02:43.24]
Ned: It's actually been a while. I've been doing a lot of runs on the Delaware Canal Trail, which has no rocks, or very few of them. It's mostly flat. But this time I decided to venture out once again into the hills and mountains, and they punished me for it.

[00:03:01.29]
Chris: Once again, we get to one of the most important rules of life. Don't try.

[00:03:08.27]
Ned: So important that I have it tattooed on the inside of my eyelids. Every time I blink, it's there to greet me. Let's talk about some tech garbage, shall we?

[00:03:19.29]
Chris: If we must.

[00:03:22.18]
Ned: The main in the frame stays mainly in the IB aim? Close Close enough.

[00:03:30.29]
Chris: Nailed it.

[00:03:32.04]
Ned: I appreciate the assist. Last week, we had a little fun examining the technical accuracy of the movie Hackers. There is one particularly ridiculous scene where the titular hackers band together to hack back against the evil Eugene, the Plague, Belford, and crash the Gibson. Hack the Planet. So what is the Gibson?

[00:04:04.18]
Chris: It's a guitar. Duh.

[00:04:06.09]
Ned: That's the first thing I would think of. But in the film, it's a fictional mainframe that is responsible for managing some portion of the Ellington Mineral Corporation Corporation where the plague works, and from whom he is attempting to defraud serious amounts of money in that salami slicing scam that we explained is not a real thing that you can do. If you haven't watched the movie, just go do it. I would classify it as a romp. And Fisher Stevens, criminally underrated at the best of times, check out My Science Project, he turns in a great performance as The Plague.

[00:04:49.12]
Chris: Yeah, I mean, go watch the movie. We'll wait.

[00:04:51.21]
Ned: Yeah, it's fine. I actually don't have time. So while the Gibson is a fictional computer named In Honor the author William Gibson, mainframes are a real thing.

[00:05:06.22]
Chris: So Chris-Barely certain everybody knows that.

[00:05:09.08]
Ned: You'd be surprised. When I say mainframe computer to you, Chris, What's the first thing that pops in your head?

[00:05:18.04]
Chris: A room full of boxes that all look exactly the same.

[00:05:24.09]
Ned: Yeah, I mean, that's definitely one of the classic images. If you're a person of a certain age, your mind might have gone to a big box sitting in the middle of a room with blinky lights.

[00:05:37.05]
Chris: Oh, shit. I forgot about the blinky lights.

[00:05:40.05]
Ned: And you would be thinking of Wargames, another classic movie with just about as much accuracy as Hackers, except for the phone freaking, which is actually very accurate in both movies. I don't know why Hollywood nailed that one thing and won't shut up about it. Go back to Mannecan 3, Hollywood. Andrew McCarthy's waiting. Forever. What was he talking about? Yes, the Wapper from WarGames, which stood for War Operation Plan Response. I read in several pieces on the movie that this was a joke on a real computer called BurgerBRGR that was used by Norad. But as far as I can tell, it's completely unsubstantiated by any credible source. But it didn't make me really want a hamburger. Flamed broiled, if possible. The Wapper was built to be a hulking mass of inscrutable blinking lights and random bloops, and it was modeled after real mainframes of the 1960s and '70s. You might think that such a relic would have no relevance in our modern computing landscape. We've embraced X86 commodity servers, Linux in the cloud. Mainframes are your dad's of computing. So lame, I say, as a dad with a teenager who thinks I'm lame.

[00:07:08.21]
Chris: He's right.

[00:07:10.18]
Ned: And yet, at the Hot Chips conference in 2024... Yes, Chris, it's named Hot Chips. No, I will not be taking questions. Ibm announced their forthcoming Z17 mainframe, powered by the, and I don't know how to pronounce this, but I'll try, Tulum or Tell 'Em. I like Tell 'Em because it's funny. I think it's Tell 'Em, too, Chip.

[00:07:37.08]
Chris: Yeah.

[00:07:38.26]
Ned: Go run Tell 'Em about it. Not only are mainframes still a real thing, IBM is making new ones. Yeah, you heard me correctly. Ibm is making something new. I'll wait until you've gathered your composure.

[00:07:56.10]
Chris: I don't think I had it in the first place. It's the problem.

