Going back five years before cloud computing was really a part of our vocabulary, there was a degree of tension in most IT departments that often included two factions that butt heads. In the red corner were your developers focused on pushing out code, evolving from waterfall development to agile methodologies that were all about speed and time-to-market. In the blue corner were the System Administrators or “Sysadmins” - those who had the power to say “no”, or I “don’t have enough”. Neither corner understood what the other did (or in many cases, wanted to understand). As a result, this led to some catastrophic failures and some really epic finger pointing in corporate IT.
If we look at the role of the Sysadmin, it fundamentally includes two responsibilities; (1) provision the infrastructure for IT projects - including procurement, logistics and actual provisioning; and (2) once in production, feed and water the infrastructure to ensure the developers app stayed up, which in most cases are wrapped with some SLA/OLA.
Cloud has turned this on its head. Now the Sysadmin and developer roles have merged. Let’s review this example. A developer needs ten servers stood up for a new app. In the cloud, the developer simply goes to the provisioning interface or reaches for an API, and with a few clicks or calls, ten servers now exist. Provisioning time = 3 to 5 minutes/server; experience needed = very little; the outcome = pretty much unlimited capacity.
Since the process is now so easy to provision, the next step is monitoring. We are seeing many frameworks to take this cloud approach to monitoring. These tools replace the Netcools and Openviews of the world with simple web interfaces and the ability to subscribe (RSS/ATOM) or poll (RESTful) to find the health of a platform, the combination of a virtual machine, and an application.
This trend can be seen in full effect in the web2.0 community. I spent some time a few weeks ago with a large photo/video sharing shop. The folks who ran the infrastructure joked about how they had never been in a datacenter, because they really had no need to. IT to them was simply an API call away. Servers, storage and how many cores was not a concern.
So does this scare us? Absolutely not! Quite simply, cloud is an evolution of hosting and we have really always had devops. If a customer wants a new server they open a ticket in a portal, servers are moved from inventory to production, racked and stacked, OS installed and handed over. Cloud has optimized this workflow dramatically and removed much of the manual provisioning – and selfishly speaking, has allowed us to reach billable events much faster.
So who should be afraid? The corporate IT folks who fought to keep applications inside their four walls. As devops becomes defacto, they now need to either invest and build private clouds to sustain this, or step aside.