Earlier this month during our Cambridge Wireless Small Cells SIG event, I presented a small quiz in the final session. The first part of the quiz was titled "LTE Deployment Dilemma" and it generated lots of interesting discussions. After the event, I did a more detailed writeup of that and Cisco has kindly published it in their SP Mobility Blog. Since many people have told me that they cannot anonymously post comments there, I am now bringing it to this blog. I am interested in hearing what others think.
Here is the complete post
Having personally been highly involved with the 1st GSM, 2G, 3G and 4G rollouts; one would think the problems would be solved after 15+ years. This is not the case!
ReplyDeleteMainly I see two issues with LTE which are severely overlooked: (There are more!)
1. Scale-ability, current implementations of 3G under-estimated the saturation of users.
Now there are many 4G areas that cannot provide adequate coverage due to active population growth.
2. Total bandwidth available will not handle the near-future requirements. 2 million new users were added the first weekend the new iPad came out, just as an example. More and more people are using the internet to stream video, view online data and cloud sharing uploads/downloads, just to name a few resource hungry apps.
Surely the issue is not interference minimisation / and throughput maximumisation in the sense outlined, as end users hardly consider such entities.
ReplyDeleteThanks Zahid! excellent analysis. IMO, since Cell edge or location of UEs can not be predicted acurately, for best use of spectrum HetNet approach (Same spectrum for Macro and small cells) would be used. Secondly, the ABS option to be kept open for the SON to take care initially and once the network matures and Rel 8 devices go down, it can be fine tuned depending on the location of the small cells (geographically). Please share your views.
ReplyDeleteManoj, there are no correct answers but I agree with your view.
ReplyDeleteI would consider a scenario with a few ‘large’ small cells to have operator 2 type of deployment whereas within a cell with lots of ‘smaller’ small cells the operator 1 type of deployment would be better.
ReplyDeleteBack to basic, why small cell. To complement the macro, hot spot coverage and data offloading. In 3G small cell, we have interference issue when deploying shared frequency small cell when macro signal is too strong. So less headache would be option 1.
ReplyDeleteThe ground reality is generally a lot different than theory. Metrocells often face interference not just from Macrocells but also other Metrocells. The ABS patterns are not just straightforward Macro to pico case but even pico to pico and multiple macros to pico.
ReplyDeleteUntil all the handsets and other dongles could be upgraded with advanced interference cancellation receivers, there would be many scenarios where deployment option 2 may be chaotic. Deployment option 1 can serve the users well in the meantime.
We can sacrifice efficiency for reliability in the meantime.
Thanks S.C., I like your viewpoint.
ReplyDeleteGood Blog ,
ReplyDeleteHowever , Deployment model should considered the Centralized SON or Radio Network Optimization Solution to Guide the resource planning of Small Cell and Macro Network. With Deployment model , most of use-case would be dynamic in nature and with Centralised and Coordinate SON solution , My believe operator should deploy with Second Options.
Sanjay
Zahid,
ReplyDeleteI have been involved in radio network planning since 1996 on the financial and entrepreneur side, working for both operators and equipment vendors. On the one hand we have seen sophistication emerging in planning tools, and on the other side also in the air interface technology. Still, the terrain remains the same, albeit we may have seen substantial additions to building mass, population densities, and even to roads. The Services offered by the Services Delivery Networks have dramatically changed, as we have moved from SDN optimized voice codecs to full IP devices. The need for small cells has emerged out of the need for bandwidth, the limitation in user device uplink capabilites, and the spectrum interference issues. Spectrum licensing models in use today make spectrum a scarce resource. As 3G is introduced in Bangladesh, 4G is introduced in Italy. The world is complex, and planning wireless networks did not get boring.
Once driving down the freeway in the Seattle, WA, area, my client suddenly pointed to a cell tower in the middle of the Woods, near the freeway. He told me the story of how that cell site had been a matter of pondering by the network operator staff, as it always reached capacity and caused problems in the afternoon. It was in the middle of nowhere, and nobody had considered configuring the site for other than minimum load. It turned out that there was an alltime traffic jam every afternoon as people got off one freeway and on to another, at rush hour. Only by getting out of the office was the capacity planners able to understand the issue. People took time out to do their calling while moving at snail pace, if at all.
When I plan networks, I make a capacity plan, and a roadmap of services. My objective is to plan funding requirements over time, and as part of that I need to get the capex Investments right over time. If I have spectrum resources, I check on the feasibility of the plan in the light of available spectrum over the planning period. I would expect the network planners to take the capacity plan and services road map into account when they configure the network. The war stories on how this link between financial and marketing planning on one side and network planning on the other side breaks down are legion.
In general, there is a lot of static planning made by new venture founders and project owners that are not coherent with the lifetime of the asset. As network assets become cheaper on a per unit basis, the human resources needed for planning, decommissioning, redeployment have increased in price. The cost of being wrong is not insignificant, and "doing it right the first time" is still relevant.
I understand that the deployment dilemma you are highlighting may be a matter of principle (damned if you do, damned if you don't), but it seems simplistic. My principal concern would be that you are going to find the deployment model proven undesirable for reasons that you do not take into account in your deployment planning - the business evolution by cell, and the implications for user experience under each deployment scenario.
Thanks for providing LTE network planning insights.
Trond
Thanks Zaffareen, Sanjay and Trond for your comments.
ReplyDeleteZaffareen, I agree with your viewpoint but the deployments are generally complex, with little choice about the size of the small cells. Most cells will have a mix of 'smaller' and 'larger' small cells so it is generally hard to predict. Also, different cells will have different combinations of 'smaller' and 'larger' cells which make things complex.
Sanjay, agree that SON and especially C-SON will play a big role in LTE deployment scenarios. While SON is a great technique to optimise the networks, I still believe it has some way to go before it can handle the interference scenarios properly. One of my worries is, from a network point of view the Interference may look like its being handled but the UE may have a bad QoE. This is why some of the operators are still doing extensive trials to do Interference management in HetNet scenarios.
Trond, I have been a party to similar experiences as you have described on freeway in the Seattle, WA. I have seen an operator panic and throw more small cells to handle such a situation. Sometimes a real world experience of understanding what people do and how they behave can help alleviate problems. The SON algorithms have not yet reached that stage where they could read our minds and make their decisions based on that. I agree with your cautious statement that its better to get it right the first time or there is a price to pay. This is what I am trying to bring to the attention of everyone. Just because a feature is available from 3GPP doesn't mean it would suit one's requirements. Every solution has to be checked on a case by case basis and then a decision has to be made. The problem is that the more you look at things on a case by case basis, the less automation is possible and then the costs rise significantly.
Generally speaking, from something similar to Erlang's law, there is a trunking effect: when you merge two spectrum pieces the total capacity is larger than the sum of each one. This is all the more relevant here that the total capacity is 20MHz which a single mobile can handle for itself. So I would focus on not splitting the spectrum and deploy all the most advanced dynamic spectrum/interference management features of LTE. I guess this would be the most futureproof deployment. If you are interested we could work on theoretical models that assess the gain.
ReplyDeletePierre, thank you for your comment. Agree with your point that "total capacity is larger than the sum of each one." If we split the spectrum, there is additional guard band which we will have to think about and each of this smaller band will have its own reference symbols and other information that will in effect waste the resource blocks. The question is more of a theoretical one that asks whether its worth going through the headache of interference management or instead do a simple CA deployment. In an ideal world where we will have all devices equipped with advanced receivers, HetNet deployment would be great but what now?
ReplyDelete