Change.org, the platform behind all those online petitions — the justice for Trayvon Martin petition is one recent example — uses a ton of technology, just not a lot of it in-house.
The company now claims it serves 40 million users, up from 20 million from a year ago and 25 million last January, and it still doesn’t have a server in sight. Well, that’s a slight exaggeration. “We have Meraki(s csco) running our office internet in one server closet,” says CTO Tom Hughes-Croucher, a Yahoo and Joyent veteran, who’s been on the job for about 8 months.
Now, that 40 million number is highly variable — it’s not like all of those people “live” on change.org — but it’s still impressive and the company says it supports 20 million unique visitors monthly and adds 3 million users a month. No matter how you slice them, those are big numbers.
Now, the organization which fields 185 people worldwide (but won’t disclose how many serve an IT function) is an extreme example of a born-and-bred-in-the-cloud company. Its highly variable workload is ideal for running on highly-variable rented resources from Amazon Web Services(s amzn) and Rackspace(s rax). More traditional companies with steadier workloads often say that AWS is a great place to start but, once the applications are designed and running, it’s often more cost effective to bring them in-house.
Not so with Change.org, which runs a ton of compute jobs using AWS Elastic Map Reduce (EMR) and is starting to move more to AWS’ RedShift data warehousing service. It also uses a lot of “bare metal” capabilities at Rackspace.
One of the facts of life at Change.org is that some of the workload spikes are entirely predictable. It knew, for example, that the Trayvon Martin petition was coming well in advance of its launch, and could plan accordingly. And it sees fairly typical traffic patterns. It sees waves of users coming online at lunchtime PDT, dinner time EDT and evening hours European time. But then there are also the surprise spikes — when something unforseen goes viral. Spokeswoman Charlotte Hill said a petition asking the Malaysian government to work with the United Nations to validate the recent election caused a huge spike at midnight to 4 a.m. PDT.
And that’s why it’s key to stay agile and make best possible use of cloud resources.
Hughes-Croucher, by virtue of time spent at Yahoo where he was a front-end engineer working on the home page and then a tech evangelist, has seen IT from all sides now. “At Yahoo, we had incredibly large server deployments and if we needed to we could just switch data centers. Now, he feels Change.org has to be as fast as Google, as scalable as Amazon — but without its own data centers. “That’s our challenge if we want to be successful; we have to pitch like the big guys.”
And, in his view, Change.org’s use of these cloud services makes that possible.
Last September, when I spoke with Kyle VanderBeek, Change.org’s manager of infrastructure, he talked up its server-free status but also left the door open for more in-house IT. Indeed, he said he might buy some servers in fairly short order.
Apparently that didn’t happen. Maybe next year.