Saturday, March 12, 2016


Figure 1: Visibility and the software defined data center
In the land of microservices, the network is the king(maker) by Sudip Chakrabarti, Lightspeed Venture Partners, makes the case that visibility into network traffic is the key to monitoring, managing and securing applications that are composed of large numbers of communicating services running in virtual machines or containers.
While I genuinely believe that the network will play an immensely strategic role in the microservices world, inspecting and storing billions of API calls on a daily basis will require significant computing and storage resources. In addition, deep packet inspection could be challenging at line rates; so, sampling, at the expense of full visibility, might be an alternative. Finally, network traffic analysis must be combined with service-level telemetry data (that we already collect today) in order to get a comprehensive and in-depth picture of the distributed application.
Sampling isn't just an alternative, sampling is the key to making large scale microservice visibility a reality. Shrink ray describes how sampling acts as a scaling function, reducing the task of monitoring large scale microservice infrastructure from an intractable measurement and big data problem to a lightweight real-time data center wide visibility solution for monitoring, managing, optimizing and securing the infrastructure.
Figure 2: sFlow Host Structures
Industry standard sFlow is the multi-vendor method for distributed sampling of network traffic. The sFlow standard is model based - models of entities such as interfaces, switches, routers, forwarding state, hosts, virtual machines, messages, etc. are used to define standard measurements that describe their operation. Standardized measurements embedded within the infrastructure ensure consistent reporting that is independent of the specific vendors and application stacks deployed in the data center. Push vs Pull describes how sFlow's push based streaming telemetry addresses the challenge of monitoring large scale cloud environments where services and hosts are constantly being added, removed, started and stopped. In addition, sFlow Host Structures describes how the data model allows telemetry streams from independent sources in network, server and application entities to be combined at the sFlow receiver to provide end to end visibility into the microservice interactions and the compute and networking services on which they depend.

The challenge in delivering network visibility to microservice management tools is not technical - the solution is fully deployable today:
  • Applications - e.g. Apache, NGINX, Tomcat, HAproxy, ADC (F5, A10, ..), Memcache, ...
  • Virtual Servers - e.g. Xen, Hyper-V, KVM, Docker, JVM, ...
  • Virtual Network - e.g. Open vSwitch, Linux Bridge, macvlan, ...
  • Servers - e.g. Linux, Windows, FreeBSD, Solaris, AIX
  • Network - e.g. Cisco Nexus 9k/3k, Arista, Juniper QFX/EX, Dell, HPE, Brocade, Cumulus, Big Switch, Pica8, Quanta, ... – visit for a complete list
Network, system and application teams working together can enable sFlow instrumentation that is already embedded throughout the infrastructure to achieve comprehensive visibility into microservice interactions.
Incorporating sFlow analytics into the microservices architecture is straightforward. The sFlow-RT analytics engine processes the raw telemetry streams, combines data using the data model, and delivers visibility as a REST based microservice that is easily consumed by new and existing cloud based or locally hosted orchestration, operations, and security tools.