COTS. Commercial Off The Shelf. I’m not a huge fan of industry jargon, but I must admit these four little letters are music to my ears. COTS = Go to the shop and buy it. No questions asked, no special contracts, or agreements, or NDAs, or backroom deals. If I can buy it, so can you – and so can everyone else.
I love COTS so much, I’ve combined it with two of my favourite things from the early to mid-2000’s: Harry Potter, and Flaming Text.
(The fire symbolises how ‘hot’/cool COTS is, not that it’s on fire and about to wreak havoc throughout your facility. The Harry Potter font is just because I’d probably rather be reading Harry Potter right now.)
Honestly though, it’s not something I generally think about – yet it influences my work in subtle ways. I’ve never been particularly fond of special purchase agreements. If you have to call or email someone for the price, I’m not interested. If you have compulsory on-site installation from a vendor, then I’m sorry but your product is simply too complicated. Last week, someone asked me to sign a NDA before even getting a list of features or price. I have another four little letters for these sorts of people: ‘NOPE’, not going to happen – I’ll find another way, thank you very much, have a good day.
When my personal purchasing philosophy intersects with the broadcast industry, it sometimes confuses, offends and upsets. “How dare I suggest you should just sell me an OVF template instead of a 2RU box with a motherboard in it“, vendors sometimes imply through their piercing stares at tradeshows. Seriously – sometimes it’s as if Julie Bishop is coaching the salespeople and vendor reps…
I mean, what’s so bad about providing a transparent pricelist on your website and selling software unhindered by expensive hardware?
When I was preparing lwSDS for the Raspberry Pi a couple of weeks back, I posted about it on Twitter:
Here’s a photo of my ‘#Livewire Simple Delegation Switcher’ running on a Raspberry Pi with touchscreen! Setup instructions and new configuration options will go up on my blog once I’ve finished testing. @kharnack @TelosAlliance pic.twitter.com/wQDULP07vM
— Anthony Eden (@anthony_eden) September 11, 2018
To me, this was a pretty logical thing to do. We need a simple way to control routing in-studio, and the Raspberry Pi is about as COTS as you can get. Not only is it off-the-shelf, but it costs so little money it’s not even funny.
Imagine my suprise when I got this reply:
Using a raspberry pi in a broadcast application is akin to drinking milk on a hot day, it’s a bad decision. Just ask Learfield Sports. For the love of everything holy, use the right tool for the job pic.twitter.com/c69aaWUwzx
— Andrew Stern (@astern) September 12, 2018
Call it Andrews’s Axiom:
“The Raspberry Pi is ALWAYS the wrong choice.”— Andrew Stern (@astern) September 13, 2018
That’s some nice use of a GIF, and I don’t know what the Learfield Sports reference is about, but it seems we’ve hit a classic philosophical difference in opinion. Sorry Andrew.
Andrew seems like a nice guy, and I enjoy chatting with industry colleagues online, but I figured it’s about time I document the case for COTS being a GOOD idea for the broadcast industry. In fact, not only do I believe this is GOOD, but I also believe it’s BETTER than many of the ‘black box’ broadcast solutions out there.
While doing so, I’m going to take a leaf out of Andrew’s book, and coin ‘Anthony’s Axiom‘:
“If a task can’t be accomplished reliably on commodity computer hardware, it probably hasn’t been engineered well enough”
That’s a pretty big statement, eh? I’m thinking of getting it printed on a T-Shirt. Much like Atwood’s Law, I’m hoping it reveals itself to be true in years to come.
Sorry to all those devoted product engineers out there that are still writing FPGA code and shoving motherboards in fancy 2RU boxes. Can we still be friends?
