Real-time cloud monitoring too challenging for most providers, TFL tech lead says
Getting solid data on what’s happening in your application in real-time seems to be a fairly big challenge for most cloud services providers out there explains Simon Reed, head of bus systems & technology at Transport for London (TFL).
TFL, the executive agency responsible for transport planning and delivery for the city of London, manages a slew of technologies designed to support over 10 million passenger journeys each day. These include back office ERP, routing and planning systems, mammoth databases tapped in to line-of-business applications as well as customer-facing app (i.e. real-time travel planning apps, and the journey planner website), line-of-business apps, as well as all the vehicle telematics, monitoring and tracking technologies.
A few years ago TFL moved its customer facing platforms – the journey planner, the TFL website, and the travel journey databases – over to a scalable cloud-based platform in a bid to ensure it could deal with massive spikes in demand. The key was to get much of that work completed before the Olympics, including a massive data syndication project so that app developers could more easily tap into all of TFL’s journey data.
“Around the Olympics you have this massive spike in traffic hitting our databases and our website, which required highly scalable front and back-ends,” Reed said. “Typically when we have industrial action or a snowstorm we end up with 10 to 20 times the normal use, often triggered in less than half an hour.”
The organisation processes bus arrival predications for all 19,000 bus stops in London which is constantly dumped into the cloud in a leaky-tap model, and there’s a simple cloud application that allows subscribers to download the data in a number of formats, and APIs to build access to that data directly into applications. “As long as developers aren’t asking for predictions nanoseconds apart, the service doesn’t really break down – so it’s about designing that out and setting strict parameters on how the data can be accessed and at what frequency.”
But Reed said gaining visibility into the performance of a cloud service out of the box seems to be a surprisingly difficult thing to do.
“I’m always stunned about how little information there is out of the box though when it comes to monitoring in the cloud. You can always add something in, but really, should I have to? Surely everyone else is in the same position where monitoring actual usage in real-time is fairly important. The way you often have to do this is to specify what you want and then script it, which is a difficult approach to scale,” he said. “You can’t help but think surely this was a ‘must-have’ when people had UNIX systems.”
Monitoring (and analytics) will be important for Reed’s team as they expand their use of the cloud, particularly within the context of the journey data TFL publishes. Reed said its likely those systems, while in a strong position currently, will likely see much more action as TFL pursues a strategy of encouraging use of the data outside the traditional transport or journey planning app context.
“What else can we do to that data? How can we turn it around in other ways? How can other partners do the same? For us it’s a question of exploiting the data capability we have and moving it into new areas,” he said.
“I’m still not convinced of the need to come out of whatever app you’re in – if you’re looking at cinema times you should be able to get the transportation route that gets you to the cinema on time, and not have to come out of the cinema listings app. I shouldn’t have to match the result I get in both apps in order to plan that event – it should all happen in one place. It’s that kind of thinking we’re currently trying to promote, to think more broadly than single purpose apps, which is where the market is currently.”