Speaking Engagements & Private Workshops - Get Dean Bubley to present or chair your event

Need an experienced, provocative & influential telecoms keynote speaker, moderator/chair or workshop facilitator?
To see recent presentations, and discuss Dean Bubley's appearance at a specific event, click here

Wednesday, February 03, 2010

Traffic management and offload - diverging solutions

One of the major trends I'm seeing at the moment is that of mobile network offload - typically "dumping" traffic onto WiFi or other networks to avoid congestion from mobile broadband.

I'm expecting it to be a huge feature of this year's MWC / 3GSM in Barcelona.

But at first sight the offload trend seems to be a confusing mish-mash of technologies and techniques, all designed to reduce the impact of bulk traffic on cellular networks, but applied in very different ways. I'm trying to start to categorise these various areas, and also work out how they are being prioritised.

One important fact to note is that "congestion" is itself quite complex. It can be congested radio (RAN) networks in terms of downlink capacity, uplink capacity or signalling. It might be congestion in the backhaul from a cell-site to an aggregation point or the core network. It could be various elements of the core itself - SGSNs, GGSN and so forth. It could be a supporting IT system that handles IP networking (eg the DNS), or the billing/rating engine.

And then we have the various classes of solution. I'm still thinking about a full taxonomy (and terminology), but an initial starting point might be:

- Macro offload onto femtocells (in homes, offices, hotspots or outdoors)
- Macro offload onto WiFi (in homes, offices, hotspots or outdoors)
- Local IP breakout for WiFi & femtos (in the premise or at the broadband DSLAM / cable head-end)
- Managed offload (eg where the fixed or cable access provider, used for the WiFi or femto connection, actively assists in the offload process)
- Macro backhaul offload (eg direct connection to a cache or CDN for "bulk" Internet video)
- Core network offload / bypass
- Content compression (eg video format transcoding)
- Traffic-shaping (eg selectively degrading / capping specific flows or traffic types - sometimes misnamed as "applications")
- Policy control (prioritising specific users / subscribers, or administering specific profiles to some)
- Optimisation of capacity utilisation at the IP level, by changing the TCP protocol or packet scheduling, or by introducing "scavenger class" traffic

EDIT - MAY 21st - New research paper published - see below.

I'm sure there are others - while these segments can undoubtedly be fine-grained more as well. The bottom line revolves around a reduction or delay in the need for capex to enhance network capacity, and as a corollary an improvement in the more nebulous quality of "user experience" - at least for the bandwidth non-hogs.

What's not clear to me is which of these techniques is the most effective or important overall. I suspect it probably varies by operator, maybe even by cell or time. From a top-level ROI perspective, which of them enables spending on network upgrades to be minimised? Or perhaps introduces new revenue streams.

My gut feel is that for networks dominated by *PC-based* mobile broadband, the best option is some form of radio offload. All notebooks have WiFi, and they also tend to have the most complex applications and mashups, as well as being able to spot any degradation of quality or operator "interference" most readily. Having a PC user say "hang on - why's this video performing better over my ADSL line than over my HSDPA, what's going on?" is likely to lead to damaging PR, if you're trying to present your mobile broadband as a direct replacement for fixed connectivity.

I reckon it's simpler and safer just to dump PC traffic to a standard Internet connection as close to the device as possible. For iPhones and some smartphones, the same may be true. But for other devices, it may make more sense to route the traffic via the operator core and play around with subscriber policies, or adapt traffic at the application layer. But I'm sure there are exceptions in both cases - for example, where a PC is actually an operator-controlled netbook.

I'm going to try and resolve the picture more clearly over coming months.

EDIT - MAY 21st - New research paper published - see below.

NEW Mobile Broadband Traffic Management Paper

NEW Broadband Business Models Strategy Report


Anonymous said...

good one dean - the one that makes least sense to me is the "Core network offload / bypass".

Anonymous said...

Why core network offload? A few thoughts:

With a growing set of nomadic wireless users on flat rate data plans generating evermore Internet traffic, does it make sense to bear the expense (and latency) of traversing mobility-focused network elements in the core? Moreover, once offloaded, mobile operators can look to fixed-line transport providers for additional services on Internet traffic at lower costs then they may be able to offer themselves.

A step closer to an All-IP model.

Anonymous said...

more offload http://www.rcrwireless.com/apps/pbcs.dll/article?AID=/20100208/WIRELESS_NETWORKS/100209972/network-optimization-focus-of-junipers-new-offerings


Anonymous said...

Distribute intelligence and delivery to the Radio edge which will allow mobile operators to intelligently offload, establish their own CDN strategy, save both on the core and RAN resources and distribute centralised control to the edge allowing the alignment of content delivery to real-time radio conditions. Offload in the wireless space is but a feature of a new intelligent network edge architecture that will gradually lead operators into ownership of content relationships, provide cost effecient delivery and allow subsciber QoE differentiation.