Discovery Scalability...

[SlightlyLoony]

A perceptive customer couched a question to me in an interesting fashion. She noted that Discovery populated the CMDB automatically, which was essentially the same thing as hiring enough people to do the same thing. She estimated that her CMDB would have about 20,000 items in it. To re-enter all those items weekly (as she'd like Discovery to do), she'd have to hire 80 people who each could enter 50 CIs per day. Cost and practicality aside, she noted that hiring those people would put a heavy load on her Service-now instance — and she wondered if Service-now could handle that kind of load, and if so, how? Would we just add lots and lots of boxes, like the fellow at right trying to scale up his wireless capability?

Good question. But of course we can, and without all those cables and antennae! Here's how we do it...

You probably know that we host Service-now for a wide variety of customers and needs. Our smallest customers run on hardware that's shared with several other smaller customers. As a particular customer's use of Service-now increases, we'll move their instance to hardware shared with fewer customers. If the needs increase further, we'll move the instance to its own hardware, and then to hardware of increasing capability.

All I've described so far is standard hosting strategy; there's no magic involved. But what if a customer's needs increase to the point where a single computer system can't handle the load?

In those cases, we take advantage of a capability built into the Service-now product from its earliest days: the ability to add independently-hosted nodes to form an application cluster. As I write this post, the largest such cluster I know of has four nodes running on moderately powerful Sun hardware. The four nodes share an Oracle database, which is itself clustered on two Sun computers. This Service-now cluster is for an organization with over a half million devices and over 50,000 employees, and they use Discovery.

We haven't hit the "scalability wall" yet, and we're looking forward to meeting the next challenger...