One of the things we’ve observed from the Internap customer base is that businesses with real-time big data applications often start out on a cloud platform – typically a virtual cloud platform – and then essentially outgrow it. Whether it’s in terms of performance needs or even in terms of overall price performance needs, they tend to gravitate to our bare-metal cloud offering.
Gaming providers are great examples. If you’re an online game publisher, you live and die by latency on one side, and you live and die by the competition and resource intensity on the other side. And if you’re a big game provider you probably have massive big data use cases where you’re mining the data and running Hadoop kind of jobs.
This same kind of scenario exists in the advertising technology world. Particularly with mobile ad tech, location-based ad delivery is pretty compelling as a business model, but again you’re making real-time decisions based on real-time data from essentially millions of inputs.
Another kind of use case could be where you’re running big data jobs, and Hadoop tends to be the most common one. Typically, your performance needs are great, but the needs themselves are sporadic and unpredictable. You may have a Hadoop job that you run to mine a bunch of data, and it may take 3 days for you to run that job, and you don’t need the infrastructure for 3 years or 3 months, you just need it for 3 days.
What our customers find is that when they move to the bare-metal servers – an elastic cloud model – they get 3 to 5 times better performance. And by the way, they’re paying 40% less on average than what they were paying for similar compute needs from a virtual cloud environment. Essentially, they find that their monthly bill has been cut while they have immediately been able to scale the application to their desired performance needs.
To learn more, download our white paper, Stop paying the cloud performance tax: Go beyond virtualization with bare-metal cloud.