The last twenty years of networking have left much to be desired in terms of network visibility. Archaic least common denominator SNMP based approaches have been industry norm but only provided the bare minimum insight. Not much has changed since the 1990’s. Infrequent polling intervals completely miss networking events that take place on a sub second basis and inconsistent and inflexible MIBs gather only a limited amount of actual network state.
All of this has left the network operator essentially blind to what is actually going on with their network. In today’s cloud networks this legacy approach is clearly inadequate and a new approach to network telemetry is needed.
Arista’s telemetry solution has been improving visibility for today’s cloud networks with their telemetry tracers portfolio. Arista have expanded their telemetry solution with their new streaming telemetry platform within CloudVision.
What’s possible with a modern approach?
- The first place to start is the ability to do real-time monitoring. What is actually going on right now? Not just looking at averages or approximations but discrete state changes. Every interface counter. Every microburst and not just for one box but aggregated across the whole network and certainly not repeating the same command again and again to try to catch an event change. By looking at the real-time flow of state changes so that you can see an event right when it happens. A modern approach will provide a new level of granularity never before achievable. The ability to capture and view statistics from the network. Even if they take place in sub second intervals.
- The second place to improve on with a more modern approach to visibility would be the historical view. When a network event happens and the team is working on a root cause. They are essentially an investigative unit trying to piece together what happened. But they are limited to the info that was collected at that time. Certain MIBs. Certain log files. What would it be like if we had access to all the information found in the more than 2,000 commands in Arista EOS. Not just now but from that previous point in time. This is not possible with today’s tools so a new forensic troubleshooting approach is needed.
- The real-time examples also extend to other technologies and industries . Like the area of predictive security. Where we want to provide more complete data from the network to security tools to take action more quickly or in the area of broadcasting or IP storage where the need to monitor for a lossless transport requires visibility not typically found in MIBs. All of these areas require a new modern approach to visibility.
The key to Arista’s Telemetry solution is the state based software architecture of a Arista’s EOS. From the beginning Arista EOS devices store all real-time state in one common database SysDB. Easily accessible through powerful api’s and SDKs. With the software infrastructure enhancements to netDB we are now able to stream every state change from every device into an aggregated network-wide view. Since the collection mechanism is based on a streaming technology we’re talking about granularity down to the very instant the change is made. This makes the approach vastly superior to the traditional polling models that have a granularity on the order of every 10 to 15 minutes. Unlike traditional predefined SNMP MIB based approaches we are able to stream the full device state. This includes details such as; – Configurations, Counters, Errors, Statistics, Tables, Environmentals, Buffer utilization, Flow data and much more.
This streaming architecture is the basis for Arista’s yang base open config implementation as well. By collecting every state change in the network Arista customers will have access to both real-time and historic telemetry views of the network in one place and a level of granularity never before achievable. To leverage this rich network data the CloudVision platform is now enhanced to provide both the analytics engine and the telemetry visualisation for this network-wide state.
- The CloudVision platform includes a built-in state repository via a scalable storage infrastructure built on open-source HBase. Basically a big data environment for centrally collecting this stream network state.
- CloudVision now includes an analytics engine for processing the raw stream data into actionable information. More than just separating out the noise. This analytics engine digs through the data to track trends, correlate data across devices and layers and it even detect anomalies. Arista provided easy access to the stream state via open restful APIs. As a point for integration into a variety of partner solutions and customer specific infrastructure. Customers or partners can stream the state natively to their own back-end storage infrastructure or they can access the aggregated state via API to the cloud vision back end.
- Arista have added new telemetry Apps to the cloud vision portal. Including the workstream analytics viewer. Providing simplified visualization of network-wide state for faster time to resolution.