Cloud migration comes with numerous challenges, from picking the right architecture to service orchestration and network automation. However, enterprises often overlook the change needed in the operationalization of the cloud, especially from the perspective of monitoring, troubleshooting and fault handling. In the rush to the cloud, how do you ensure that you maintain operational visibility?
Understanding the Cloud Connectivity Environment
Migrating workloads to IaaS platforms or adopting SaaS applications means that the Internet is the underlying communication and connectivity engine for service delivery. The dynamic nature of Internet connectivity is a fundamental shift compared to relatively static enterprise network architectures of the past. The connectivity path across the Internet involves multiple ISP networks that are not directly under the governance of enterprises. SaaS delivery architectures add another layer of complexity to the connectivity matrix. Connectivity is also impacted based on end-users’ geographical placement. Across all of those elements, Internet routing updates can cause path changes with no warning. With the cloud, which from a connectivity point of view is synonymous with the Internet, any provider, network, service or location can impact end-to-end performance, availability and user-experience.
Debunking Cloud Monitoring Myths
Given this new connectivity environment, enterprises need to move past unrealistic expectations and squash cloud monitoring myths to effectively operationalize their cloud deployment.
Myth #1: There is no need to prepare for the cloud
One way of dealing with the uncertainty of the Internet is to simply ignore it as a factor. Contrary to popular belief, there’s no steady state in the cloud. In an ecosystem that’s constantly changing, proactive preparation holds the key to successful execution. Baseline performance pre-deployment. Proactively identify and establish performance metrics that are pragmatic and relevant to the cloud. Continuously monitor, validate and fine tune when necessary. Prevention is better than cure.
Myth #2: Troubleshooting the cloud is not my problem
Enterprise IT teams may be tempted to feel that, because they don’t own the cloud, they don’t have to worry about troubleshooting the cloud. Network engineers already spend the majority of their troubleshooting time identifying where a problem lies. In the cloud, this dynamic is compounded. Even though you don’t own the networks, infrastructure and apps, you still own the user-experience. In many cases, you’ll have to identify one of many possible organizations as the “owner” of the root cause issues.
Myth #3: If I can find it, I can fix it
Assuming you’ve identified where the problem lies, next you must try to resolve it. Even if you have a clear escalation process with your third party providers, fixing the problem is well beyond your control. This means you need to present evidence that you can effectively escalate to that third party and drive problem resolution. Fault resolution in the cloud tends to exacerbate the gap between troubleshooting and resolution time.
Why Traditional Network Monitoring Breaks in the Cloud
Traditional network performance monitoring (NPM) and application performance monitoring (APM) solutions fail to fully meet the demands of today’s distributed and highly dynamic cloud architectures. Traditional NPM techniques flatline outside the perimeter of the enterprise, while APM code injection loses its relevance in the world of SaaS.
These solutions tend to be siloed because both perform data collection and correlation tuned to the perspective of the particular app and network operations teams that maintain and deploy them. While both of these techniques are valuable for enterprises that own the end-to-end infrastructure and applications, they fall short in cloud and SaaS environments. The APM and NMP solutions you’ve previously relied upon leave a visibility black hole in your cloud universe. To be successful in the journey toward cloud migration, enterprises need new sources of monitoring data.
Five Tips for Cloud-Friendly Network Monitoring
- Take an active monitoring approach to understanding cloud apps and services. Active monitoring utilizes simulated transactions and requires no passive data collection or instrumentation, which makes it a perfect monitoring technique when you don’t own the application, infrastructure and networks. Unlike packet capture and flow-based monitoring, active solutions can provide useful app and network-layer data outside of the boundaries of your enterprise.
- Move network performance monitoring to the left on your cloud migration timeline. This is where active monitoring starts helping by getting visibility ahead of deployment. For example, with VoIP monitoring and UCaaS adoption, you can perform pre-deployment benchmarking with active monitoring, whereas passive monitoring via CDRs and packet capture won’t help you in pre-deployment since there’s nothing yet to measure.
- Cloud monitoring solutions need to be collaborative. Sharing data with and across vendors eliminates finger pointing, builds trust through transparency, and speeds resolution time by providing hard evidence to escalate to third party providers.
- Remove application infrastructure dependencies from your monitoring stack, to avoid cascading failures. For example, if you have workloads in AWS, you should monitor them from another infrastructure outside of AWS.
- Combine various types of monitoring data from multiple vantage points. Application performance is paramount, but when it’s correlated with the vagaries of the underlying network infrastructure and path, the combination yields far more than the sum of the parts. Active network monitoring is best when paired with passive techniques like BGP routing protocol and route table monitoring that can enrich understanding of Internet paths that carry cloud traffic. When you can get that rich, correlated view from multiple, geographically diverse monitoring vantage points, such as your branch locations or geographies that represent your remote users or online customers, you can triangulate issues more quickly.
Enterprises are increasingly finding themselves in the midst of cloud migration without having fully thought through the intricacies of managing an external network, service and code dependencies. Make monitoring and visibility a higher priority early in the cloud lifecycle. Change is constant in the cloud, so once you have access to the right data, continuously adjust your performance metrics to match that change. By mixing the right data with realistic expectations, you can create an effective network monitoring strategy for the cloud.