Pushing Packets the Last Mile

“The “last mile” or “last kilometer” is the final leg of delivering connectivity from a communications provider to a customer. The phrase is therefore often used by the telecommunications and cable television industries. The actual distance of this leg may be considerably more than a mile, especially in rural areas.”

I spent the last 10 years of my career behind the firewall in the LAN and server farm data center; my new job is a little different. Now, I work at a service provider in the customer-facing support group that supports top 10/100 business customers. This support is where the “last mile” comes into play and the CPE demarc that we provide, whether it be traditional WAN serial circuit hand-off with Smart Jack to plug into a router CSU, or an Ethernet hand-off.

Here is a short list of the devices that I get to touch at the CE or PE to provide this Ethernet handoff to the customer, so that all they see as their WAN uplink is a plain old Ethernet port, no muss no fuss. Maybe later we’ll dig in a little deeper to see how we get that Ethernet port out to the customer, no matter the distance from the central office (CO).

Equipment list for Ethernet demarcation at the CE and aggregation at the CO:

  • Hatteras
  • Overture
  • Accedain
  • Aktino
  • Tellabs
  • Actelis

I had never even heard of most of these 6 months ago when I started this job. I have learned from the “old guys” here in the NOC that these products have been merged or acquired into one or two remaining companies of the same names. The devices use bonded pairs that support MLPPP and are sometimes terminated on a router from vendors such as Cisco, Juniper or Alcatel. In most markets, the normal configuration for these bonded pairs is to terminate on an aggregation head-end CO device of the same manufacturer bonding over the copper pairs using standards such as 2BASE-TL, which is IEEE 802.3ah standard for Ethernet transport over G.SHDSL and Enhanced SHDSL. Usually the CPE demarc will be an Ethernet hand-off to the customer and will be limited anywhere from 1.5Mbps to 20Mbps by the physical line speed of the bonded copper pairs, distance to the CO, and line quality.

Often, we enforce QoS at the demarc to prioritize VoIP over the bonded pairs, but sometimes this capability is missing or has been determined not to work. In that case, the enforcement of QoS is done at the PE router, which is not optimal, but serves the purpose of QoS for the customer’s traffic across the provider backbone.

Working with this equipment leads me to being on the phone with CO technicians out in the field wiring these on to cross connects and main and intermediate distributions frames (MDF and IDF) in the CO. Mixing in all those devices with VLAN tagging and QoS at just about any given point on any piece of hardware along the path, depending on the engineering and provisioning teams in a given market, makes it even more interesting.

This is a short and sweet post just to bring this into the fray of topics that I thoroughly enjoy reading here on Packet Pushers. If this post is read by people who blog about such things, I would love to see them comment with what they are writing about.

About Garry Baker

"Keep it simple. When in doubt during design, choose the simplest
solution." - RFC1958 On Twitter @networkdongle

  • Mikerochip

    Very interesting; would like to hear more on: “Maybe later we’ll dig in a little deeper to see how we get that Ethernet port out to the customer, no matter the distance from the central office (CO).”

    • http://twitter.com/networkdongle garry baker

      I mention that as a “maybe” as to future thought and posts to dig deeper, and was hoping to spark interest in both myself and anyone else out there that has more knowledge of such things…I mean the SP i work with can get ethernet to any point, just depends on cost at some point, or so it seems to me in my short 6 months working here in SP…

      I would also be interested and if i can get access to the equipment in the algorithms and or compression beyond MLPPP in the “bondings” over the copper pairs and of course what is called Transport where i work that includes tons of equipment and things like SONET and RPR rings that i do not deal with…

  • Michael Gonnason

    Lots of the tech in the “last mile” are pretty extraordinary. You might become frustrated with your copper though, I hope you have good install and short loop lengths!

    • http://twitter.com/networkdongle garry baker

      no doubt about it…but fiber seems to cause just as many frustrations…IT always boils down to “good install” doesn’t it?

      • http://profile.yahoo.com/WH3X43KRJI7QX2T65YLLM6BD7M batron352

        Yeah, no kidding about “good install”… even in a plain copper LAN… tap the patch cable at the switch and the light goes out, tap again and it comes back on.

        • http://twitter.com/networkdongle garry baker

          right, just today i was sitting in my cube and my monitors started going on and off, i looked under my cube and the power plug was coming out of the modular desk i set at…

          could of freakin’ been electrocuted…

      • Michael Gonnason

        I agree it does, but when you have a customer at 11k loop length, you need to have every bridge tap pulled off, make sure it isn’t wet/damp cable, verify that a squirrel didn’t eat part of the casing.

        Also hope there isn’t any gauge changes, those can kill signals quickly.

        It is a whole new world of external factors compared to a data center.

        • http://twitter.com/networkdongle garry baker

          yeah that was my point it getting this post written, and let the word get out about pushing packets out there in the “wild” has its own set of problems…

          I love to read about SDN and Openflow and all the wizardry going on…

          just wanted to keep it real, or try anyway and this is my attempt anyway…

          so much good stuff going on in the field…