It’s no secret that most IT services are either hosted in the cloud, or they are headed that way. In fact, the proliferation of cloud-based services in today’s corporate infrastructure environments will be looked back upon as one of the most compelling and important trends the IT marketplace has ever witnessed. While the massive advantages gained in terms of scaling, agility, and elasticity are impossible to ignore, cloud computing can (and typically does) introduce some visibility and control challenges. Whatever the flavor (IaaS, PaaS, SaaS, STaaS) or architecture (public, private, hybrid, transitional), Infrastructure and Operations (I&O) leaders are finding out that most cloud providers don’t provide (or even allow) adequate visibility.
Going forward, how will I&O leaders commit, measure and attain given user-experience SLAs as they have done in the past? How will they accurately assess utilization levels to determine over-subscription or under-subscription of capacity? How will they spot potential problems and bottlenecks as they are forming, versus after they have had a chance to fester and ultimately affect production traffic?
Dealing with the Opacity of Most Infrastructure as a Service (IaaS) Offerings
Typically, IaaS vendors tend to cloak activity ‘behind the curtain’, and assumes that you will be satisfied with vague, aggregate roll-up data of web workloads. You may have to take matters into your own hands. Therefore, consider tools that allow access directly into your core cloud providers via REST-APIs, Webhooks, or other means of integration. However, even the best tools won’t be up to the task if certain requirements aren’t met. First, you verify that your existing management and cloud vendors support these methods of data collection. Next, start persistent polling of sample transactions and workloads. Lastly, tie it all together. You’ll need to model these polled samples as they occur in your actual production traffic, incorporating each step, call, and hook to give a complete picture. This picture will show you exactly what components need to be monitored to give the visibility you need.
Filtering Out the Noise
Too much data is worse than not enough data. When presented with a tsunami of inbound notifications I&O personnel will go into ‘alert fatigue’ mode. They’ll begin to ingore important problems as well as the false alarms and redundant alerts. Therefore, it’s critical to keep the signal-to-noise alert ratio very high and avoid this syndrome. It is crucial to note that many isolated events are benign when they happen on their own, but when they occur in concert with other specific events, they can spell trouble. You don’t want to flood your team with these isolated and inconsequential events, so start to consider which need to be monitored as a collective grouping. Map out all the application dependencies, and strategically tie together related events via conditional, sequential or other patterned trigger statements. The net effect is to only receive filtered, intelligent alerts that mirror what your actual stakeholders are experiencing when transacting with your cloud-based applications.
Single Screen for All Elements
Keep the visibility that you are able to achieve into cloud-based resources front-and-center, stacked alongside other I&O metrics. If personnel must sift through various screens to get to cloud-specific content, then those metrics won’t get the attention it needs. Today’s cloud environments can be tricky to keep control of but putting the right visibility strategy in place can set you on the right path.