The wireless industry is wise to embrace end-to-end IP networks. 4G is IP, exclaims an infrastructure vendor. As a long-time IP advocate, I am pleased to see the movement in this direction. Carriers that implement the newer wireless access technologies are positioned to capitalize on operational simplicity, reduced cost, and flexible service creation that IP/Ethernet networks offer. Although we’re off to a good start, we have much to accomplish before fully realizing the value of extending IP to the edge.
Before examining the gaps, I’ll describe my desired end state. I want to design, operate, and maintain a “wireless” network using the same practices as Tier 1 ISPs, the best of which know how to cheaply and efficiently transport bits. Note the use of quotation marks in my previous sentence. Wireless describes the access mechanism only. Why should we invent new ways to run IP networks? The solutions are already out there.
To reduce my costs of offering an IP service, I need infrastructure that supports wireless access to function like IP routers. We’re building IP networks, so let’s push vendors to implement the feature sets available on carrier routers. If the provider’s engineers can interact with packet core gateways and base stations as they do with today’s routers, I can adopt the best practices honed over the last fifteen years at the Tier 1 ISPs. Think about the advantages in terms of staffing, augmenting capacity, and ability to quickly launch new services that would accompany this shift.
I’ll get specific about what I seek from network designers, vendors, and standards bodies. Here’s an inexhaustive list.
Routing – To extend IP to the edge in a meaningful way, the same dynamic routing functionality should be available to the RAN, including the cell site. Why to the cell site? Packets always head to the core, right? This is true…for now. Deploy the same routing protocols used in Tier 1 ISPs. Robust implementations of OSPF and ISIS avoid one-offs in routing design. ISIS, in particular, is favored by some large carriers for reasons such as ease of extensibility for IPv6.
Backhaul Avoidance – Let’s avoid the backhaul bottleneck that some providers have experienced in 2G/3G networks by localizing routing decisions. If I’m using my 4G-enabled device at my neighborhood park to share content with a passer-by, why should traffic traverse the core and backhaul network? Designs could be jiggered to accommodate network-layer mobility at the cell site. The longer term fix will likely be to modify the reference architectures to meet the business realities of production networks.
Infrastructure Management – My goal is to manage the entirety of the infrastructure using similar means. Operators modify router configurations with scripts, command-line interfaces, and API. Routers are accessible with SSH, and users are authorized with TACACS+ or RADIUS. Routers use syslog for logging to a server. Router capacity is displayed with open source tools such as MRTG and Cacti. I have yet to encounter a base station, for example, that does all these things (No, having a vendor use hidden Unix/Linux shell access to create a hack doesn’t count).
IPv6 – Ignore or half-heartedly implement IPv6 at your peril. Over the last ten years, routers have made significant strides in being able to implement a mature IPv6 transport service. Mobility elements that I’ve worked with do not match the IPv6 capabilities of carrier routers. Being able to forward IPv6 packets isn’t enough. Operators need the full range of capabilities for which they currently rely on IPv4. Too often vendors issue a cavalcade of “buts” in discussing IPv6. “Our gear is IPv6-ready…but it can’t forward packets at line rate…but we didn’t implement an IPv6 version of that firewall rule set…but the infrastructure protection mechanisms that you use today aren’t ready.” Push for anything less than like-to-like functionality or prepare to be tripped up by that one feature that you didn’t realize was critical in offering a service…or billing for it.
Visibility – A simply designed L1/L3 network maximizes visibility into the network elements, thus reducing outage duration and the overall complexity of the service assurance activities. The value of simple IP tools like ping and traceroute can’t be understated. Imagine a scenario in which your NOC receives an alarm that a cell site is unreachable. What next? An engineer familiar with troubleshooting IP networks might first attempt to ping an IP address closest to the customer. This could be a base station or cell site router. An enormous amount of troubleshooting is eliminated when IP connectivity to the cell site is confirmed.
I often discuss how the mobile operators will find themselves behaving more and more like the Tier 1 ISPs. I see this as inevitable based on the challenges the providers face. I believe the ones that quickly adopt this mindset will thrive. What I’ve described above is one step in getting to the all-IP nirvana. Hope to see you there.