Continuous Delivery Accelerates Developer Velocity & Cloud Migration
Over the past three years, Harness has helped customers like NCR, OpenBank, and McAfee master Continuous Delivery (CD) as they migrated their monolithic applications to microservices running in the public cloud.
Deployments in the past were managed by IT Operations or dedicated release teams. Today, that responsibility has ‘shifted left’ to engineering teams who use a self-service CD platform like Harness, where engineers can deploy on-demand to the public cloud.
As developer velocity goes up and time-to-market comes down the next big challenge customers face is cloud spend. Corey Quinn, the Cloud Economist at Duckbill Group summed it up nicely:
At last year’s AWS re:Invent, Abiadde Adedoyin from the AWS Cloud Economics team noted in her cloud financials session that “35% of cloud spend is waste”.
Now think about your last cloud bill, divide it by 3, and that is what AWS considers waste. A pretty sobering thought.
Cloud Consumers Lack Basic Visibility of Cost
As cloud consumers, engineering teams need to understand their impact on cloud spend to request and manage cloud resources efficiently.
Today, the majority of these teams are 100% reactive to the teams that do have access to cloud cost data– the teams who pay the bills (Finance/FinOps) or the teams who manage, secure, and govern the cloud infrastructure (Cloud Center of Excellence, Cloud Engineering).
Simply put, engineering and DevOps teams don’t have self-service access to cloud cost data. Here are some examples of how self-service works in other areas for engineering teams:
- APM (Performance) – self-service access to AppDynamics, Datadog & New Relic
- Log Analytics (Quality) – self-service access to Splunk, Elastic & Sumo Logic
- Incident Management (Availability) – self-service access to PagerDuty & EverBridge
- Cloud Spend (Cost) – no self-service access to tools and have to manually tag infrastructure
Just like engineers optimize performance or quality, with cost visibility, they are more likely to optimize cloud costs without their CFO having a coronary every 30 days.
Introducing Harness Continuous Efficiency
We built Continuous Efficiency to empower engineering and DevOps teams so they can proactively manage and optimize cloud costs, allowing finance teams to “shift down” the responsibility of cloud costs to the people who consume cloud resources.
In a nutshell, this is why Harness Continuous Efficiency can enable organizations to “shift down”:
Complete Developer Context & Visibility
This means engineers get self-service visibility to the real costs of their applications, microservices, clusters, and environments. They’re not constrained by the limited infrastructure or cloud service views of the world and get the context they need when managing cloud costs.
Engineers also get this visibility without manually tagging cloud infrastructure and creating unnecessary toil. For example, here is a screenshot showing cloud costs by microservice:
Root Cost Analysis for Kubernetes & AWS ECS Clusters
Engineers can also deep dive into their container orchestration, viewing the actual utilized, idle, and unallocated costs of clusters, namespaces, workloads, nodes, and pods. Again, no tagging is required. Think of this as an x-ray of where cloud resource is and isn’t being consumed, so engineers can resize and optimize their clusters proactively to reduce cloud spend.
This view is extremely useful for engineers who want to see how much of their ‘requested resource’ is actually utilized, idle and unallocated. Just because an engineer requests 16 CPUs and 256GB RAM in their manifest doesn’t mean the application or service is actually using it. You can see in the above screenshot that only 33% of the cluster resources are utilized.
Engineers can also identify zombie namespaces or workloads on clusters that may not be required making cleanup of cost an easy exercise.
Hourly Time Granularity
The norm for cloud providers and legacy tools is to report cloud costs thru daily or monthly time windows. Harness Continuous Efficiency takes this one step further with hourly time windows so engineers can truly understand cloud resource usage and cost trends over-time. Aggregating cost into daily or monthly windows means abnormal spikes may go unnoticed and are blended by the averages.
Correlate Cloud Events with Cloud Costs
Engineers also need situational awareness of cloud events that may contribute to cost spikes. New application or service deployments, modified resource manifests (e.g. replicas changes, compute spec changes), and activation of Horizontal Pod Autoscaling (HPA) or Node Autoscaling are all events that impact cloud consumption and costs.
For example, in the screenshot below, you can see that all related events are overlaid and correlated with cloud spend and spikes for clusters. Events can be clicked and detailed in the pane below, providing detailed insights into what changed in the environment (in this example, someone changed the Kubernetes Replicas count from 2 to 3).
Continuous Delivery (1) + Continuous Efficiency (1) = 3
By integrating Continuous Efficiency into our CD platform, engineers can now access cloud cost visibility for their applications, microservices, and clusters in seconds. Engineers don’t need another tool or log in; they simply deploy and get hourly feedback on how their deployments or changes are consuming cloud resources and cost. Instead of waiting weeks or a month for another team to highlight a cost spike, they can get cost feedback the same day and act on it, thus helping organizations eliminate cloud waste and reduce their cloud spend.
In addition to the above unique capabilities, Continuous Efficiency also provides standard features you expect in any cloud cost solution:
- Multi-Cloud Support for AWS, GCP, and Azure
- Visibility of Non-Clusters Costs (e.g. EC2, S3, RDS, ELB, CloudWatch, …)
- Executive Dashboard
- Budgets & Alerting
- Historical Reporting, Digests & Data Export
- Slack & Email Integration
Harness Continuous Efficiency installs in seconds so it’s an easy lift to get started. Whether you’re an existing Harness CD customer or not, your teams can be up and running the same day getting unique visibility into their cloud costs.
Eating Our Dog Food and Drinking Our Champagne
Harness recently ‘shifted down’ when our run-rate for the public cloud was $1.2m per year. A few months later, we empowered our engineering team with Continuous Efficiency. The result? We reduced our public cloud annual spending by 40% to $700k, which proved that the 35% stat from AWS was real and, if anything, a little conservative.
Continuous Efficiency isn’t a one time fix
The adoption of Continuous Delivery means that engineering teams are continuously deploying to the public cloud. Every single change has the potential to alter the way an application or service consumes cloud resources, and ultimately cloud costs. Every new application or service which is onboarded to the public cloud also has the potential to be over-provisioned by its owner.
“Shifting Down” the responsibility of Cloud Cost to engineers helps organizations achieve Continuous Efficiency, and in doing that can help eliminate the 35% of cloud spend waste that organizations are incurring every year.
Click here for more information or to request your Continuous Efficiency demo.