The Large ISP: An Introduction

1 of 5

By large ISP we mean an ISP with a physically large network, serving many external customers and providing a broad range of network and hosting services. We have in mind ISPs with national up to global reach, including the so-called 'Tier 1' ISPs.

Before considering what a Next Generation ISP might look like, we need a reasonably simple model of a current ISP where we have:

• a number of geographically separated "Points-of-Presence" (PoPs) connected by the ISP's "Core Network". The Core Network will use a variety of routing protocols-notably OSPF or IS-IS, and iBGP-which may be overlaid over MPLS and/or Layer 2 networks.
• at any given PoP, there may be:
  • "border" connections to other ISPs: peers and (except at Tier 1) transit providers-either directly or via Internet Exchange Points (IXPs). These connections all use eBGP.
  • a local Data Centre, where there may be some quantity of:
    • ISP internal services, including: internal management and monitoring systems (OSS), DNS, email servers, etc.
    • customer equipment or equipment provided for customer use.
    • local Content Distribution Network (CDN) equipment: content cacheing. (Connections to external CDNs are, essentially, peering connections.)
  • customer connections: either simple (Default Route) connections or Transit Customer connections (using eBGP). Transit customers may be other ISPs or multi-homed end-customers.

Within the PoP the Site Network will connect things locally and to the Core Network. That Site Network will use a variety of routing protocols and lower layer networks.

Not explicit in this model of an ISP are:

• network management and monitoring: the equipment for these will be distributed across the PoPs and connected by some internal network within a PoP, and across the Core Network between PoPs and to one or more Network Operation Centres. There may be some entirely separate way of reaching some PoPs, for disaster recovery.
• network services: services such as VPNs will overlay the connections and networks shown. General Internet Access is also provided over the connections and networks shown.
• the infrastructure for the PoPs: the buildings, their security, the reliable supply of electricity and cooling, etc.
• the network infrastructure between PoPs: from the fibre upwards.

The Data Centre component of the PoP will vary in size and complexity. For this component, this use case overlaps the Data Centre and the Heathrow use cases. What distinguishes this use case are:

• geography: the ISP's PoPs may be widely geographically spread, so the network between those PoPs may have significant latency and be less reliable than the network within a PoP.
• interconnection with other networks: which is a quite different from connections within a network.

Continue