Multicloud: the new monitoring silo

Enterprises are embracing cloud platforms, often having more than one provider to best suit the given application. But how to monitor across this hybrid environment?


Most IT professionals would agree that cloud computing is more about a paradigm than a physical displacement of computing resources. The “utility” paradigm is transformational by virtue of on-demand, elastic resources. But unlike being stuck with just a single local cable provider, the range of choice in cloud computing options also provides selectivity and customization best suited for the delivery of individual applications.

Research from Edwin Yuen at ESG suggests enterprises have adopted multiple public clouds for production workloads at a remarkably high rate. His research says 81 percent of current cloud infrastructure users are running two or more public cloud infrastructure providers. Of those users, 61 percent are using three or more. The greatest driver for multicloud utilization is application or workload specificity.

As companies adopt multiple cloud providers to support individual applications, management increasingly becomes a difficult task of specialization based on cloud/app pairing. The resulting silos may be monitored for performance and reliability by the native tools in a cloud stack, or from an application-oriented view via APM.

Hybrid cloud in a devops world

But if the devops movement in IT has taught us anything, it’s that silos are antithetical to agile practice and create brittleness. In the case of operations management, the brittleness manifests in the shallowness of native tools. While most of the dominant public cloud infrastructure providers offer logging and metrics services, none is especially strong when it comes to the depth of data, its context relative to the rest of the stack, nor presentation in terms of dashboards or alerts. For example, Amazon Cloudwatch provides 16 metrics for EC2 Container Services (ECS), whereas third-party tools offer greater insights through dozens of additional metrics pulled from secondary AWS APIs and tertiary application APIs.

The lack of comparative visibility among clouds can also lead to blind spots. Assessments about performance, fit, and cost made during cloud migration quickly become invalid as apps change, as do their requirements for resource utilization. How can you tell if the cloud of choice for your apps yesterday remains the same today?

To break the silos, you need to think differently about the way in which you manage and monitor across clouds. You need to first start by decoupling data collection from the data science and performance analysis. This powerful concept lets you manage and monitor your multicloud infrastructure in an agnostic light.

A standardized, service-oriented approach to system data collection enables:

  • Constant technology currency.
  • Comprehensive coverage across all layers of the IT stack.
  • Data normalization across disparate metrics.
  • Contextual metadata—or data dimensionality—needed to support advanced component dependency-aware analytics.

This next-generation data collection approach is often delivered through a monitoring- integration-as-a-service (MIaaS). MIaaS is an emerging set of tools for connecting monitoring platforms that are deployed in different IT environments, so they can analyze the health and performance of the entire ecosystem. MIaaS is often used by large enterprises that need to integrate on-premises infrastructure performance metrics with cloud service-performance metrics or to monitor multicloud environments.

Here are five multicloud management strategies where a MIaaS across multiple public cloud providers pays off:

1. Run entire production applications or workloads on Azure and/or AWS, including customer user interfaces, compute and data processing/archival, and storage.

Advantages of MiaaS: Empower Central IT with a topology view (see all objects) of workload distribution across clouds on their preferred monitoring platforms when cloud-hopping.

2. Use clouds for your development, test, and failover/recovery workloads and operate production applications using on-premise software and infrastructure.

Advantages of MiaaS: Manage alerts in one place, including those generated by AWS, Azure, or on-premises datacenters.

3. Use cloud only when your on-premises applications/servers need additional compute, network, and storage during periods of peak demand (bursting).

Advantages of MiaaS: Enable faster incident response in the event of loss of cloud service and informed strategies to bring services in-house or switch clouds.

4. Use multiple clouds simultaneously—use Azure for SaaS and PaaS and use AWS for IaaS. Use these clouds during early release cycles as features are evolving and when demand spikes.

Advantages of MiaaS: Consider the risk of making capacity planning calculation errors (moving workloads) due to context switching between Azure Log Analytics and CloudWatch—no single pane of glass.

5. Use cloud to merge/spin off a new business unit or buy a company and let the unit run workloads/applications on its choice of cloud.

Advantages of MiaaS: Give central IT full visibility across heterogeneous stacks from new business unit—eliminate finger pointing between central IT team and siloed IT/management teams from new unit acquisition or spinoff when a problem or a service issue surfaces—this increases recovery times (MTTR).

Once again, the new paradigm provided by cloud computing brings its own set of new challenges, particularly how to monitor multiple applications effectively across this hybrid environment. The industry must evolve to meet these new challenges.

This article is published as part of the IDG Contributor Network. Want to Join?