Can cloud help Tube ride out Wi-Fi peaks?

Well-intentioned plans to provide travellers waiting on London Underground platforms with Wi-Fi access could unravel without robust ways of coping with the peak loads.

The announcement that Wi-Fi access will be available at 80 stations on the London Underground created a buzz, but this connectivity could have a greater impact on Transport for London's employees than on travellers.

In 2009-2010, excess journey time averaged only 6.4 minutes according to Transport for London's Travel in London report. Those minutes, probably spent waiting in stations, may not seem a lot of time to hook up to the internet and shop, grab email, or send yet another "I'm on the Tube" tweet.

Yet in terms of digital time, 6.4 minutes is forever and may lead to some unintended consequences given the volume of people who use the Tube every day, particularly in the morning peak hours between 7am and 10am.

Because the Wi-Fi service will only be available at stations it is likely to involve bursts of access to services, as travellers attempt to grab the latest news or download their latest email to keep them occupied on the next stage of their journey. These bursts can be problematic, especially if organisations and services are unprepared.

Unintended attack profile

The Travel in London report estimates that 379,000 people use the Underground during morning peak hours — a far cry from the roughly three million daily passengers, but still a non-trivial number.

Plans for Wi-Fi on the London Underground are welcome, but the 'bursty' nature of rush-hour requests will be hard to address.

Assuming some percentage of these peak passengers will certainly be taking advantage of wireless access during their average 6.4 minutes of excess journey time there is a very real possibility that the burst of connections suddenly emanating from London Underground stations could be seen as a DDoS attack.

Consider that if even four percent of those travellers accessed the same service at the same time, those 15,000 connection attempts would rival 2010's largest DDoS attack — measured at 15,000 connections per second.

Many services simply cannot handle that kind of load at one time, potentially resulting in a loss of service, as infrastructure and applications melt down in the face of overwhelming traffic.

While individual organisations are unlikely to see this level of activity coming from the Underground, they are likely to see daily spikes in access to email or secure remote access solutions as employees take advantage of the time to prepare before they arrive at work for the day.

These bursts in demand may have the same impact as an attack because...

...of the timing of the access and the restriction on access to Tube stations.

Services and organisations are unlikely to be able to justify the costs of provisioning additional full-time resources to meet such demand for only a small portion of the day — and yet failure to address that demand could be disastrous.

If a sudden spike in access to one service disrupts critical shared infrastructure along the way, other services will be affected with consequences across the entire user base — not just those coming from the London Underground. Such is the nature of shared resources — a failure hits all dependent services.

Cloud to the rescue

Cloud computing — the model — can be a boon to eliminating the risks associated with bursts of activity as well as mitigating concerns regarding costs associated with full-time resources for part-time services.

By employing a cloud-based model, computing resources can be redirected on-demand — or even on a scheduled basis as access bursts are anticipated — to meet demand without requiring additional dedicated resources.

The ability to provision and reprovision resources as services require (rather than dedicate those resources) gives organisations of any size the ability to make services more elastic and meet demand, effectively sharing the cost of those resources.

Cloud-bursting may also serve as a viable resolution to sudden bursts in access, especially if those bursts are predictable based on travel schedules.

Cloud-bursting today on-demand is technically possible but rarely feasible. However, predictable daily bursts provide the ability to spin up additional capacity in the cloud before morning peak hours and spin them back down afterwards on a schedule that allows for pre-positioning without the pressure of live migration.

The decision to enable wireless access only at stations along the Underground will have some interesting and potentially adverse effects on organisations.

Such a strategy also enables organisations to provision the capacity needed to deal with the added load without requiring resources to be acquired locally, where they may sit idle more often than not.

The decision to enable wireless access only at stations along the Underground will have some interesting and potentially adverse effects on organisations.

The bursty nature of requests that will come at specific times based on Tube schedules will be a challenge for organisations to address. Cloud computing models provide a good strategy to ensure service availability without incurring too many additional costs.

Lori MacVittie is responsible for application services education and evangelism at application delivery firm F5 Networks. Her role includes producing technical materials and participating in community-based forums and industry standards organisations. MacVittie has extensive programming experience as an application architect, as well as in network and systems development and administration.