Why observability is important in multi-cloud environments
Increased use of multi-cloud environments is creating a need for specialized observability methods and tools for tracking and measuring cloud performance.
The days of companies going all in on the public cloud are largely over, and many enterprises still keep much of their critical infrastructure on premises. Moreover, companies that have maintained substantial investments in public cloud are spreading their risk across multiple providers due to factors such as cost savings, the desire to take advantage of innovations from competing providers and hedging against lock-in.
Today, the notion of cloud has become malleable. Cloudlike environments that provide benefits such as scalability and consumption-based pricing are increasingly common in on-premises locations controlled by the customer, helping to create a common experience across locations. Public cloud providers now also offer on-premises versions of their environments that serve as extensions of the public environment. The term multi-cloud has become a common way to refer to these multifaceted offerings.
But the paradox of choice looms large over this market. As customers take advantage of more flexibility when designing their multi-cloud strategy, the challenges involved in monitoring these mixed environments for security and performance are mounting. This is where a comprehensive set of observability technologies should come into play.
What is a multi-cloud environment?
Multi-cloud refers to the use of more than one public cloud service by a customer, as well as hybrid cloud deployments that combine on-premises cloud software infrastructure with public cloud services. Moreover, public cloud vendors have moved to recreate their native experience inside customer data centers with products such as AWS Outposts.
The reasons enterprises adopt multi-cloud have to do with factors such as cost management and access to a wider range of proprietary services. While some enterprises have based their multi-cloud strategy on a provider's geographic ubiquity in the interest of data sovereignty laws and lower latency, the coverage gaps between hyperscalers and other public cloud providers have narrowed in recent years. Providers are also developing smaller form-factor data centers that enable rapid expansion into new regions, particularly via partnerships with colocation providers. There is also a growing number of emergent public cloud vendors, primarily focused on core services, such as compute and object storage, that provide additional coverage options, often at a significantly lower cost.
In the early days of public cloud, many enterprises saw these platforms as targets for lift-and-shift deployments -- essentially, rehosting an on-premises application using public cloud VM infrastructure with few changes. Later, more sophisticated and ambitious approaches involved refactoring an application to not only live on the public cloud, but to take advantage of its native services. Refactoring also gives an enterprise the opportunity to dig into an application's core code to improve efficiencies. A third step sees companies overhaul an application's structure -- or write a new application -- with a distributed, microservices-based architecture, which provides scalability and other benefits over older, monolithic application designs. While many distributed applications live in a single public cloud provider's infrastructure, multi-cloud offers additional opportunities for distributed application development.
What is observability and what does it mean in the cloud?
Observability builds on traditional application performance monitoring (APM) tools. The two are often used in tandem, but each emphasizes a different core area. APM tools tend to focus on how an application's behavior affects the user experience, serving up alerts and error messages when it detects potential problems. Meanwhile, observability tools employ three core methods: collecting log data for use in historical analysis and troubleshooting; performance metrics; and traces, which analyze an application process from start to finish. The most comprehensive tools cover the entire application stack, from core infrastructure to databases, application servers and client UIs. Ultimately, the goal of a successful observability implementation is to provide enterprise teams with a holistic view of an application's overall health.
The importance of multi-cloud observability
As discussed, there are many reasons for enterprises to pursue multi-cloud and hybrid cloud strategies, and in some cases, they may have no choice, due to factors such as regulatory compliance and required low latency for critical applications.
As companies expand their use of multi-cloud -- and seek to integrate the services with on-premises assets -- it grows increasingly critical for them to maintain a unified view of these locations and the assets running on them. The automation and efficiencies provided by the public cloud can be undermined if DevOps teams find themselves struggling to consolidate and maintain an accurate, ongoing view of system performance, costs, security and other factors.
Common multi-cloud observability challenges
While major public cloud providers all offer a set of observability tools, in many instances, they are understandably oriented mostly around their own systems, rather than serving as general-purpose platforms that can plug into any third-party environment as well. This presents enterprises that pursue a multi-cloud strategy with a potential conundrum, as teams must contend with a disparate set of observability tools that may not use the same standards and conventions, such as for logging and tracing. It's also possible that aspects of a company's multi-cloud posture were not well planned, with disparate teams entering contracts with different providers, thereby building up assets that then have to be included in the observability plan.
Observability tools can generate massive amounts of data, which is a good thing in terms of giving enterprises the most sophisticated view of their systems, but it can also be detrimental if not throttled carefully, given the costs of storing and securing the data. Moreover, enterprises that create a central repository for observability data can incur egress fees from public cloud platforms. However, egress fees are becoming less onerous in recent years, due to efforts such as the Bandwidth Alliance, an industry consortium of cloud platforms.
More recently, in response to pressure from regulators, AWS, Microsoft and Google announced that customers who wish to migrate data out of their platform and to another provider or an on-premises location would not be charged. While this was a welcome move for customers that want to leave a provider, the policy shift doesn't necessarily have a positive impact on those who are maintaining multi-cloud implementations and conducting a multi-cloud observability strategy.
What to look for in a multi-cloud observability tool
Many observability vendors position their products as full-stack tools, tracking aspects such as core infrastructure, applications, databases, application security and microservices. Some also target additional domains, such as business analytics. While they're often sold as a suite, it is possible to purchase individual modules in many cases. Also, the observability market has seen a fair amount of consolidation in recent years, so it's important to consider how well a vendor has integrated any acquisitions it has made. Prominent multi-cloud observability vendors include Dynatrace, Splunk, Grafana, AppDynamics and New Relic.
As with any enterprise software, multi-cloud observability platform buyers should conduct a thorough evaluation of each candidate to ensure it provides an intuitive user experience. One way to think about it is that, while a multi-cloud observability platform must provide users with enough knobs and buttons to achieve sophisticated analyses, it should not overwhelm them. Also, multi-cloud observability vendors tout their platform's AI and machine learning (ML) capabilities. Seek proof points of their efficacy for both general-purpose and domain-specific needs. Moreover, evaluate vendors who have a focus on observability for AI and ML applications themselves.
Meanwhile, enterprises may seek out a single multi-cloud observability platform for most or all their needs but should consider the role of native observability services associated with a given cloud platform, as they will likely provide the finest-grained experience for that system. This requires careful examination of available budget resources before making purchase decisions. In turn, weigh the number of supported third-party integrations an observability vendor offers, and map them to your existing systems. The more "out of the box" a cloud observability platform is, the more value it presents in flexibility and ROI. Consider also whether a given product is available as SaaS, on premises or both, as this may be critical to a successful multi-cloud observability strategy.
Finally, multi-cloud and distributed application development are rapidly evolving spaces; ensure that prospective vendors have a solid product roadmap that they can easily explain.
Chris Kanaracus is a former news director at TechTarget and former research director at IDC.
