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 discuss Dean Bubley's appearance at a specific event, contact information AT disruptive-analysis DOT com

Tuesday, May 19, 2009

Evolved Packet Core in LTE

Many people in the mobile industry are unaware that LTE, the radio Cinderalla, also has an ugly sister, called EPC (Evolved Packet Core - originally it was called SAE).

Many things about EPC are actually worthy - "flatter" IP network, lower latency, the ability to mesh between base stations, getting rid of a lot of other legacy equipment like SGSNs and GGSNs and so forth.

However, there is one thing that appears to be missing - the option for local offload of traffic. Everything is still backhauled via various gateway and policy boxes. It's not possible to say "that's a 3GB web video download - just short-circuit it straight to the Internet". It still has to go through the core network.

If I'm understanding this correctly, this has a couple of important implications:

- Roaming traffic will still need to be backhauled via the home network, even if it's just plain-vanilla Internet data. You can't just dump it to the web in the visited country. So you can forget about LTE roaming being able to compete with local WiFi (or a locally-acquired LTE SIM) on price, because there's still a huge extra overhead of network elements and unnecessary transport involved.
- In LTE femtocells, the option for "split tunnels", for example to offload web traffic or operator services like IPTV in the home gateway, or DSLAM / cable headend, won't work. Again, the traffic needs to go through the mobile EPC core. You can also forget about using LTE instead of WiFi for UPnP or other types of home networking via a femto - the traffic will need to be "tromboned" in and out.

9 comments:

Anonymous said...

Come on Dean, why would any incumbent vendor who promotes the LTE vision and sells core equipment want to offer a break-out solution close to the user? There are answers, they just don't come from this crowd.

Zigurd Mednieks said...

Back when I was working on GGSNs, I calculated that traffic through a metro area's GGSNs would be about the same as through one edge router on a cable or DSL network. Surely it is more now, but not in proportion to the size of fixed-line backbones.

It all might add up to a small-enough amount of traffic you would have to come up with a very specific use case - like wanting to get to a CDN - and wouldn't the commercial CDNs be on the mobile core networks anyway?

Anonymous said...

Actually, local breakout is possible - you should really get some good book on EPC :)

alexander Esser said...

In roaming cases, breakout in visited network has been in the standards and supported by SGSN/GGSN network elements since the first days of GPRS. However, operators decided to always backhaul to the home network for "business reasons".

Martin said...

Hi Dean,

Home Network: The traffic always HAS to go through the Serving-Gateway and the PDN-Gateway to hide the mobility of the user as he roams to the network. It's a basic need of IP. As one of the commenters above, I think it doesn't make much of a difference, the PDN-Gateway is mostly a router like in fixed line networks as well. O.k. you can put applications on the router such as traffic shaping, prepaid charging interfaces, etc. but it's still a router and not so much of a problem.

Roaming: Here, the backhauling of the plain Internet access is mostly necessary because of charging, prepaid, etc. But the standards also allow for a local breakout in the visited network. It's just not used.

I don't think the cost of backhauling to the home network is so prohibitive that it's impossible to compete with local Wi-Fi, even it's free, because it's usually not where I need it :-)

Cheers,
Martin

Dean Bubley said...

Thanks for all the comments here - looks like I need to delve a little more deeply into the precise capabilities of EPC.

I still think the roaming backhaul vs. offload issue is important - it also makes it more difficult to do clever things like sponsored/"free" LTE access paid by 3rd parties.

Dean

Dean Bubley said...

Also - Zigurd, is there not a case for CDNs to be further distributed, eg out towards the edge of the network? For example, at localised RAN backhaul aggregation points, DSLAMs linked to femtos, or even at some of the larger cell sites?

KMitch said...

I believe EPC and SAE are different terms. EPC includes the new functional elements of the MME, SGW, PGW. SAE includes the EPC and the various access networks: 3GPP access (LTE included), non-3GPP trusted access (e.g., CDMA, DSL) and untrusted access (Internet). SAE would include the ePDG for tunnel termination of the untrusted access network.

Anonymous said...

EPS = EPC + EUTRAN per 3GPP. Though LTE/SAE has stuck on due to historical reasons. SAE was rechristened EPC and LTE as EUTRAN. Together they form EPS or Evolved Packet System.