What if you were friends with your servers on Facebook? What if you followed your servers on Twitter? What if you could just instant-message a command to a server?
It may seem silly to consider social networking as a big part of data center infrastructure. After all, social networking is usually something you do on your own time, under a pseudonym, while vehemently denying any affiliation with your employer. But the shrinking use of pagers in the enterprise, recent conference demos and even some shipping products all make me think the social network isn’t far away from infiltrating corporate IT.
Enterprise, Meet the Social Network
Arista Networks uses the Extensible Messaging and Presence Protocol, or XMPP—originally called Jabber, an instant messaging protocol—as part of its CloudVision switch management framework. Many enterprises are using XMPP-based IM servers internally so that humans can communicate with one another professionally and securely.
Arista Networks has enabled its switches to chat, too, by joining conference rooms on a corporate IM server. Interacting with switches is as simple as sending a command to one or more chat rooms they’ve been configured to join. This makes it easy to perform administrative tasks on thousands of switches in parallel, in a way that is secure, audited and natively designed to handle massive amounts of communication.
At VMworld 2012, VMware demonstrated using Socialcast to map relationships between servers and virtual machines. A server’s “friends” were the virtual machines running on it. When an event occurred in the cluster, a server posted a status update, and all other servers affected in the same way “liked” the update. It made it easy to identify the scope of a problem, which is often hard to determine quickly.
Similarly, by using hashtags, events and other system data can be collated quickly for consumption by IT staff. After all, consumption by humans is what all this log data is for, right?
In IT, monitoring systems sending pages has always been an expensive, kludgy and temperamental process, requiring special modems and subscription services to send extremely short messages. Pagers are also bulky and have enjoyed almost zero hardware innovation over the past two decades.
Now that smartphones are ubiquitous and have large or unlimited data service allotments, most IT staffers don’t want to carry pagers anymore. The problem is how to get messages to those smartphones. The easiest way to send a message from a monitoring system is through a phone carrier’s email-to-Short Message Service (SMS) gateway, but every SMS message sent incurs individual charges. A single oversized alert message sent this way is often broken into three or more SMS messages, each costing money.
What about Twitter? Every smartphone platform has a Twitter client. IT professionals often hang out in places with working Wi-Fi, so there is no charge, unlike with SMS. I’ll stop short of suggesting that your monitoring system post something on your Facebook wall, but why couldn’t it use Twitter to send you an alert? Unlike paging and SMS, the Twitter application programming interface (API) is well known and documented, there are thousands of clients, it is bidirectional, and it is quite timely for most applications. After all, users get upset when their friends don’t get back to them.
Given these examples, it doesn’t seem like social media and the data center are as incompatible as corporations may think they are. In fact, social media is built on things enterprises tend to value, like open and extensible APIs and nonproprietary standards. They also have some of the same goals, such as timely delivery of messages to thousands of subscribers. Certainly, some thought should be put into implementing these ideas—like making your monitoring system’s Twitter account private—but the flexibility and ubiquity of social media makes it highly likely that we’ll see these technologies in our data center soon.