The computing power needed for many business needs is stable. That makes it relatively easy to forecast what server and network capacity you'll need. There are some industries, however, where demand can be unpredictable and skyrocket in seconds, before dropping back just as quickly. It's a challenge faced by retailers running seasonal sales or ticket agencies releasing concert tickets for an artist who rarely performs. It's also a problem faced by many digital business when they launch. They simply don't know what the demand is going to be. Becoming a viral success could be as much of problem as sinking without a trace.
Until recently, the answer to dealing with spikes in traffic, such as Christmas, was to put in place enough server and network capacity to handle the expected peak demand. The downside was that you'd be paying for most of that capacity to go unused for most of the time. The lead times for sourcing, configuring and deploying new hardware were simply too long to make adding servers only when things got busy into a viable option.
All that's changed with the advent of "serverless computing". It's a way of providing cloud-based infrastructure that hides the detail of the clusters and nodes running your applications. All you see are application-level constructs, while the cloud provider takes care of everything about the underlying server infrastructure, from monitoring and logging to OS management, rolling out security patches and load balancing, as well as spinning up servers (and closing them down) to match your current load.
With serverless computing on the Google Cloud Platform, new servers – configured exactly the way you need them – can be spun up automatically in less time that it would take you to notice there's been a spike in traffic. When demand drops, they'll be closed down just as quickly. You'll only be paying for the servers and capacity you use: if an application doesn't have any load, it won't be consuming resources, and you won't paying for them.
Music discovery service Shazam is one company that's taking advantage of serverless computing for both real-time and batch applications. It's using GCP to handle spikes in demand, such as when a new song is first played on the radio or during events like the Eurovision Song Contest. It's also finding serverless computing a great solution for batch processes, including regular re-indexing of its data.
If you want to find out how you can take the headache out of handling unpredictable demand for computing power, while cutting your costs, come and talk to our GCP experts about serverless computing.