[00:08:00.13]
Ned: The TELM 2 processor will reportedly have 43 billion transistors on the die with eight cores and a built-in DPU and AI accelerator, all clocked at around 6 gigahertz, which if you compare to the latest Intel processors, is... Well, at least the clock speed is fairly impressive. The normal clock speed for an Intel processor tops out 4 gigahertz. But in most other respects, the most recent Xeon processor has more transistors on the die. But the Z17 or the TELM 2 is allegedly more reliable. So I guess that's a feather in its cap. It's also really expandable, the mainframe, not the chip itself. The Z17 can have up to four drawers of CPUs with four sockets in each drawer and two chips chips per socket, which if you do the math, is a lot. It's 32. It's 32 telem chips in a single system with 256 total cores and 64 terabytes of memory. 64 terabytes of memory in one system.

[00:09:18.18]
Chris: That's with a T. Yeah.

[00:09:20.25]
Ned: That does not include the additional 12 drawers dedicated to expansion slots for storage, cryptographic processors, or AI accelerator cards. Ibm is still creating new mainframe versions, and people are still buying them, at least according to IBM's marketing. Take that with the attendant amount of skepticism one reserves for marketing and statistics. What they claim is that 45 of the top 50 banks and four of the five top airlines use mainframes as part of their core platform. Given the recent track record of airlines and banks, not sure I'd single out those industries. But still, a study commissioned by IBM claims that 70% of the world's production IT workloads are handled by mainframes. I'm sure there's caveats bound with that statement, but the point is mainframes are still a real thing. Not only that, but 2024 marks the 60th anniversary of the original mainframe, the system/360. Which I'm just going to say system360 from now on, the slash is implied. That system was also created by, I think you can guess, IBM.

[00:10:40.29]
Chris: Oh, I was going to say Chiquita.

[00:10:43.27]
Ned: Yeah, I mean, it's very possible that United Fruit Company or whatever they're called did, in fact, design mainframes. I mean, they overthrew governments. I don't see why they couldn't build their own silicon chips.

[00:10:59.10]
Chris: Bananas can I do a lot of stuff, man. That's all I'm saying.

[00:11:02.25]
Ned: Quite possibly the world's most perfect food and circuit designer. Who knew? So let's take a little scroll down memory lane to learn what a mainframe is and how it's still relevant today. Chris, at this point, I'm fairly certain I was born in the wrong decade. I know that you secretly pined for the 1770s. I'm more down with the 1970s, or maybe the 1960s because that's when computing was groovy.

[00:11:34.17]
Chris: That does explain why you're wearing bellbottoms and there's a smell.

[00:11:40.04]
Ned: Leave my petulia alone, sir. It's your fault for installing smell a vision on your camera.

[00:11:46.11]
Chris: Admittedly, that was my bad.

[00:11:48.29]
Ned: Indeed. The system 360 was by no means the first large scale programmable computer, as we've mentioned in previous episodes. We can probably assign that particular honor to the Colossus constructed during World War II to decode German communications. After the war, military and civilian groups alike continued to create new computing edifices. For instance, the creation of ENIAC and the IBM 702. The ENIAC, in particular, was a bespoke machine with the software being closely bundled to the hardware. That was a problem with the early computers. You couldn't take your program that you wrote for the ENIAC and go run it on a different machine. The software and the hardware were intrinsically linked. When the time came to upgrade your computer, say from the IBM 702 to the 705, even though it's the same vendor, the same model line, you might need to rewrite several of your programs. God help you if you were switching to a different vendor, say to a Univac 1107 from the Sperry Rand Corporation, you're going to have to rewrite everything from scratch.

[00:13:09.19]
Chris: In assembly.

[00:13:11.12]
Ned: Yeah, or something very, very close to assembly. I think there were a few very basic programming languages at that point beyond assembly, but yeah, it was painful. In 1961, IBM made a big bet on bringing a new unified architecture architecture to their line of computers. They spent $5 billion over the next four years, and that's 1960's $5 billion, which would be like 80 quadrillion today or two bitcoins. Something like that. They spent 5 billion over the next four years to produce the System 360 series of mainframes. That line replaced all previous lines. They had about five previous lines that they were selling. Critically, they included a compatibility promise that future versions of the mainframe would be backwards compatible. The system was also expandable, so you could add more capacity as you needed to. The system 360 début in 1964, and luckily for IBM, sales took off, selling more than a thousand units in the first month. Think about this, this is 1964. Selling a thousand computing anything was a miracle. That essentially made IBM and System 360 the standard for mainframe computing overnight. Speaking of standards, the System 360 also standardized the use of the 8 bit bite.

