Distributed Network Monitoring

Networks Are Distributed

Enterprise networks are distributed systems by nature which are supposed to operate autonomously after correct deployment and configuration. I want to emphasize the “after correct configuration and deployment” part because if this happens the network engineer’s responsibility is fulfilled and he can go to bed, worry free, after a 3:00 am network change. We all know that hardly ever this is the case because he won’t be 100% sure that everything went fine after his change until next morning that users come in the office. This is only one example of the hurdles introduced by the distributed nature of networks.

All of us know that bringing a network to the “correct configuration and deployment” goes through unmapped hills and valleys. Yet, the important thing is to be prepared with the right tools to tackle every bump and pothole in your path as soon as you face them.

Distributed and Traditional Network Monitoring

Traditional network monitoring, also referred to as up/down or SNMP monitoring, is indispensable to gather information about the status and performance of network devices (e.g. switches and routers). Traditional network monitoring is something that any network engineer is familiar with. So how it is different from distributed network monitoring?

Let’s review some characteristics of distributed network monitoring and how it can be used to complement and close the gap in the network monitoring arena. The goal of distributed network monitoring is to complement the information that is missing from up/down monitoring, and inform the network engineer  how the end-user really experiences the network and the applications. At the end of the day, enterprise monitoring is not just about the network, but about the users.

Differently from traditional network monitoring, which is performed by a central server that regularly polls network devices, distributed network monitoring is performed from remote locations. Inevitably, this kind of monitoring has to be agent based. After all, somebody at the other end of the wire has to send the information that is needed.

I am sure the first question that popups into your mind is: Software or hardware agents? The eternal conundrum. Everyone understands the most obvious pros and cons of this question. I will give you my take at another blog post about hardware vs. software agents.

Evaluating Distributed Network Monitoring Tools

So, what properties should a distributed network monitoring tool have? Here is my list in order of importance:

1. End-user monitoring: this is the main goal of distributed network monitoring, so it couldn’t be anywhere else apart from the top of the list. What this “end-user monitoring” includes is a matter of what are the needs of each network engineer. It could be bandwidth, streaming or VoIP quality, routing between locations, application performance, you name it!
2. Manageability: when the number of agents increases, the manageability effort should increase slowly. Medium and large enterprises might have hundreds of locations that would require monitoring by an agent. The responsibility for this property falls in the hands of the UI/UX designer.
3. Scalability: the backend architecture and GUI of the tool should be able to support hundreds if not thousands of agents. Different industries have different needs, but if we take as an example a bank that wants to monitor all its branches we can easily think of examples that would need thousands of agents.
4. Robustness: what happens when an agent loses connectivity to the GUI dashboard? Does the agent cache the monitoring results locally and resends then upon reconnection, or does the agent have a second or third redundant path to the dashboard? If the agent fails (or the server shuts down in the case of a software agent) is there a backup or not? Most of these questions can be answered when somebody imposes requirements and cost. Military applications have different requirements that civilian ones.
5. Pricing: this is last not because it is not important, but because the pricing comes with value. If the tool costs less than the value it adds, then it is worth the investment. Nonetheless, the cost per agent is an important aspect in order for a tool to be affordable when scaling to hundreds and thousands of agents.

Feel free to comment on with what you think would be the most important properties and benefits of your ideal distributed network monitoring tool.

If you want to learn more about NetBeez, request a product tour!