Connect Blogs

Is Packet Networking Ready for Pilot Channels?

We’re in another UTC conference season, and we will undoubtedly hear more of the CE vs IP/MPLS debate; what to use for a utility network buildout?  I think we have to pull back a few levels and remember what the core issue is: Can we use this network for critical system protection traffic?  In short, yes, and here are some of the key points on my mind:

  • Both CE and IP/MPLS is an Ethernet frame.
  • Link speed has the biggest effect on latency.
  • Carriers oversubscribe their network

Both CE and IP/MPLS traffic could traverse a Linksys device you’d buy at BestBuy for home use because both types of packets still have a MAC address header.  That is, they operate within the confines of an Ethernet Frame.  {Hear that?  Yes, you can use any Ethernet switch as a transport node; meaning there’s no add/drop traffic at that particular NE}.

The main thing driving FDV and latency is speed of the link.  Once you hit the 1GB mark on a non-blocking, carrier-grade device, we’re looking at per NE latency in the 10s of microseconds range.  CE and IP/MPLS work by adding tags to the Ethernet Frame; those tags might add up to an extra 50 bytes at worst case.  You can see, at 1GB+ data rates, it is still in the dirt.  At the transport NEs that don’t look into the packet, the latency increase is just the increased wire time.  My point is: it’s really negligible for most networks considering the equipment used to bookend the transport and provide interface services has latencies around the several millisecond level.  Sounds good so far.

So, what’s up with our legacy utility grade leased circuits performing poorly?  As the carriers upgrade to packet network, they’ve optimized for big data including voice, video, and internet.  Those services can have latencies in the 100s of millisecond range, and as to get the most value out of their network, they oversubscribe their bandwidth, so at times of network congestion, some traffic gets dropped.  What does this mean?

CE vs IP/MPLS discussions aside, if you architect your network such that it is not oversubscribed on any given physical link, packet technology is as adequate as legacy TDM.  In fact, it generally has lower per NE latency and lower cost.  All that stuff about dropped frames, excessive delay variation, it’s all caused by oversubscription.

Now, you’ll need some loop avoidance and tunneling protocols to implement redundancy which CE or MPLS bring to the table, but either one could work.  Next post I’ll talk about CE and IP/MPLS and where/why I’d apply the different technologies.

About the author