STE VE CASSIDY
Steve wolfs down analysis of apps consuming our bandwidth in lockdown, before providing advice on how to clear your videoconferencing line
Steve wolfs down analysis of apps consuming our bandwidth in lockdown, before providing advice on how to clear your videoconferencing line.
It has been a month of unexpected invitations. I’m afraid you’ll have to wait another four weeks to discover what probing questions I eventually manage to lay in front of Tony Blair, but like most events these days this will be over a remote video platform. I’ve even decided to create a dedicated “two-way chat over video” device, as this is going to be the way things will be for the rest of 2020.
While videoconferencing does have its positives – not least all those interesting invitations – there’s a conflict between the type of control required in a corporate conversation versus the needs of the journalist. In corporates, especially outside the English-speaking world, it’s a case of don’t speak until you’re spoken to. This supports the function of the “ringmaster”, bringing in questions or responses in a choreographed way, politely introducing the speaker and their publication. That’s not how technical news conferences work! We’re commonly dealing in complex and newly presented technologies, and questions flow freely and easily in most meetings because a technical writer understands the field: very often, their interjections help the presenter over an especially far-fetched analogy. It’s not always the case that every interruption is malicious or point-scoring. When Barry Collins and I sat opposite the “security CEO” of BT on the train back to London, the conversation was so intense in both directions that it felt more like a confessional than a sober, controlled release of information.
I realise that for many PR people, letting us loose on their quivering, nerve-wracked customers can seem like a gladiatorial combat session, but we not only make enemies but friends too. When I’ve been on trips with Mr Winder, the pair of us ask more questions per session than the rest of the press put together. Even allowing for our exceptionally bad manners in interrupting and clarifying what speakers are saying, it’s a sad fact of WFH life that the software is not working in our favour.
Personally, I’d suggest an interface stolen from Pac-Man, with each pending question sitting impatiently in a queue of icons, ready for their turn at the microphone. I don’t expect the role of ringmaster to go away, however, for a completely unrelated reason. That is, the revealed reliability of the internet, taken as an entire system. In a 15 or 20-person two-way session, I’d expect to lose two or three of those people to bad connections, over every hour of the presentation. If the target of the videoconferencing manager’s remit in running the call is to present a smooth, well-connected session and have everybody’s audio and video nicely responsive, they have to try to take control of “the edge”: that bit of the internet farthest from the data centres.
And the people living on the edge are us. As a buzzword, the edge has been around for a good few years in specialist sectors concerned with matters such as industrial computing, telecommunications, building controls and CCTV. Like a lot of computing terms, it’s a portmanteau term, mostly reflecting the idea that the real home of the internet is not up your lamppost or in a green painted steel cabinet whirring away at the roadside: the internet lives in data centres. Everything that exists outside of those places is loosely included in the term “edge computing”. While this can seem like a rather arrogant classification by the high priests of cloud computing, it’s a fairly accurate representation of the architecture most businesses are working with.
Power play
It continually surprises me how difficult it gets to measure stuff happening inside our computers. I went looking for a power meter a few months back but couldn’t find one I thought of as interactive or directly representative of the equation of “pop-ups per watt”. The closest I’ve been to seeing where the machine’s resources are going is using quite old PCs with low-power CPUs, which slows down the jiggling about of Task Manager’s app list when you sort it by memory or CPU use. Sadly, the things that are instrumented in Performance Monitor are much more serverfocused than they are useful to the interactive user (okay, it’s useful to see SQL Server performance even on an Atom-powered laptop, because it can turn up as bloatware in the most unexpected places… but that’s an exception to the general rule).
If I move out of the PC and on to the network, two things happen. One, I feel remarkably alone because this is not a common thing to do. And two, I get the idea that it might be quite difficult to unpack the stream of bits on the wire and reach some simple, headline conclusions about good or bad traffic, large or small sites, and so forth. I might be a little unfair about this but I remember seeing several brave announcements a few years back, claiming that small network-monitor boxes sitting beside your router were only a few short months of coding away. These all quietly vanished, because it turns out the problem is not nearly as simple as they think.
So if you can’t tell what’s hogging your machine, and you can’t see what’s hogging your network, and there’s no way to relate your experience inside an app to what you see anywhere on your home PC, then how can you figure out what to change to make things better? Enter, rather to my amazement, Uswitch ( uswitch.com). Yes, the service price-comparison site. It has been analysing its data from pre-lockdown to lockdown, along with the relative demands of popular entertainment services. While Uswitch has produced a selection of visually attractive images, it’s the less visual stuff, such as the table of “bad apps”, that’s more interesting.
