Since the day enterprises first connected PCs to a LAN and built client-server systems, network engineers became the scapegoat since when users and developers cast blame for any application problem, it's always the network. The reality of connected devices is that the network is guilty until proven innocent and the burden of proof is on network administrators to absolve their systems of responsibility. Sadly for beleaguered network managers, today's era of cloud infrastructure, SaaS applications, WFH employees, edge computing and composite applications relying on a multitude of external components, makes the burden almost insurmountable.
With the world of application development, cloud operations and network administration converging, formerly specialized and siloed tools for measuring performance and identifying bottlenecks have become interdependent, with each realm dependent on data from the others. Finding the source of performance problems amidst the maze of application connections and dependencies is virtually impossible without software assistance. Cisco describes the situation well in its 2020 annual report,
In our view, over the next several years, customers will be increasingly writing modern software applications that can run on any hybrid cloud, and will be adding billions of connections to their environment. They will need to be able to build applications quickly, deploy them nearly anywhere, monitor experiences, and act in real time.
Such interdependence and the synergies, correlations and the insights made possible by combining software IP and development expertise was the motivation for Cisco to acquire two leaders in performance measurement over the past few years: AppDynamics in 2017 and ThousandEyes last year. Perceptive readers of my columns are familiar with ThousandEyes as the source of one of the most detailed and valuable analyses of cloud network performance (see here from 2018, this 2019 update and a post-lockdown analysis here). Indeed, its study of performance differences revealed some surprising variability in network routing choices among the major cloud providers.
Long before ThousandEyes carved a niche in cloud network performance monitoring, AppDynamics became a leader in application performance management (APM), by using machine learning (ML, in its Cognition Engine) to baseline performance levels, identify deviations and correlate anomalies across different components to find the root cause of performance bottlenecks, minimize the time required for problem resolution and proactively flag aberrations before they become significant. The company focuses on so-called "full-stack observability," namely the ability to monitor a user's view of application performance, including all the web, IaaS, SaaS and on-premises components that comprise a modern composite application.
The availability of cloud services to provide data and functions previously operated in-house has turned applications into a mash-up of custom code, external API calls and embedded cloud services, making products like AppDynamics a requirement for fully understanding application performance and finding bottlenecks. Although creeping complexity has been endemic to IT for generations, three-quarters of the respondents to an AppDynamic-sponsored survey said that the situation has gotten worse over the past year with the accelerating shift of workloads and infrastructure shifted to the cloud.
With cloud services entwining the realms of network engineering and APM, it underscores the need to tightly link the monitoring and optimization tools used in each area. Thus, the Cisco Live announcement that it is integrating the AppDynamics and ThousandEyes products is well-timed.
Better together, better still as SaaS
As my previous columns detailed, ThousandEyes probes the network plumbing from a user's perspective, from a client device, through the ISP, various intermediate Internet infrastructure (like DNS, CDN, network exchange points) to an IaaS availability zone or SaaS instance. As the accompanying Cisco Live slide illustrates, AppDynamics picks up performance measurement from an application or SaaS API, through the executable code to the runtime environment, whether SaaS, IaaS or on-premises. Together, they provide the full-stack holy grail of end-to-end performance analysis for cloud-native applications.
Extending AppDynamics' ML capabilities to network monitoring and equipment management is fundamental to Cisco's vision of delivering the central nervous system for IT, one that combines visibility (network and application monitoring and benchmarking), insights(ML-powered analysis and predictions) and action (automation and guidance for infrastructure and application management products).
Attuned with the zeitgeist of the modern enterprise, both AppDynamics and ThousandEyes are available as SaaS. indeed, as Cisco CEO Chuck Robbins noted in discussing the company's Q2 2021 earnings, cloud delivery is the future for all Cisco services (emphasis added),
"We plan to transition the majority of our portfolio to be cloud-driven, cloud-managed and delivered as a service. If we can deliver from the cloud, we will. For example, we are looking at offering SD-WAN plus cloud security as a service, along with creating other new solutions. We will also provide simplified end-to-end networking with security, reliability, control and automation plus seamless on-ramp capabilities to the cloud that no one else can deliver."
Cisco's Q2 earnings show significant progress on this front, with 76 percent of software revenue coming from subscriptions.
Cisco's first steps towards full-stack performance management
The first tangible fruits of Cisco's performance management acquisitions, announced at its Cisco Live event, are the integration of ThousandEyes with AppDynamics Dash Studio and its Catalyst 9300 and 9400 switches.
- Dash Studio is the company's data visualization and reporting platform that uses a visual dashboard editor to create layouts, embed pre-built widgets and define bindings and queries to data sources. Integrating ThousandEyes data provides details about the performance of connections to various cloud resources previously obscure to application developers.
- The 9300 and 9400 are in the middle of the Catalyst switch lineup with the stackable 9300 designed for campus and large ranch deployments, while the modular, chassis-based 9400 targets either access and distribution layer deployments within large enterprise LANs. ThousandEyes uses enterprise agents on switches, OSs, VMs and containers to measure performance such as loss, latency, jitter, and path topology via a combination of active and passive monitoring techniques. Agents use TCP and ICMP tests to service endpoints or another agent to measure path performance and map the network topology. Building the agent into the switch OS eliminates the friction of installing after-market software and runs the agent in a more efficient containerized environment.
Incorporating ThousandEyes data in Dash Studio improves the manageability of composite application by letting developers and operations teams:
- Correlate application and business process performance with network and internet measurements.
- Identify whether a problem is caused by an external dependency, Internet bottleneck or internal code problem.
- More thoroughly troubleshoot problems by using network diagnostic information in ThousandEyes.
APM and network monitoring software are like chips and dip, better together. Cisco's marriage of AppDynamics and ThousandEyes is sure to be fruitful for frustrated IT administrators and developers tasked with maintaining the performance and reliability of modern applications reliant on multiple cloud services. At 11% of revenue, applications are only a small fraction of Cisco's business, however, they present the best opportunities for growth and recurring revenue as its enterprise customers shift more of their infrastructure to the cloud. Consequently, APM and optimizing the end-user application experience is one of the six strategic pillars that Robbins highlights on most of Cisco's earnings calls.
As Todd Nightingale, Cisco's SVP of Enterprise and Cloud Networking noted at a recent Morgan Stanley telecom conference, companies are awash in so much telemetry that the challenge is figuring out what it means and how best to act on the data, noting that "analytics plus algorithms equals insight. And then you have to be able to take action." With merchant silicon and open network operating systems decreasing the differentiation between network hardware and enterprises reducing their data center footprint, Cisco's best shot at distinguishing itself from competitors comes from software. Whether it's full-stack APM, SD-WAN and SASE platforms or security products, software is eating Cisco's world. Linking its AppDynamics and ThousandEyes products is an excellent example of the synergies an integrated infrastructure provider like Cisco can bring to the complexities of today's IT.