I think COTS hardware (such as the Raspberry Pi, Mini PCs, and standard desktop/server/network hardware) changes the way we think about building broadcast technology solutions. Products in this class of hardware have the following characteristics:
- Instant replaceability
So cheap you can carry as many spares as you need. Many ‘black box’ products we fill our racks and studios with are far too expensive for anyone but the best-equipped broadcasters to keep multiples of. - Simple, predictable design
The Pi gives us an ARM Processor, HDMI Output, one single Ethernet port, Bluetooth, WiFi, GPIO Pins, and a few other odds and sods. This is a well-defined, but incredibly powerful set of parameters to build on top of. And because these platforms are predictable, almost every known hardware issue has already been experienced. As of mid-2017, 14 MILLION Raspberry Pi’s have been sold. That’s crazy! If you think that’s a lot, in Q1 2018 alone, Dell shipped something like 10 MILLION PCs. You’re never going to sell that many hardware DSP engines, no matter how hard you try. There’s much less testing and battle hardening that’s going on with your broadcast hardware. In one generation of broadcast gear (which can be 5+ years), how many hundreds or thousands of PC models have Dell built, tested, released, supported, and EOL’d? - Scaleable
If I run everything in a VMWare or Hyper-V cluster, I can easily add more resources and scale up as needed. The same applies to commodity PC hardware, although with a few days delay to actually source the hardware (opposed to weeks to get something in from overseas broadcast hardware manufacturers). Imagine being able to add remote codecs of studio mix engines or phone hybrids or playout channels or audio processors at the click of a button! - Open Troubleshooting/Support
I understand companies wanting to make world-class support available to customers, but I don’t understand why they do so by channeling everyone via email and not publishing support resources online. If 14 million people have bought something, there’s going to be plenty of resources online. No need to wake up the poor ‘on call’ 24/7 support dude from the other side of the world.
When we use COTS hardware, we are literally standing on the shoulders of technological giants. You and I and all our industry buddies can’t come close to the sheer number of hardware R&D hours poured into PC hardware. Not even close. But we can harness all this R&D, combine it with our industry knowledge, and churn our really solid products that are going to move broadcast forward into the next era.
These hardware characteristics force us to think a bit differently about product design:
- Focus on software
Hardware is sorted. If you’re creating products, you should almost never need to build hardware again (one key exception is probably broadcast console surfaces, where tactile control is important and existing MIDI/OSC surfaces probably don’t give us a suitable UI). Focus on creating software that works across a variety of hardware platforms. You can still charge expensive licensing fees, but you can cut down on hassles such as manufacturing and spare parts. This should also dramatically reduce your time to market. - Redundancy by design
Hardware is so cheap – you should be running the same applications across multiple units at the same time. Don’t have redundancy built into your software? Too bad – you should. - Utilise existing components whenever possible
There are plenty of existing components out there now to handle IP audio. Software such as NodeRed change how we think about logic control systems. Arduino and Raspberry Pi already give us ways to handle GPIO and HDMI displays. USB SDR’s allow us to tune into FM & DAB bands, and pass that RF off to software for analysis. Even systems such as Q-Sys give us incredibly flexible audio DSP in commercially available boxes.
Thankfully, we already have some examples in the industry where COTS is being embraced (at least partially):
- Audio Over IP has all but removed the need for TDM routers and big patchbays
- IP Audio Drivers remove the need for broadcast soundcards, meaning you can run most automation systems on any Windows-compatible hardware you like
- StereoTool and BreakawayOne replace traditional FM audio processor boxes with software
- Lawo R3lay gives us software-based audio mixing
- Logitek’s JetLink is a software-based, bi-directional, low-latency audio codec
- Broadcast Launch provides SMS/Email/Social functionality for studios, and I’ve heard they’re working on building call-in phone system functionality
- Broadcast Bionics’ PhoneBox4 lets you run software-based phone hybrids
- HyperStudio is a proof-of-concept way to run some Telos Alliance products as VMs
- SDR-based modulation software (MPX Tool) is currently under development by Leif Claesson and co.
- Shameless plug: My new product, MetaRadio, is software-based (so far I’ve resisted the desire to build and sell a $10,000 appliance version)
- I’ve got a whole article with open source apps for the radio industry (including my very own lwSDS which proves to be quite popular)
Sadly, in the whole wide world of broadcast products, that’s a fairly small list. There could be more, and I sincerely hope I’ve overlooked some significant development or product announcement – but I’m not holding my breath.
Since the days of patchbays and reel-to-reel players, we’ve come a long way. But there is still so much further to go. Every time NAB or IBC comes around, there’s a new range of boxes on offer. More things to fill our racks. More PSUs to fail. More parts to become unavailable after a couple of years.
I’m committed to building and supporting COTS solutions to broadcast problems. Will you join me?