[00:14:48.19]
Ned: You and I take it for granted that a bite is a collection of 8 bits. It's like one of the first things you learn in computer science. But that was not always the case. It's not like the 8 a bit bite is a natural law or something. The bit itself is pretty fundamental to the binary nature of computers, but there's no reason that we grouped them by 8s. Could have just as easily been a 4 bit bite. Now for a brief message from our sponsor, Integro1. Our technology landscape is constantly evolving, and it's getting difficult to separate the hype from the reality. Is AI really the next big thing? Will WebAssembly change application deployments forever. Is the era of public cloud over or is it just getting started? I find the best way to discover the truth is by talking to other practitioners and hearing from folks who have their finger on the pulse of enterprise IT. That is exactly what you'll find if you'll be able to connect with like-minded professionals, get one-on-one demos from technology experts, and attend over 60 interactive sessions. You might even get a keynote from me, Ned Belivance, on the reality of in 2024.

[00:16:01.13]
Ned: It's two days of learning, connecting, and growing with fellow IT professionals. If that sounds good to you, join me at the Wind Creek Casino on October first and second for OneCon 2024. Now back to the show. In fact, the Intel 4004 was a 4 bit-based processor, and 4 bit-based computing was common in early calculators to save on money. You only needed 4 bits to represent a binary-coded Adorably, the 4 bit grouping is called a nibble instead of a bite.

[00:16:37.14]
Chris: I get it.

[00:16:38.23]
Ned: Yeah, just soak that in. That's good.

[00:16:40.17]
Chris: I get jokes.

[00:16:43.07]
Ned: Anyway, For big iron and later personal PCs that followed, the System 360 made the 8 bit bite the standard. The physical architecture of the System 360 was a central processing unit and peripheral devices that hooked into it. You talked about a room that has a whole bunch of big metal rectangles. You said they were all the same. They wouldn't all be exactly the same. Slightly different heights. Some They have worrying or spinning magnetic tape. Some might look like a dishwasher, but actually have hard drives in them because a hard drive was two feet across at the time. The big rectangular box that all of those boxes would connect to is the main frame. It's literally a large metal frame housing the processor and system memory. That's where the name came from. Isn't that delightfully dumb?

[00:17:48.05]
Chris: I don't think I knew that, but now that I think about it, I should probably have been able to figure that out.

[00:17:52.21]
Ned: That one's not cloaked in mystery. The word dog, however, is, and there's a really good YouTube video about it. Anyway, speaking of the memory, the system 360 didn't use silicon transistors for memory storage. Instead, it used magnetic ferite core memory, which was a type of physical memory that relied on magnetizing ferite cores to hold a one or a zero. Now, these were a big upgrade from the previous use of vacuum tubes. Since they were non-voletile, meaning they would keep their state without power, and they were just a lot more stable. Magnetic core memory was replaced with silicon transistors with the release of the system 370 in 1970. However, the language of magnetic core memory lives on in our lexicon. When a system halts unexpectedly and it writes its memory contents out to disk, we call that a core dump. The core is a reference to magnetic core memory of yesteryear. Mainframes and core dumps. What more could you ask for? Don't answer that. I touched on it before, but the expandability of the system 360 was a really big deal. If you were a medium-sized business in 1964, looking to dip your toe into the world of computing, the other systems on the market were all of a fixed size.

[00:19:23.29]
Ned: You had to just guess how much compute you would need and hope that you got it right. If you were wrong and you needed a bigger computer in a couple of years, you were not only going to have to replace the system that you just bought, but also rewrite a whole bunch of your code. If you were wrong about the size and you need something smaller, well, you just paid a whole lot of money for a compute you didn't need. Maybe you can rent it out.

[00:19:51.11]
Chris: That was not a thing.

[00:19:52.25]
Ned: Actually, it was. If you look at some of the documentation and advertising of the era, there's actually a price for computing for System 360.

[00:20:03.22]
Chris: Oh, you mean rent it from IBM?

