Distractions - I've got some

[administrator]

I’ve been a bit quieter than normal on the forum because I’ve been distracted by some hardware. Meet my new hardware platform…

It’s a beast, although the LEDs are a bit bright so it’s not that easy to see it. Also whilst the picture fills the frame, this is only a Mini-ITX case so pretty small for what’s in it.

It’s a 4-node cluster. It’s got 24 CPU cores, 40GB RAM, 96GB EMMC, 2.5TB of SSD disk (part SATA, part M2 NVME). It’s also got 12 TOPS of NPU (used for AI stuff). For the technically minded the details are:

Nodes 1 & 2 - Raspberry Pi Compute Module 4 - 4GB RAM, 16GB EMMC, 4*ARMv8 CPU cores, 500GB SATA disk connected by mPCIE. Running a hybrid of Raspberry Pi OS with DietPi.

Node 3 - Turing RK1 - 16GB RAM, 32GB EMMC, 8*ARMv8 CPU cores, 1TB of M2 NVME disk, 6 TOPS NPU. Running Ubuntu 22.04 with XFCE4 desktop

Node 4 - Turing RK1 - 16GB RAM, 32GB EMMC, 8*ARMv8 CPU cores, 500GB of M2 NVME disk, 6 TOPS NPU. Running Ubuntu 22.04

The board that this is all plugged into is a Turing Pi 2 cluster board. That’s got it’s own processor, memory and OS with SSH access and a Web GUI to the Board Managment Controller (BMC). The BMC allows you to turn the individual nodes on and off, change which one currently owns the ‘control’ USB port and you can also flash an operating system onto any node from the BMC. The board also has lots of blinkenlights so it looks cool sitting next to me flashing away as it does stuff.

Whilst there’s a lot of oomph in the box it’s quite frugal with power, peaking at about 60w when everything is doing stuff at the same time.

I’ve just finished getting the OS flashed onto each board today and also have them running from disk rather than their slower onboard EMMC memory. Next step is to get Docker Swarm and GlusterFS installed. This will allow me to deploy containerised apps using multi-node storage. It’s got a level of fail-over, so if one node fails, the apps running on that node are moved to another node. The storage is spread across the nodes too, a bit like RAID, so if apps have to move they’ll still find the storage they need.

What can I do with it? Well what can’t I do with it!

My first plans are to add a MariaDB container so that all the apps in the house that need a database have a centralised one available. I’ll also be adding a centralised MQTT server for handling weather and smart home data. I’ll also move the weather station software across to it. I’d like to move the smart home (Home Assistant) software to it too, but I need to do some research about how that will work first.

The Turing Pi 2 has a full set of Raspberry Pi compatible GPIO pins attached to Node 1, so I’ll also be investigating whether it’s feasible to move my Software Development Pi onto it.

After that…well I’m still thinking of ideas of how to use it.

[bitsostring]

Wow! (And that’s because I can only understand one word in ten.)

For comparison, here’s my Spectrum 48K built-in to a full-size keyboard, with twin disk drives.

It’s now in the Royal Museum of Scotland, Edinburgh - somewhere

[tpfrs]

I still have my 48k Spectrum lol

[broadstairs]

We’ve come a long way since the 1960s. The first computer I worked on was an IBM 1400 with 8k ( yes 8k) memory which was the maximum it could have. It was about 10 by 4 by 6 feet high if my memory serves and needed air-conditioning! Followed not long after by the System/360 model 30 with 64k memory!

Stuart

[administrator]

I never owned a ZX Spectrum, although I sold a lot of them way back when I worked in a computer shop.

Another project that I have underway is a Z80 based computer. That’s built from a backplane and multiple PCBs, so I have a lot of soldering to do. I did guesstimate how many pins needed soldering, but I’ve forgotten now.

It will run a version of CPM (and other operating systems) from a 128MB Compact Flash card. It’s got 512MB of ROM, a real time clock, serial ports, WiFi, some digital IO ports and probably some other stuff I’ve forgotten now. That’s a task for the summer months though. I can sit outside and solder in the sun…and not be too bothered about fumes.

The Z80 system is a nostalgia project. My first personal computer was Z80 based (Sharp MZ80K) so I did a lot of Z80 programming when I should have been studying for Uni exams I also worked on a number of other Z80 based systems, Cromemco, Tatung Einstein and others. That initial Z80 work is what got me into IT as a career.

I don’t think I’ll build an IBM 360, although arguably my cluster is considerably more powerful than that ever was. I remember using an IBM Mainframe at Uni. Possibly a System/370 of some kind but I don’t remember the exact model. I do remember loading stacks of punched cards containing my Algol-W source code via HASP (?). I still remember my username from 45-ish years ago - QXEC4.

[broadstairs]

The biggest System/370 I worked on in my hardware days with IBM was a model 168 but there were several other slightly less powerful ones. By the mid 70s I had moved from hardware to software support still with IBM, during that time the IBM PC arrived so I played with that as well but my main job was support of mainframe software. I retired (well took voluntary redundancy) early in 2002!

Stuart

[bitsostring]

@tpfrs

I still have my 48k Spectrum lol

I still have one, too, with various joysticks, a couple of ZX thermal printers and a box of tapes - not all games, although my daughter certainly had a few! Not sure where the old mono tape deck is, though

@broadstairs

We’ve come a long way since the 1960s.

