Test Results Raise Femto Service Concerns

Policy management techniques used by broadband service providers are a serious threat to the quality of services -- particularly voice -- delivered over femtocell connections, Unstrung has learned.

The issue, which came to light during recent service tests, is a major concern, as the use of policy control technologies could potentially cause major service quality problems for mobile operators when they rely on broadband connections from other operators to backhaul femtocell traffic.

When third-party providers are used for that portion of the network, mobile operators do not have complete control over how their traffic is treated. And, as a result, the quality of femto services, especially voice, could suffer.

Broadband test specialist Epitiro recently evaluated femtocell voice and data service quality over the top ten broadband networks in an unidentified country, which cannot be named for confidentiality reasons.

Epitiro found that even the lowest-end consumer broadband service had enough bandwidth capacity to support a femtocell service, so capacity was deemed not to be an issue.

But in four out of the ten cases, the test results showed poor voice service quality. A closer inspection into the degraded voice services revealed that packet loss was the culprit for the quality problems, according to Jon Curley, chief technology officer at Epitiro.

In Curley's analysis, he concludes the reason for the packet loss in those four cases was the policy management used to control IPsec traffic: That has a direct impact on femto traffic, which is encrypted in IPsec tunnels.

"You can tell [traffic policy] is in place by the number of packet drops and higher latency relative to unmanaged traffic. It's in evidence by the amount of packet loss and jitter levels," Curley tells Unstrung. "If you don't have a traffic policy for IPsec in place, it's fine -- it's only when [ISPs] put traffic policy in place and they look at IPsec and they decide they want to control that." So, the femto traffic was being treated like BitTorrent Inc. peer-to-peer traffic by those four service providers, Curley explains.

Nothing to see here...
Femto vendors are quick to dismiss this policy management issue.

"We have heard this question raised from time to time, but it’s not really an issue, and one that is easy to blow out of proportion," says Josh Adelson, director of product marketing at Airvana Inc. , in an emailed response to Unstrung.

"In general, femtocell network designs do not assume special treatment from the fixed broadband provider or ISP, something which is vital to their easy deployment," states Adelson.

He notes that reviews of Sprint Corp. (NYSE: S)'s Airave service, which uses other ISPs' broadband connections, have reported "excellent call quality."

Adelson also thinks it's unlikely that ISPs would intentionally block or degrade femtocell traffic on their networks.

Meanwhile, ip.access Ltd. says policy management isn't a problem for now.

"It's too soon to say definitively that this is not an issue, but so far we have not seen a problem with customers complaining about the quality of voice calls on femtocells running over third-party broadband connections," says Andy Tiller, vice president of marketing at ip.access. "In fact, people often comment that voice quality is better than on the outdoor network."

And Randy Fuller, vice president of business development at policy server specialist Camiant Inc. , says he hasn't seen this problem arise for femtocells.

"The more common problem is insufficient bandwidth," says Fuller.

But Epitiro's test results show that in a worse-case scenario, users would lose service during peak periods, according to Curley. Mobile operators need to be aware of this potential problem, he adds.

— Michelle Donegan, European Editor, Unstrung

JP Curley 12/5/2012 | 4:03:37 PM
re: Test Results Raise Femto Service Concerns

In the US it's not going to be such as issue as the FCC forbids such Traffic Management - which should ensure that IPSec doesn't get downgraded in the US ISP networks (although let's not forget Comcast's well-documented downgrading of certain traffic classes). This might change though - the recent thinking is that they're softening their line on this kind of activity if the ISPs clearly state their policy on such matters.

More worryingly what about the rest of the world where there's little or no real regulatory interest in traffic management.

It's simple - If an ISP sees a lot of IPSec activity, where they are effectively backhauling a Mobile Operator's voice traffic with any payback then it's very, very easy to simply downgrade IPSec's allowed bandwidth and reduce voice quality - and, perhaps more to the point, reduce HSDPA throughputs rates to sub-Macro network performance levels. It could mean that you're better off wandering out to the street to use your iPhone than surf indoors on your Femtocell.


paolo.franzoi 12/5/2012 | 4:03:37 PM
re: Test Results Raise Femto Service Concerns


Femto delivery of voice will not be greatly different than UMA.  I found my cable to be a bad delivery mechanism.  I was not sure if it was a policy issue or a congested upstream issue.  When I was at homes served via DSL I did not have the same problem.




alchemy 12/5/2012 | 4:03:36 PM
re: Test Results Raise Femto Service Concerns For this to work properly, you need end-to-end QoS on the IP network. With a cable modem, you're supposed to push a service flow down to the cable modem using PacketCable MultiMedia to QoS-enable the access network. In most MSO routed IP networks, you then use DiffServ to get to your destination with the appropriate DiffServe code point. Of course, you need the MSO to cooperate with the cellular provider to get this to all work properly. Otherwise, it will be the usual best-effort VoIP that varies wildly in quality depending on local contention for resources. In my experience doing commercial voice over the top of a best-effort broadband network all over the US, 80% of the access network runs just fine running best effort. If you're in that unlucky 20%, you'd better have a good quality fiber alternative or have a very short copper run to the DSLAM.
JP Curley 12/5/2012 | 4:03:36 PM
re: Test Results Raise Femto Service Concerns

Alchemay is missing the point here entirely. The goal of a Femtocell is to provide Voice quality on-a-par with the Cellular network, MOS levels of 3.4 are perfectly acceptable in this space - it's not a replacement for POTS/PSTN. Additionally the model is parasitic by nature - end-to-end QoS will never be an option - how on earth can a Cellular Operator hope to define SLAs with all the potential fixed-line operators potentially backhualing their traffic.

Additionally there's a big requirement to offload the mobile data traffic (think iPhone, etc..) from overly contended macro basestations - that's when traffic management really burns the potential.

rfulleremp 12/5/2012 | 4:03:35 PM
re: Test Results Raise Femto Service Concerns HomeNodeB is on the right track. Not all policy management techniques are the same. As John Curley describes, Epitiro is describing a situation where there is plenty of bandwidth available but yet IPSec tunnels are selectively treated negatively. This can only be done with DPI, not other sorts of policy control, and only then when allowed by the regulator. Note that there are policy management techniques that could create the reverse situationGÇöhigh quality femto-based traffic even in a congested network.
paolo.franzoi 12/5/2012 | 4:03:35 PM
re: Test Results Raise Femto Service Concerns


Here is all I can tell you.

I turned off my UMA phone because the voice experience was much worse than it being on my single bar Cellular hookup.  Then, T-mobile made it all moot a few months later with a new cell site somewhere nearby.



t_newt 12/5/2012 | 4:03:27 PM
re: Test Results Raise Femto Service Concerns

This shows the advantage of having your own network to connect the femtocells to.

Just build a femtocell into every DSL or cable modem, and every wired customer becomes a basestation for their cell network. Since it's their network they can use bandwidth above what the customer is normally contracted for, so the customer doesn't see any slowdown when someone connects to his femtocell wtihout his knowledge, and they can control the priority to ensure a good phone experience (no dropped packets, no dropped calls).

I'm surprised AT&T, Verizon, or the cable companies in a deal with Sprint haven't implemented this yet. Though femtocells are so new that I'm sure a lot of the problems that show up when they become concentrated in an area aren't yet known.

Sign In