What Apple might do with distributed computing

I’m excited to see not only async-await making it thoroughly into the language (in Swift 5.6), but also the extensions that enable actors and distributed actors with this general sweep of the Swift language embracing concurrency. It’s been several years in the making, and the past year has been building much of these base pieces through the open-source side of Swift, which has been fantastic to watch. The next version of Swift (version 5.7 I presume) is likely coming next week, but even that doesn’t yet contain all of the outlined goals for concurrency within it. There’s clearly more to come – so where will it go from here? And how will it get used?

I have some background with distributed computing, back from the early days when OpenStack was starting to become a thing, several years before Kubernetes rose up to provide (in my opinion) even better abstractions, even if it stopped short in just the right way to bring with it a hell-scape of complexity. I’ve been programming using the Swift language recently, primarily to be delivered on Apple platforms, but I keep my fingers in the linux distribution, and a close eye on “server-side swift” in general. I see a huge amount of opportunity in distributed computing and the capabilities that it could provide.

With more hope than any real knowledge, I’d love to see a replacement API structure that’s swift focused and distributed across devices as a successor for XPC. Expanding the cross-process communication capabilities to cross-device, and the supporting infrastructure from Apple operating systems to identify and broadcast to each other, could enable some interesting scenarios. XPC has long been critical in terms of sandboxing processes, and generally enabling cross-process communications. It’s used all over the place in macOS and iOS, in the background where we don’t see it. Enable that across devices… and you’ve made the cluster of Apple devices that I have in the house more powerful because they could leverage each other and what they can individually do.

To me, the obvious starting point is Siri interactions. I have a HomePod mini in my kitchen, and usually more than one other device nearby. When I say “Hey Siri”, everything lights up and tries to respond, and one of them usually runs with the request – but sometimes several make the same attempt. I’ll ask Siri on my phone to activate a shortcut, but the HomePod grabs it – and because it doesn’t have the installed apps or intents, it says it can’t fulfill my request. Wouldn’t it be nice if Siri could know all the intents of all the active devices (where you have permissions, naturally) in the area where the request was made? That kind of thing is only possible when you have a robust and dynamic clustering solution that can let devices come and go, and share information about what’s nearby and what their capabilities are. In short: distributed computing.

Another place that would be useful is data collection and presentation with HomeKit. It feels dumb to me that each device and some iOS app has to manage the history of data for the in-house sensors I have installed. It’s making a silo of every sensor, and restricting my ability to even see, let alone use, the information together. The most effective solution to this today is cloud service integration, but the “advertising freemium” model that has dominance is primarily focused on extracting value from me, not providing it – so no thank you! Instead, I’d love to have a device (such as a HomePod with a bit of storage) stashed in the house that just collected the data from those sensors and made it available. A little, local persistent cache that all the HomeKit things could send to, and iOS apps could read from. A collection center akin to what Apple developed for health, and that doesn’t need to be stashed in a secure enclave. Imagine that as a HomeKit update!

Those two ideas are just based on what’s available today with a bit of evolution. My wildest dreams go full 2001: A Space Odyssey and imagine a rack where you can plug in (or remove) additional modules to add compute, storage, and memory capability. Make that available to other devices with something distributed, authenticated, and XPC like. It could expand the capability of small devices by offering progressively larger amounts of offloaded compute and storage available to devices that are getting lighter, smaller, and more invisible. In short, it’s a place where you could do the offload/tether that used to exist for iPhones to Macs, or watches to iPhones. If the current equivalent of those little white boxes in 2001 had the internals of the low-end current iPhones, it’s a truly amazing amount of compute that you could made available.

Yeah, I know that last one is getting pretty out there, but I love the idea of it anyway. I know it’s possible – maybe not practical – but definitely possible. I helped create something akin to that idea, based on the lessons from NASA’s Nebula Compute (but with a LOT less integration of hardware and software about a decade ago – we were trying to ride it all over commodity gear). That company failed, but the similar vision was replicated later by CoreOS, and a variation included within Canonical’s Ubuntu. I haven’t seen it really come to fruition, but maybe…

In any case, I’m looking forward to WWDC starting next week, the announcements, and mostly seeing the work a number of friends have been quietly working on for a while.

Published by heckj

Developer, author, and life-long student. Writes online at https://rhonabwy.com/.

%d bloggers like this: