July 2018 Summaries

The first Dash conference, organized by Datadog, took place recently in New York City. It featured keynotes from industry leaders, product launches, and breakout sessions on topics such as performance, scalability, and teams. Highlights included a discussion of how Datadog was founded to bridge the gap between development and operations teams, new features announced by the company's product teams, insights into observability as a business concern, and lessons learned from NASA's approach to reviewing all operations. The event concluded with an announcement that plans are underway for an even bigger Dash conference in 2019.

The first-ever Dash conference was held in NYC, featuring keynotes from Datadog CEO Olivier Pomel and former NASA flight director Paul Hill. Key topics included product launches such as Watchdog, App Analytics, and Logging without Limits, which aim to improve observability and reduce costs. Another focus point was the importance of approaching observability as a business concern, ensuring good service for customers by monitoring systems from end to end. Breakout sessions covered performance, scalability, and teams, with speakers sharing their experiences and lessons learned in building and managing complex systems, including using circuit breakers and experimentation-built-in CD. The conference aimed to share knowledge and get to know the community better, with plans already underway for a bigger and better Dash in 2019.

Datadog has introduced Live Process monitoring for Windows, allowing users to gain granular insights into resource usage across their infrastructure including Windows Server hosts from version 2008 and up. The Live Process view provides real-time data on CPU and memory usage of each process, with the ability to search and filter by various parameters. Historical process metrics can also be viewed for additional context. Users can add Windows process monitoring to existing dashboards and easily pivot between dashboard graphs and Live Process views. To enable Live Process monitoring for Windows, users need to update their Datadog Agent to version 6 and configure the settings on each Windows host.

Datadog has announced the availability of Live Process monitoring for Windows, which provides real-time insights into resource usage across infrastructure. The feature allows users to explore a detailed inventory of processes running on Windows Server hosts, version 2008 and up. With Live Process monitoring, users can gain granular visibility into process resources, including CPU and memory usage, as well as view historical data and add it to their dashboards for further analysis. To use the feature, users need to upgrade to version 6 of the Datadog Agent and follow specific configuration instructions. The new feature is now available in a single view that combines Windows and Linux processes, allowing users to search, filter, and group process metrics by various criteria.

Datadog APM has released support for monitoring Node.js applications, joining its existing support for Java, Ruby, Python, and Go. This feature provides detailed performance overviews of applications, tracing requests across distributed services and hosts to identify bottlenecks and debug errors. Datadog APM automatically instruments requests to commonly used modules in the Node.js ecosystem, including web frameworks like Express and data stores like MongoDB, Redis, MySQL, and PostgreSQL. The tool also allows users to set up automated alerts for potential issues such as high error rates or anomalous request throughput. Additionally, Watchdog uses machine learning to surface potential problems without requiring manual configuration of alerts. Datadog APM is compatible with OpenTracing, the open standard for distributed tracing, making it vendor-neutral and easy to port applications from one backend to another.

Datadog APM has officially released support for monitoring Node.js applications, providing detailed performance overviews and traces to help troubleshoot issues effectively. The tool automatically instruments requests to commonly used modules in the Node.js ecosystem, including web frameworks and data stores. With Datadog APM, developers can quickly identify bottlenecks and debug errors by viewing latency distribution and percentile statistics for each service running in their application. Additionally, the platform offers automated alerts and Watchdog to surface potential performance issues without requiring manual configuration. Developers can easily set up the tool by installing the Datadog Agent and `dd-trace` library on their Node.js servers, or by deploying APM on Docker using a provided image.

Docker is a rapidly maturing container technology that provides easily-configured, lightweight virtual machines ideal for microservice architectures and scaling environments. Amazon Web Services' EC2 Container Service (ECS) automatically manages Docker containers, offering load balancing, health recovery, and scaling automation. However, monitoring containers can be challenging due to their short lifespan and frequent changes. Datadog is designed to monitor highly dynamic infrastructure, including containers, and has partnered with AWS engineers to create a tailored ECS integration. This allows for automatic tracking of container metrics, flexible aggregations, and alerts for service-level problems. With Datadog, users can track both Docker metrics and the software running inside each container, as well as easily correlate metrics from any part of their infrastructure.

App Analytics is a new feature in Datadog that provides distributed tracing for application performance monitoring (APM). It allows users to explore and analyze all their spans in one place, making it easier to find specific traces for troubleshooting or debugging. The tool supports filtering and aggregating data using high-cardinality dimensions like customer ID, user ID, or checkout value. App Analytics enables users to drill down and analyze every span quickly, even with thousands of services or millions of users. It also allows for slicing performance aggregates on the fly and provides a brand-new analytics interface for visualizing data using high-cardinality attributes. Users can add these views to their dashboards for continuous monitoring and pivot between related data sources using common tags, unifying the three pillars of observability: metrics, traces, and logs.

