Beyond the Data Center: The Edge and the Cloud
Data Center Maintenance
When it comes to the edge, it’s all about the zettabytes.
Most IT pros don’t like to be trend chasers, at least in the sense of Instagram fads, but the IT trade press routinely foments debate about which technology is beating the latest incumbent. Horserace coverage aside, these stocktaking exercises are helpful in predicting the advancements worth adopting.
For some time, discussion has centered on the decline of the data center. At first, many seemed to expect the cloud to subsume all IT functions, and the corporate data center was being given its last rites. Then emerged a more nuanced view, in which the cloud is good for some things but on-premises IT and hybrid public/private cloud capabilities are necessary for others.
Now, new lines are being drawn, this time pitting edge computing against the cloud. We’ve heard from Gartner pundits that “The Edge Will Eat the Cloud” and from venture capitalists that “Edge Computing Will Blow Away the Cloud.” What’s going on?
First, What is Edge Computing?
When it comes to the edge, it’s all about the zettabytes. Cisco predicts that by 2021, various devices—from phones to IoT sensors—will store more data than our data centers, four and a half times as much! It’s a ton of information to be pushing over internet connections, and edge computing promises to deal with the fallout. It does so by moving some of the data processing away from a centralized cloud and out toward the network’s fringes. Raw data is processed much closer to its source; i.e., whatever sensor, cell phone, or smart refrigerator originally created it.
This reduces latency in ways that enable devices to work when realtime or near-realtime decisions are required. Think self-driving cars, where a moment’s lapse could cost lives. The edge will also reduce the sheer volume of data running around corporate networks, hopefully keeping it at manageable levels.
The Limits of Edge
Few people believe that all data processing will happen at the edge, however. As much as driverless cars will need to “think” and “interact with” other cars without constantly consulting centralized resources, developers still expect to collect, process, and aggregate data in the cloud.
The cloud will remain the most efficient site for non-time sensitive processing tasks and big data analytics that build knowledge out of the bits and bytes provided by hundreds or thousands of disparate devices. Thus edge computing is better thought of as a complement to existing cloud technologies.
In considering where and how to deploy edge computing, it’s vital to keep the possible pitfalls in mind. Purpose-built edge applications for managing a facility’s physical security, for example, can detect and respond quickly to alarms. Such systems already exist, and they make a lot of sense. On the other hand, putting a retail chain’s inventory controls at the edge—meaning on point-of-sale (POS) devices—could leave them more open to attack. And there’s really little reason to do it, as latency is not a key concern for inventory management but cloud-friendly data analytics are.
At present, good edge implementation requires a specific and narrowly focused goal and involves purpose-built systems to fulfill specialized needs. Cloud computing is a more general-purpose platform that can also aggregate data and implement higher level business logic. It’s also great for media and entertainment, backup and disaster recovery, and more. Keeping their relative strengths and weaknesses straight will enable better architecture and technology adoption decisions.
The Balance of Power
IT pros who have been around the block a few times have seen this happen before. We went from centralized mainframes to more distributed client-server applications to centralized cloud processing and maybe now out to decentralized edge computing. From this history, we can be pretty confident that edge computing will not destroy the cloud, but there will be changes. Such providers as Amazon are already adapting their offerings to stay ahead of the curve, a good sign forward-looking enterprise IT leaders should do so as well.