[00:20:05.13]
Ned: Yes. They would host the system for you and rent out fractional portions of the mainframe to you.

[00:20:10.28]
Chris: Right.

[00:20:11.28]
Ned: That's how far back cloud computing goes.

[00:20:14.28]
Chris: Well, that's also how a lot of computing happened for people that were non-professional. You would schedule time whenever it wasn't being used, 3:00 in the morning, for example.

[00:20:29.22]
Ned: Right. Because any idle time would be wasted time on the person who owns the system. They can get somebody else to pay for that idle time. That's a bonus. Imagine that, yet another thing that was not invented in the year 2012. The system, 360 mainframes, allowed you to start small and expand as needed. I mean, up to a point. They did have a maximum. When it did come time to upgrade, if you hit that maximum, you could get the system 370, and you didn't have to rewrite your software. That modern Z17 system I mentioned earlier, you can start with a single drawer of Compute and maybe one or two of peripherals and two telem chips, and then scale up to four drawers with 32 telem chips, and it's all the same system. Which brings me to why mainframes still exist.

[00:21:27.10]
Chris: Why do mainframes still exist?

[00:21:29.04]
Ned: Well, thank you. God, I was waiting. Even though a mainframe can be expanded, it's all part of the same system. What do I mean when I say system? When we're referring to a single system, where do you draw the boundaries between one system and another? You could start with a processor and memory. As we've discussed in past episodes, check out our episode about the von Neumann architecture. All you really need to be a computer is processor and memory. That's basically a complete system. The processor and the memory are connected together via a bus, so anything else on that bus could be part of the closed system. So much so that that particular bus is called the system bus. We're good at naming things. We can expand to peripherals using a separate bus, which in modern systems is going to be the PCI bus. The The Z-Series systems have a system bus to interconnect processors and memory called the X bus and the A bus. Essentially, as you add processors to your drawer or more drawers to your system, the telem processors in Each drawer can talk to each other on these dedicated busses and talk to the memory.

[00:22:50.28]
Ned: To talk to the peripherals, the chips can engage with the PCIe5 bus, which is providing connectivity to the other drawers in the system. Technically, PCI can be used as both an internal and external bus type, but in this case, it's just internal. That's one way to draw the border for a system. It's whatever that system bus and PCI bus can directly talk to. Another way to draw the border for a system is through its operating system. You would expect a single system to be under the management of a single operating system that has complete view of all the attendant hardware. The operating system can schedule jobs and select specific processors, memory, and peripherals for that job to use. A job could be a batch process, a web application, or even a virtual machine, which also existed in the 1960s. I feel like I'd have to hammer that home. All this stuff existed 60 years ago. The unified view of such an immense amount of hardware is hard to get in any other type of system. Even the most robust Commodity X86 servers tend to max out at four processors. If you want to build a larger machine, well, what you actually end up doing is creating a cluster of machines with each node in the cluster running its own operating system, and then some overarching cluster management application that will handle scheduling.

[00:24:22.05]
Ned: Kubernetes does that, for instance. That's also what supercomputers do. If you need a really big honking system to process a massive amount of data or house extremely large and important databases, then a mainframe might be your only option. Mainframes are also ridiculously over-engineered for uptime. A mainframe is meant to run at 98% plus utilization levels with eight nines of uptime. That's like one breakdown every 11,000 years or something.

[00:25:00.25]
Chris: That's better than a lot of AWS services for reference.

[00:25:05.17]
Ned: Indeed. That's not to say that parts in the system don't fail. They can, and they do. But it has an incredible amount of redundancy, so the system itself stays up and everything in the system can be hot-swapped without taking down the operating system as a whole. You can just say, section off this portion of the hardware to the operating system, and it will do it. Then you can swap out that piece of hardware everywhere. It's amazing. Mainframes are also considered super secure since the entirety of an application sits within the moat of the mainframe. That includes the use of cryptographic accelerator cards and AI cards if you're generating a model. Rather than sending traffic out across the network to distribute a workload across multiple nodes, everything is happening inside the closed system. Is it really more secure than a well-maintained cluster of x86 servers? I have my doubts, but this is the narrative that IBM is really pushing on why you should still use a mainframe. The companies that are most likely to use mainframes today are financial institutions, healthcare organizations, government agencies, and retail establishments. In fact, the only time I had to deal with an actual mainframe in my career was when I was working at the main office for a retail chain.