I know what you’re thinking: how on earth could there be a way through the personal information and data-retention regulations that could deliver this grade of information, to a business whose entire raison d’être is its aloofness from the markets it analyses? What combination of datagathering kit and inter-company cooperation would be required for Uswitch to be able to see that (as one table shows) Zoom’s app is way out in front when it comes to increases in bandwidth consumption? Surely, as we have been told in decades of stock answers in technology forums all across the internet, all that data is encrypted?
The answer is about applying various filters and processes at each stage of the process of analysis. The first and most obvious exercise is anonymisation. Once an internet company has the log files of the servers its customers connect with most often, it can easily substitute junk, random strings for each subscriber name: suddenly you can’t see who was on Zoom or Disney or Hulu: all you can tell is that they
accounted for a lot of packets. Then there’s summarisation: making a pass across a dataset to make a table of figures means you lose tonnes of the contributory (and identifying) details. Of course, you then also lose all those sexy data-analysis features such as drilldown, but for an exercise such as this one the headline is the objective, rather than a forensic ability to get back to the original records from the headline number.
So, should we actually be concerned about the various apps bad-mouthed by their own traffic statistics here? Does it really matter that Disney comes somewhere around tenth place for change in bandwidth from before Covid-19 into lockdown? The first answer is, yes, it does matter if you’re on a data-limited contract. Profligacy by app developers and teleconference service providers is a privilege issue, because increasingly the cheapest deals for home connection are subject to bandwidth caps low enough to make the mobile data marketplace look comparatively generous. Suddenly, the number of videoconferencing calls you can make over the course of a month becomes a material limit to how many movies your kids can stream from one service or another. Go over the limit and you’re looking at excess payments. Now we’re looking at a change to salaries, with supplemental payments for homeworkers, dependent on a better and more manageable class of internet connection, and not put at risk of meeting failure because they didn’t have enough “internet credit” left in their account.
This came to me as quite a shock, I have to say, because I’ve been unmetered since the days of ISDN. I pay £70 per month for a one-gigabit fibre pipe, because the residential density around me is off the scale and a specialist provider decided to offer high-speed access here. The unholy marriage of bandwidth capping and devil-may-care app developers is going to be a major inhibition to successful adoption of working-fromhome and teleworker principles for the UK, and that’s why I’m grateful to Uswitch for its unexpected comparative efforts. We might be living on the edge, most of us, but that’s a passing label that now only reflects a historic, bygone way of working.
That edge thing
I’ve been looking at edge computing, probably without knowing it, for decades. Smaller local area networks, with servers and the ability to persist autonomously without the cloud to talk to, have been my stock in trade for decades. So I was amazed by a release from Amazon trumpeting the availability of an Amazon-branded edge server device, for deployment out in far distant corners of the planet where instant cloud access is suffering from a few challenges.
I’m afraid I can’t yet give a model number or a hardware specification, because for the next few months this particular object will be available on Amazon’s North American clusters only. But, to be clear, this isn’t for deployment inside Amazon’s data centres. It’s for you to have in your business or businesses, and it effectively extends the Amazon cloud environment right into the messy old real world, where people like you and me live.
For Amazon to give this idea a moment’s thought, a mere flash of consideration, there has to be something pretty badly wrong with the mass of packet-passing infrastructure stood between your business and its servers. I don’t believe it did this as a response to the changes in internet performance following lockdown - it’s too soon for those developments to hit the market. Edge is supposed to be industrial or maybe functional: not the sort of jobs that you would associate with an on-demand, scale-up cloud elastic product range. Nonetheless, I’m sure Amazon has found that certain developers have decided to use the architecture as if it had no limits and no escalating rate-card, and the best fix for that behaviour is to cheat and put more compute closer to the cause of the problem.
This makes me think that the state of the edge could be the driving force behind winner and loser products in 2020/21. At the moment, homeworking users are the canaries, indicating the suppliers and locations with poor service (Uswitch says the worst in the UK is Southampton, followed by Belfast and Liverpool). But soon, if you’re 100% cloud resident and already chafing at the restrictions of your connection, you’d better start thinking about how to circumvent sub-par edge equipment getting in the way of your business, because it’s all a bit too large and widespread and of varying ages to be handily upgraded just because someone complained to their MP.
“The state of the edge could be the driving force behind winner and loser products”