We are excited to announce that Splunk has partnered with AWS in launching a new AWS Service Ready program – Lambda Ready. This designation recognizes that Splunk provides proven solutions for customers to build, manage and run serverless applications. AWS Lambda Ready designation establishes Splunk as an AWS Partner Network (APN) member that provides validated integrations and proven customer success with a specific focus on observability and monitoring of Lambda Functions.
Organizations are adopting Lambda Functions to rapidly deliver new features, focus on their customers and accelerate innovation. Serverless architectures provide the following advantages:
- Offloading operational concerns: Running serverless applications on AWS Lambda does not require managing a long-lived host or application instance; infrastructure is completely abstracted out, enabling teams to focus on their application code.
- Efficient scalability: The Lambda platform creates just enough runtime environment to handle any load. If one instance of the function is required, then Lambda will instantiate only one such environment. On the other hand, if millions of separate instances are required, Lambda will scale efficiently without any effort from DevOps teams.
- Cost: Lambda environments get closest to the true usage-based cost in the cloud as they are priced by the number of invocations, duration of function execution and memory required to run the function.
These benefits have propelled the use of serverless technologies among enterprises in every industry. According to Gartner, by 2021, 90% of enterprises using IaaS will also have some serverless applications in production. The latest CNCF survey cites that at least 41% of organizations are currently using serverless.
Serverless Poses New Challenges
While Lambda enables companies to increase the pace of software innovation and control cost, it also presents its own set of monitoring and observability challenges:
- The agent-based approach no longer works. Since serverless separates code from the infrastructure running it, agents cannot be installed to collect and transmit observability data.
- There is no place to batch the data, so traditional monitoring tools that use batch and query architecture don’t work in serverless environments as they are not designed for streaming data analytics. Using them in serverless will result in dropped data points and, consequently, incorrect analytics.
- Scraping metrics from CloudWatch results in slow insights at the typical five-minute resolution, or if called too frequently, may consume AWS API limits.
Lack of Real-time Visibility Leads to Cost Overruns and Sub-optimal User Experience
When you invoke a Lambda function, the request is processed in a new container environment. When a function has not been executed for some time or when you need to process more concurrent requests, AWS Lambda will start a new environment to execute your functions. The initialization of the runtime and your code leads to additional latency. This latency is usually referred to as a cold start. If you have several Lambda function invocations in a chain, latencies add up and result in poor user experience. In fact, a study conducted by Akamai shows that every 100-millisecond delay in website load time can hurt conversion rates by 7%. Since Lambda functions are billed for the duration in milliseconds, your AWS cost can easily balloon up if latencies are not managed. Real-time visibility into latency is required to control cost and deliver a flawless user experience.
Real-time Serverless Observability
Splunk solutions enable DevOps teams to easily understand the performance, usage and bottlenecks across the entire application spectrum. Whether your applications consist of 100% serverless or a mix of serverless and traditional apps, you can monitor your entire cloud stack with Splunk in real-time.
One Aggregate View For All Serverless Functions
A pre-built dashboard with key metrics across all serverless functions provides a bird’s eye view to understand how your serverless architecture is performing. Quickly search, filter, and sort to drill down into functions with errors, high latency, or cold starts. Tag and group functions based on usage, location, account, runtime, resources, or any other dimension.
Quick Drill-downs to Optimize Performance and Cost
To avoid cold starts, development teams can use the Provisioned Concurrency feature. Functions can instantaneously serve a burst of traffic with consistent start-up latency. However, you pay for the amount and the duration of provisioned concurrency. It is critical to understand how much of provisioned concurrency is used and the percentage of spillovers. Our pre-built dashboards give deeper insights into cold starts, concurrency, throttles and spillovers to manage under-provisioned functions, optimize latency and reduce cost.
The above chart shows the opportunities to increase the provisioned concurrency as spillovers are high, provisioned concurrency utilization is 100% and we are observing cold starts with high duration.
Service Visualization, Dependency Mapping and Troubleshooting
Visualize your services and their dependencies with automatic service maps created by SignalFx Microservices APM based on real-time data interactions between your services – serverless functions, traditional apps, third party APIs, database calls, etc.
Get granular details on every transaction with trace view. SignalFx Microservices APM captures every transaction across all services with NoSample full-fidelity trace ingestions to analyze every trace.
Use open-source tracing libraries to instrument code in most popular languages and leverage automatic instrumentation for common frameworks.
Wrappers for Real-time Insights on Lambda Metrics and Business KPIs
For situations when you need to track function invocations or errors at a level more granular than what CloudWatch can deliver, we provide a function wrapper that includes calls to SignalFx with the count of invocations and errors, the execution duration and cold starts. You get real-time insights with wrappers within seconds while CloudWatch Lambda metrics, with a standard resolution, are available with at best 1-minute, typically 5-minutes+ resolution.
Wrappers also provide an easy mechanism to instrument your code for custom metrics that matter, e.g. business KPIs – simply add a few additional lines within your function to capture and send those metrics to SignalFx without incurring any performance overhead.
Get Started with AWS Lambda Observability
We are excited to partner with AWS and looking forward to accelerating the journey to serverless for our joint customers. Future-proof your observability investment with an enterprise-grade solution trusted by enterprises for most advanced use cases at a massive scale. Serverless Monitoring is included with SignalFx Infrastructure Monitoring. Get started with a free trial today.