[00:26:29.17]
Ned: We had an AS400 that was responsible for processing sales, inventory, and payroll, I think. It did a lot. Basically ran the whole ERP system. It was pretty important. What all these companies have in common is the need to process massive amounts of data in a secure and reliable fashion. A lot of them have homegrown legacy code written in cobalt or Java that was never designed for the modern, disaggregated model of cloud applications. These are monolithic applications that need to run on a single system really fast. So you have to make that system as big and as reliable as possible. Could you go take that application and rewrite it in a more modular way to allow horizontal scaling of individual services? Yeah. Would that cost a shit ton of money? Yeah. There's like a trail of of failings of companies trying to do this and eventually just giving up, which means it introduces a lot of risk. If there's one thing we know about government agencies, health care, and financial institutions, they love risk. Mm-hmm. Yum, yum, yum. Hence why the Z 17 is coming to a major bank near you in 2025. Chris, I assume you've had some interaction with mainframes in your career?

[00:28:01.24]
Chris: I've only ever actively worked on two. I almost said one, but the answer is actually two. My first job at the university, they ran their student record system off of The mainframe. Mainframe that had been upgraded many times, but was still green screen connectivity. Like you said, this is a set of systems that does one thing, and it never goes down at all, ever. There's more to it than what we talked about. Mainframe programmers are tearing their hair out talking about the different ways that the mainframe programming environment is specifically designed to be more secure. Applications are basically crash proof. You literally dedicate, I am using these memory sticks for this application. There's a specificity in mainframes that you don't get with commodity computers either. A lot of that really does go to the utilization and uptime numbers that we were talking about. But it also means you're not going to get that code on GitHub.

[00:29:11.27]
Ned: Yeah. It's extremely hard to write to it if you're not familiar with that environment. If you take a programmer who's familiar with front-end JavaScript and you say, I need you to go update this cobalt from 40 years ago, it's going to be a struggle.

[00:29:30.08]
Chris: It's to the point where people have these applications that are running on mainframes and they have for so long, like you said, changing them out for something else, something in a container or whatever, a pass, what the kids are doing these days. People are looking at that and going, nah. So you know what you can do? You can run IBM Z-Series in the cloud.

[00:29:57.17]
Ned: You sure can. There's also a non-zero number of vendors out there that I've seen when I go down to conferences in DC that specialize in mainframe emulation. So if you are running on an older Z-Series and you want to migrate stuff to the cloud or even just on the X86 hardware, they will emulate the hardware that's in that Z-Series on X86 so that you can move your stuff off of that aging hardware because the support contracts for mainframes, they just get more expensive the older the system gets.

[00:30:41.04]
Chris: Right. Mainframes themselves Not cheap.

[00:30:49.00]
Ned: Iwateringly expensive would be the phrase that comes to mind.

[00:30:56.06]
Chris: One last thing that's just not really about the technology, more about the manufacturing. Remember, we talked about the uptime and the utilization and all that stuff. One of the other things that companies do, IBM in particular, is test the physical reliability of these servers. I shouldn't say server, but you know what I mean.

[00:31:17.17]
Ned: I know exactly what you mean.

[00:31:18.22]
Chris: They have basically a boot camp on their manufacturing floor that stress tests these systems, these frames, physically. If you've ever wanted to watch a 2000 pound frame be shaken around violently to simulate a 9.0 Richter scale earthquake, head to your nearest YouTube and look that up because it's an adventure.

[00:31:43.21]
Ned: It's awesome. Well, hey, thanks for listening or something. I guess you found it worthwhile enough if you made it all the way to the end. So congratulations to you, friend. You accomplished something today. Now you can go sit on the couch, fire up your Z-Series emulator and play a game of Pong. You've earned it. You can find more about the show by visiting our LinkedIn page. Just search chaoslever. Com, or Chaos Lever, or go to our website, chaoslever. Com, where you'll find show notes, blog posts, and general top foolry. We'll be back next week to see what fresh hell is upon us.

[00:32:17.08]
Chris: Ta-ta for now. Back in 2011, you could get a business class low-end Z-Series for the low, low price of $75,000.

[00:32:37.02]
Ned: I'll take two.