By Nancy Wang and Pranava Adduri, Advancing Women in Product

2020, The Hybrid Wars, Observability, & Putting the Ops in DevOps

Disclaimer: Thoughts below are our own and do not reflect that of any past or present employers. 

Moving into 2020, we foresee several emerging infrastructure trends and needs that will come into primetime and shape the future of cloud computing and devops as we know it. 

#1      Let the Hybrid Wars Begin

Gone are the days when cloud providers could ignore customers demanding hybrid or purely on-premises computing environments. As customers encounter more challenges in their on-premises to cloud migrations, the willingness to ‘lift and shift' all applications to the cloud diminishes, cloud providers are incentivized to enable their offerings to be on-premises and edge-compatible. Case in point, AWS Storage Gateway, Snowball Edge, and Outposts. Prior to this year (2019), Azure was unique in that it had a unified offering across cloud and on prem with Azure stack. The introduction of GCP Anthos and Amazon Outposts this year marks the beginning of cloud providers focusing their efforts on building out additional functionality that expands their current reach to on-premises workloads. This marks the end of the phrase, "everything must live in the cloud". 

While all three major cloud providers (Amazon, Google, and Microsoft) have their sights set on extending their oversight to on-premises, it should be noted that how they go about doing this provides insights into how they view on-premises workloads. For example, Google Anthos, by focusing its efforts on easing the process of running on-premises container workloads demonstrates Google's preference of making the container the standard unit of compute. In contrast, Amazon's Outposts hints at AWS desire to extend its existing platform services to on-premises in a seamless manner. And finally, Azure Stack, most recently rechristened as Azure Hub, has, for a while, maintained its stance as something developers could deploy on any hardware - effectively bringing Azure to their own datacenters. However, part of the strategy re-tooling from Microsoft has been to certify Azure Stack on only a handful of hardware players (e.g., Dell, HP, Cisco) coupled with Azure Data Box (the Azure flavor of AWS Snowball) to migrate on-premises data to Azure. The overarching theme is that cloud players are doing hybrid, but only on their terms. 

Whether it's betting on containers (Google), creating a seamless platform stretching across on-premises and the cloud (Amazon), or establishing its ‘stack' as dominant across on-premises and the cloud (Microsoft), it will be interesting to see how companies that have built their businesses around standardizing on-premises and hybrid compute will continue to differentiate themselves (i.e., Nutanix).

#2      Observability: Everything the Light Touches 

With the number of plays on observability skyrocketing, there are fewer and fewer shadowy places we can't measure. With vendors leveraging technologies such as BPF, observability is possible even without instrumenting agents installed in workload runtimes. With tools like Istio, configuring complex network policies in a declarative fashion has become possible. 

With observability, we also see the evolution of the description of infrastructure in 2020. Currently, we use restricted syntax for ease of readability (e.g., YAML, JSON). K8s and AWS CloudFormation templates all make heavy use of YAML files. However, as savviness grows among the user base and the need for more expressive syntax and flexibility between objects, these constructs that were built for simplicity will reach boundary conditions - and they will no longer satisfy what the user wants. Users will end up scripting generations of declarative files. A company that has emerged to meet this need is Pulumi, which officially went out of Beta this year with v1.0. I'm excited about the concept of "infrastructure as code", where your infrastructure is your code base. Using Pulumi, you can schedule dry runs to ensure that changes to your infrastructure work as intended before its deployed. As we move to a world that is increasingly bifurcated and there is no clear dominant infrastructure provider platform, the need for standardization across development and production platforms becomes increasingly more urgent. 

Effective monitoring must keep pace so that operators can analyze and react to events. We emphasize effectiveness in monitoring as not all monitoring is created equal. As deployment and infrastructure complexities increase, the types of events and outages that occur are increasingly more emergent and basic monitoring/alarming on observed metrics no longer provide sufficient context to respond to situations at hand.

As organizations may deploy several observability and monitoring solutions, we also foresee the need to be able to combine and reduce monitoring data across solutions to give operators actionable insights. As the observability train kicks into full gear, in 2020, we'd like to see novel monitoring and analysis solutions arise to leverage the influx of observable metrics.

#3      Putting the Ops in DevOps

While there are many ways to categorize the tools that DevOps spans, we can broadly bucket the tools as helping with collaboration, development, testing, deploying, and running workloads. When it comes to running workloads in production, we find that current tooling focuses on observability and monitoring to detect problems that might arise. However, these tools fall short when it comes to being able to assist the operator in remediating an issue that has been root caused. Instead, operators are still relegated to repeating manual resolutions again and again.

What we need to see in 2020 is the transition of operations workloads to developer-centric tooling. As the emphasis on observability grows, the number of operations personnel does not scale. For example, as the number of workloads grow exponentially, the number of operations personnel does not. Given the exponential rate of growth of deployment complexity compared to operating personnel, the emphasis on observability makes me happy. 

Observability is the only way we will be able to deal with entrance of much more complex deployments, which will have many emergent phenomena in terms of issues. Without emphasis on observability and further tooling that enables teams to understand what is going on - issues can't be tamed.

A business' uptime can be its competitive advantage. In a world where violations in SLAs can be very costly, reducing an organization's mean time to resolution can have a noticeable dollar value. We see an opportunity for tooling to step beyond observability and monitoring to actively assist operators in resolving issues faster. Organizations are only as good as the problems they can effectively diagnose.

##

About the Authors

Nancy is the Founder & CEO of Advancing Women in Product, an international NGO that aims to bridge the gap for women in tech leadership roles, with chapters in SF, Seattle, Boston, and Paris. She also currently leads the data protection product line at Amazon Web Services. Previously, Nancy also led products at Rubrik, the fastest-growing enterprise software unicorn. With a history of building and launching large-scale enterprise systems in storage, data management, and networking, Nancy also led product at Google (Fiber) and system integration efforts for the U.S. Intelligence Community in Washington, D.C. She is passionate about the democratization of data and empowering more women tech leaders as the founder & CEO of Advancing Women in Product. Nancy graduated from the University of Pennsylvania with a degree in engineering.