Pros and Cons of Metcalfe's Law for IT Ops | ExtraHop

Metcalfe's Law is known in popular consciousness as the "network effect" and often applied to social networks online. But the concept has been around long before the World Wide Web was born.

Back in 1980, Robert Metcalfe foresaw the value of communications networks increasing exponentially as more devices connected. The formula (n²) is used to describe how five connected devices can make 10 connections, but twelve devices can make 66 connections, for example.

At first glance, IT leaders might rejoice because it means their environments are becoming that much more valuable as new Internet-connected "things" come online or as microservices are spun up. But from an IT Operations viewpoint, this represents a vast increase in complexity.

In fact, from a complexity standpoint, it's even worse because you not only have to account for the connections to a host, but also the processes (Active Directory, CIFS, SMTP, etc.) running on the host. So it's not just n-number of hosts, but also n-number of processes running on those hosts!

Seen through the lens of Metcalfe's Law, IT complexity presents a serious problem now and in the future. This is why IT teams need to become data-driven and cannot afford to maintain the status quo. It's also why IT teams cannot afford to rely on siloed tools that are only used by technical experts, but need to adopt analytics platforms that can be accessed by multiple teams.

I've offered a way forward: Namely, that IT leaders focus on lowering the friction to data as a way of combating the harmful effects of IT complexity.

This is a companion discussion topic for the original entry at
1 Like

Thanks for the reminder - - by the way - this is normally seen as n*(n-1)/2 that approaches n squared as is mentioned in the link you provided.

Also as you bring out - it not just the Nodes that interact / servers - but also the processes involved. This is what makes this such a head ache.

This is why something like ExtraHop is so valuable to have in place. For once, we can see the different interactions taking place between systems. What we need to be careful of is to factor out the common services like LDAP and DHCP - - What are the real business interactions taking place.

In the mapping of systems- it would be helpful to have a fuzz value that below this % threshold we do not care to see the interactions - or to filter out the ‘service’ interactions like LDAP or Solarwinds or other noise that we do not consider part of the real application.

If LDAP is down - then no one can use any application because they can not check their permissions for instance. We do not need to know that this is critical to each application. It is a given already.

Our silo-ed brothers/sisters need to come on board and see the wealth of data we are able to see of the interactions taking place.


1 Like