Sunday, April 19, 2020

Modern Network Monitoring with Streaming Telemetry and Flow Recommendations

I am looking for recommendations for a modern, real time networking monitoring system and platform.

Today, most of our devices still only speak SNMP, which is by no means a modern nor real time monitoring protocol. Most NMS only support 5 minute polling, while some support 1 minute polling intervals. I am looking for something that can do at least per minute polling.

These devices also respond to ICMP ping requests. Most of these machines can respond to a ping per second, so graphing this data would be beneficial. Not sure per second pings would cause a spike in traffic or cpu resources.

Overtime, more and more of our networking equipment will support some sort of streaming telemetry.Today, we have both Juniper and Cisco devices that support streaming telemetry. Juniper supports streaming with their Junos Telemetry Interface (JTI). Cisco has their Model-Driven Telemetry. I cannot find many NMS that seem to support either of these protocols.

The last metric we would like to monitor is flows. Our Juniper devices support IPFIX and many of our switches support sFLOW. I am not sure that the default flow interval is on most of these type of devices, but I assume it's per second type of reporting?

I assume most will just recommend we use some sort of opensouce system, and make our own. The TICK stack, ELK, and prometheus come to mind. This seems to be the default option, especially for the streaming telemetry use case.

However, we really don’t have an interest in rolling our own. I like the look of Grafana and Kibanna for visualizations, but sorting through thousands of MIBs, OIDs, streaming telemetry events, and other data, and then creating our own dashboards sounds like a ton of work.

We are looking for something that we can buy or implement that is fairly straightforward, and ideally built on these modern time series database technologies. Something that already has all the MIBs and OIDs, dashboards built, and hard work done.

Solarwinds and PRTG seems like popular NMS providers. We reached out to both of them, and they both don’t even have streaming telemetry on their roadmap?

Some vendors that I think fit the bill might be OpenNMS, AKIPS, Statseeker, Kentik, and SevOne. I am awaiting pricing from many of these vendors, but let me know if there are any others we should add to the list.



No comments:

Post a Comment