Watchdog is an auto-detection engine by Datadog that surfaces performance problems in applications without any manual setup or configuration. It uses machine learning algorithms to automatically detect issues such as latency spikes, elevated error rates, and network issues across microservices, endpoints, and cloud zones. The tool creates a feed of "stories" with notable findings, each highlighting the affected resource, timeframe, and impact duration. Each story links to a detail page that provides further context and aggregates performance statistics for the specific service or resource. Watchdog also identifies likely culprits by surfacing related behavior and common stack traces. It detects network issues in cloud infrastructure and helps identify the underlying cause of anomalies. The tool builds on Datadog's existing machine learning features to provide high-quality results tailored for large-scale infrastructure and applications, with plans to add more algorithms for root cause analysis and Kubernetes anomaly detection.

Watchdog is an auto-detection engine from Datadog that surfaces performance problems in applications without manual setup or configuration. Using machine learning algorithms, it detects issues such as latency spikes, elevated error rates, and network issues across various resources in the environment. Watchdog creates a feed of "stories" with notable findings, each highlighting the timeframe of interest on a timeseries graph and providing a plain-language summary of what happened. The engine automatically surfaces related behavior that might have the same underlying cause, allowing users to identify likely culprits. Additionally, it detects network issues in cloud infrastructure, enabling failover and re-routing of traffic to unaffected zones or another cloud provider. Watchdog builds on Datadog's established machine learning features and continually adds algorithms to detect new kinds of situations anywhere in the environment.

App Analytics is a new tool that provides a detailed view into application performance by capturing the level of detail across hundreds or thousands of services. It enables users to explore and analyze all their spans in one place, filter down to find specific traces using high-cardinality dimensions such as customer ID or user ID, and visualize data using high-cardinality attributes. App Analytics unifies the three pillars of observability—metrics, traces, and logs—with tags at its center, allowing users to pivot between related data sources using common tags. It provides real-time analytics views for on-the-fly troubleshooting and spot-checking, as well as the ability to add these views to dashboards for continuous monitoring. By integrating with Datadog APM, App Analytics gives users unparalleled visibility into their applications and how their customers interact with them.

Datadog's log management solution addresses the limitations of traditional logging systems by decoupling log ingestion from indexing, allowing users to collect, process, and archive all logs cost-effectively. This approach, known as Logging without Limits™, enables dynamic decisions about which logs to index for troubleshooting and analytics, without the need for server-side filtering that might miss valuable data. Users can archive logs in long-term cloud storage for auditing purposes without additional costs and utilize features like Live Tail for real-time log observation. Datadog allows users to adjust indexing policies instantly to maximize visibility during incidents and control data storage costs, offering an integrated platform for monitoring logs, metrics, and distributed traces.

Datadog has released version 1.0.0 of its Go tracer, which includes performance improvements, better compatibility with tracing standards, and a new API. The update also addresses long-term open issues and improves support for the community. Key features include distributed tracing, support for vendor-neutral tracing standards like OpenTracing and OpenCensus, and significant increases in performance when handling concurrent, high-traffic environments. A migration guide is available to help users upgrade from previous versions.

The host map in Datadog provides a visual representation of an organization's infrastructure, allowing users to quickly view information about the metrics reported by their hosts. Host maps represent each host as a hexagon, with color and size indicating different performance metrics. Users can change the metric represented by color or size, select from various color palettes, and filter or group hosts based on shared attributes such as tags. The host map also allows users to zoom in for detailed information about individual hosts. Saved host maps can be reused and continually updated with current data.

The Datadog team has released version 1.0.0 of their Go tracer, which represents a major overhaul with performance improvements, enhanced compatibility with tracing standards, and a new API. The release was tested in production for five months and resulted in significant benefits, including reduced memory and CPU usage by up to 90%. The goal of the release was to provide a stable, flexible, and simplified API that adheres to semantic versioning standards, while also improving support for vendor-neutral tracing standards like OpenTracing and OpenCensus. The tracer now includes features such as distributed tracing, distributed sampling priority, and improved performance in high-traffic environments. To make the upgrade process easier, the team has launched a migration guide and enabled parallel usage of both tracers during the migration period.

The text outlines the integration of Datadog with a Flask application to monitor performance and infrastructure. Flask, a Python microframework, is noted for its simplicity and extensibility, allowing developers to choose their preferred database and HTTP server configurations, such as uWSGI and MySQL. The document guides the reader through installing Datadog's Agent and Python tracing library to collect metrics, logs, and traces, providing insights into the application's performance. It emphasizes the advantage of using NGINX as a dedicated HTTP server for improved performance and suggests the use of Datadog's features for comprehensive monitoring, troubleshooting, and optimization of the app's stack, with a free trial available for new users.