One of the hardest things about infrastructure planning in our web-enabled world is estimating capacity needs, especially from a network perspective. Having a piece of content go viral can mean the difference between having a functioning set of web infrastructure and a completely broken one. We've all seen what happens when web platforms get overloaded in the old media world - the launch of a new blockbuster stressing the online booking systems of a cinema chain, the release of tickets for a major festival causing the website of its ticketing provider to crash, a funny cat video propagating on email grinding a corporate network to a halt - and so on. In the world of social media and streaming 4K content, these sorts of phenomena spread even more rapidly, even more virally, and can have dramatic effects. As such, the enterprises, service providers and channel businesses delivering IT services need to get smarter about capacity planning to meet the traffic forecast.
The limitations of retrospective analysis
Typically, traffic reports for a service provider, content delivery network or large enterprise might come quarterly, or monthly at best. They will also include retrospective analysis showing the loads the infrastructure has been under over a period of time. In a given month, the network might have been under an average load of 20% but spiked for a day or two at 80-90%... this in itself would trigger warning signals, and perhaps suggest the need for capacity investments for the following year. Indeed, most infrastructure investments are booked at least a year in advance.
The cost of overcoming uncertainty with over-capacity
This in turn means that massive assumptions need to be made about possible needs in the year ahead. Yes, 20% run rate means we're fine for now, but that 90% spike means we might lack 'burst' resource in the event something goes viral or requires additional capacity for any reason. Typically, infrastructure managers will have over-capacity plans representing 50% or more resource than is required. This is down to the length of time it can take to get hardware out in the field; and, after all, the worst thing that can happen for an IT team in this context is to under resource - the productivity cost, never mind the potential reputational impact - is just too dramatic. If you're a large enterprise with business critical services, a service provider delivering into sectors with high service level agreements, or a channel business providing hosted or managed services - you can't afford the risk. It would critically damage your credibility with your customers. This is an industry-wide problem, as businesses and service providers find the resource costs of continued capacity planning and deployments needed to stay ahead of demand are escalating year on year.
Predicting future needs with real-time DNS insights
Real-time analysis of DNS requests, coupled with an exploration and analysis of historical traffic growth patterns via DNS data, can give much more granular assessment of what's going on, allowing you to model growth much more effectively. Historically this wasn't possible due to the volume of data in play, but modern, real-time DNS intelligence can play a critical role in capacity planning and in dynamic resourcing. After all, if the patterns of requests indicate a spike is imminent and it is flagged in real time, infrastructure teams can spin up temporary burst capacity, adjust load balancing, or otherwise refine their infrastructure provision to withstand the onslaught. And this can be done without quite the same scale of over-capacity investment... even a 10% improvement in capacity planning could translate to millions or tens of millions of savings across a year.
As CTOs and network teams continue to be challenged to do more with less, even while digital and web technologies become more central to the operations of a business, the marginal improvements this kind of insight can give will be key to delivering a competitive edge. CIOs must be creative in their use of DNS data to deliver the kind of value that's expected of them.