Early 60s, day release from school to learn how to program an Elliott 80-something (6?) at the local tech college on 8-hole paper tape (1" wide). . .

@administrator

I remember using an IBM Mainframe at Uni. . . I do remember loading stacks of punched cards

Early 70s, Edinburgh University had a System/360, I think. We weren’t allowed to load the cards, had to leave them at reception. Then you could collect the printout from one of the dot-matrix printers strategically placed around the campus - or go back to reception to collect the job one of the operators had dropped. . .

[novawx]

Chris - that’s quite a system you’ve put together for home use!

You guys are really dating yourselves.

I learned FORTRAN 77 in college using a mainframe computer - circa 1979. I did my first real work on a computer when I was in the USAF in 1981 - part of their recruiting pitch was that they had the latest a greatest technology - that first USAF computer used punch cards! The only training they gave us was to keep the rubber band around that cards and don’t drop them. The punch card machine looked like something from the 1950’s. My job was to redraw 500 mb height lines on a northern hemisphere chart with a “puck” (mouse) based on aircraft observations and satellite imagery. After adjusting the height lines, the computer would create the punch cards that we would insert into the existing deck of cards. We’d take the cards to the basement and slide them through a small opening to a computer operator who would load the cards and run the 500 mb analysis, which would be delivered to us on large pieces of paper from a dot-matrix printer. I think the mainframe computer had several 10 MB hard drives that were about 12" in diameter and they were removeable.

I wrote my first “operational” program on a Z-248 PC when I was stationed at RAF Mildenhall in 1985. I took a hand-written fog forecast flow chart and converted it to a BASIC program on the Z-248. Most forecasters stuck with the flow chart because they didn’t trust the computer!

My first computer was an Apple II-C from 1986. Switched to PC’s in 1988 when I started grad school. I wish I still had the Apple II-C.

[ktrue]

Chris, that is a magnificent construction! Bravo!!

My ‘old guy’ computing started at University of California, Davis in the mid-60s with an IBM-7090 IBSYS/IBJOB and Watfor on cards. Also a Xerox 360-like programmed with paper tape in the Engineering lab. At Fairchild Semiconductor it was IBM 3090/MVS and IBM 168/VM (but using 3270 terminals). At Intel, it was IBM 3090/MVS systems programming and BSDi unix systems for internet infrastructure. First personal computer was a VIC20 with cassette tape and first PC was a Compaq Deskpro.

Now, I’m surrounded by 4 PCs/Laptops (Windows 10/11) and 4 Raspberry Pi running Raspian Linux.

[administrator]

I think we need to form an OGC (*) or “Old Geeks Club”. The young’uns of today don’t know what it’s like to program a 4-bit CPU that only has 256 bytes of memory attached to it using binary toggle switches for input and a row of LEDs as output

(*) Not to be confused with an OCG which were a consideration in the cyber security risk assessments that I’m happy that I no longer have to do. Actually, the risk assessements were reasonably interesting to do. It was the mitigation policies and instructions that were a PITA to write.

[bitsostring]

@administrator

. . . and a row of LEDs as output

Are you sure they were LEDs?

[administrator]

It was the 1970’s not the 1870’s LEDs were invented in 1962.

[bitsostring]

OK. I forgot you are younger than I am

They were a bit expensive at first. Wikipedia says “The first usable LED products were HP’s LED display and Monsanto’s LED indicator lamp, both launched in 1968.”

[administrator]

They took off quickly after 1968. I had a TI calculator in 1974/5 that had a LED display. The 4-bit computer with LEDs would have been 1976/77.

[administrator]

@bitsostring

I forgot you are younger than I am

I’m joining the YGC

[bitsostring]

@administrator

They took off quickly after 1968.

I saw my first electronic calculator in 1970. It was the size of a portable typewriter and used Nixie tubes. . .

Hence my avatar

EDIT: Sorry. For our younger readers I’d better translate:

Typewriter: a mechanical device for printing characters on paper; and

Nixie tube: a cold-cathode display.

[administrator]

@bitsostring

For our younger readers I’d better translate:

I’m not sure your translations will help much. A typewriter is a propelling pencil? Also, I doubt that many youngsters will know what that kind of cathode is. I suspect the closest people get these days will be a cathode in an electrolysis cell. They’re cold too, but not nixie tubes.

I wonder if there’s a market for a digital abacus with a USB interface?

[broadstairs]

I think this thread has already become the OGC … oh no I just found that’s the official acronym for the Office of Government Commerce

Stuart

[administrator]

@broadstairs

I just found that’s the official acronym for the Office of Government Commerce

Quick, run away and hide before they notice.

[administrator]

After battling with the Raspberry Pi Compute Module 4s I’ve given up on them. They are a nice solution for what they’re intended for, but they don’t suit the way they work with this cluster board and the way the cluster board works with them.

So I have to order a couple of non-CM4 replacement modules. After I get those installed the specs of the ‘beast’ will be:

• 32 CPU cores
• 64GB RAM
• 128GB EMMC storage
• 3TB SDD storage
• 24 TOPS of NPU

There will also be 4 onboard GPUs but I’m not sure what they can be used for. Crypto-mining maybe or possibly generating AI models, although they probably aren’t powerful enough to do either of these things well.