Network service providers running IP networks ‘hotter’

Global service provider switching and routing spending is falling, and we haven’t seen the bottom.

The service provider switching and routing market fell 20% in 1Q09 compared with 1Q08 and will likely continue to decline. Yet carriers continue to report robust IP traffic growth, and this is only the tip of the upcoming video tsunami. Cisco’s recent Visual Networking Index predicts IP traffic will increase fivefold from 2008–13, with the largest growth segment being consumer Internet.

We attribute some of the recent router market decline to recession-induced postponement of strategic IP transformation projects, but service providers are also delaying short-term investment to shore up financial results. The result is that operators are running IP networks hotter.

There are no standards for carrier router loading, but service providers typically run networks at 40–50% utilization. They are comfortable that core routers can run without incident at 60–70% loading, but once average loading reaches 80% capacity upgrades are needed to ensure high availability. Edge router loading is harder to quantify as high subscription rates are typical, as overloading will have more localized but equally detrimental effects.

The IP network won’t break, but you may think it has

IP traffic continues to grow by all measures, but many network operators are postponing capacity upgrades and thus running IP networks hotter, which means they are increasing traffic throughput without upgrading equipment.

There is no need to worry that telecom wire centers will burn, as survivability is ‘baked in’ to the DNA of IP packet network routers. However, performance does suffer during traffic peaks, and those peaks occur more often as router traffic loading increases. In addition to dropped packets, other performance impairments include delay or jitter (variable delays) and out-of-order packet delivery. In response, end-user devices retransmit data (or unintelligible conversations for VoIP), adding further congestion over a longer period of time and extending traffic peaks.

The net result is a degraded quality of experience (QoE) for end users, be they man or machine. Forecasts for booming Internet traffic growth rarely consider the impact of degraded service levels due to overburdened, packet-dropping core or edge routers. This becomes more of an issue as most regulators side with net neutrality proponents, unconcerned about funding infrastructure upgrades because Internet service levels haven’t historically been a problem beyond the access bottleneck. Are regulators in for a rude awakening when VoIP or over-the-top (OTT) video providers can’t do business over the Internet due to its unpredictable performance?

Savvy enterprise network administrators who have experienced distributed denial-of-service attacks on their IP networks know just how disruptive such overloading can be to their business and have spent millions to protect against such attacks. Will running IP networks hotter cross a similar usability ‘tipping point’? What happens if one of many routers fails during a traffic peak? Router software is equipped to route around failures, but no one really knows the impact on end-users. We could find out soon, however, if the current trend continues. Our advice is to not trust your mission critical applications to the best effort Internet. Nor to expect Application Acceleration or WAN Optimization gear to ‘fix’ a poorly performing, non-deterministic, best effort Internet. Instead, invest in higher quality services.

When is ‘best effort’ just not good enough?

We don’t believe service providers will allow serious service degradation to occur for their premium IP services. But we do believe public Internet–based services are at risk of suffering from service degradation, with the result that end users’ quality of experience may plummet – particularly in North America with its intense competition and minimal regulation. More centrally controlled countries such as Japan, South Korea, and France have higher-performance Internet networks mainly due to greater regulatory oversight. But North American ISPs don’t want greater regulatory oversight so they find themselves on a slippery slope.

One encouraging sign for higher investment levels is that residential broadband prices have stopped their decline and are starting to increase, but it is not clear if they have reached an acceptable profitability level for North American ISPs. Our advice is to watch just how hot ISPs can go and monitor the impact on Internet performance – and to not put too much faith in the Internet, as past performance is no guarantee of future results!