Scaling a service can be a double-edged sword. If your app, your game, your online store or your next generation payment service starts to take off, the business team is popping champagne, while the operations team is often scrambling to scale the service out.
Last Sunday was the season finale of Silicon Valley on HBO (it’s a fantastic show, and if you are not watching it, you should start!). In that episode, the fictional “Pied Piper” struggled while their rather boring “Condor Cam” went viral – thanks in part to Manny Pacquiao (I’m not going to explain any more, I don’t want to ruin the plot).
Suffice it to say, traffic on their site spiked, and they scrambled to scale the service to handle hundreds of thousands of unexpected users. At one point they were jamming circuit breakers so they wouldn’t trip, punching holes in walls for new cable runs, and even letting small electrical fires burn uncontrollably, all in an effort to scale the site.
Of course, that’s television, and they were able to keep the site running until the traffic dropped off. But it’s not too far off from the real world. While most bootstrapped startups are leveraging cloud-based servers instead of homemade racks in a garage, they still struggle to scale their own services when these bursts happen – and it’s hard enough to scale out your own services, let alone your 3rd party analytics platform.
The other day, one of our large customers started sending us data from a new cluster of servers. A lot of new data. They went from their usual ingest rate of 10TB/Day to a rate of 20TB/Day. If you are unfamiliar with the machine data analytics space, I can tell you that’s a massive increase.
In a typical on-premise implementation, doubling ingest in an environment of this size would cause chaos. The admins would have to stop collecting the data while they re-designed the architecture to support the new load. Once the design was finalized, network equipment, storage subsystems and servers would be ordered, while real estate in the data center was identified and racks installed. It will take weeks, but more commonly months, to scale this environment out. But the business doesn’t have months to scale – it needs that data right away.
The only reason our customer was able to scale the way they did is because Sumo Logic was purpose-built to handle exactly this kind of use case. It is a true, multi-tenant, elastically scalable, cloud-based analytics service – the only one of it’s kind. The customer simply installed our collector on those new servers, and started sending us data. No architecture planning, no server buys, and no datacenter expansion was required.
So, can your analytics platform do this?