Monday, November 23, 2009

Promise and Issues with LTE















While we are yet to come out of the confusing muddle of 3G, LTE is there knocking on the door. It is very important to understand what LTE is going to offer and what it is not going to offer (at least as of now) so that we dont end up with pipes ready with nothing to carry on or worse still, vice-versa. What drives this LTE evolution is the exponential data growth on the mobile networks seen in the current years and even stronger growth expected in the years to come. The SGSN- GGSN part of the network which is responsible for the data or IP side of the traffic have gained prominance in the recent time, which have seen falling Voice ARPU, data come up as the only savious in the bloodbath we see in the market. This implies the need to optimize the network for data intensive networks, eventually converging the enter network to a packet switched domain from the current combination of Circuit switched (for Voice) and Packet switch (for Data) domain. As per the research report (courtsey: Analyses research, 2007), highest data guzzling applications identified as 1. TV/Video (Average 140 mbps per sub @15 mins per month), 2. Voice (Av. 21 Mbps per sub @ 300mins per month) and 3. Audio (9Mbps per sub per month)-which clearly points out to the significance of packet data.

This has driven the global operator community to look for an upgrade from the current 3G rate of 14.4 Mbps DL and 5.72 Mbps UL to 100-173 Mbps DL and 50 Mbps UL (Upload) in the LTE scenario. With such a huge amount of data moving across the network, it makes sense to have a converged core with Voice as one of the services.
In view of the prominent driving role of IP, the architecture is defined under SAE (System Architecture evolution) under 3GPP. The specfication (3GPP rel. 8.0) defines defines enhanced efficiency with 326 Mbps peak download (20MHz spectrum) with 4x4 antenna and 173 Mbps with 2x2 Antenna and support for 200 active data users each 5 MHz cell. Cell size support in terms of spread is recommended as ranging between 5 kms and 100 kms from best to acceptable performance. What it essentially means from the network perspective is that the core of the mobile network which earlier comprised of RNC-MSC for Mobile voice and RNC - SGSN - GGSN for data traffic, collapses and merged to what is called as Evolved Packet core in LTE.
The ePC comprises of, the MME (Mobile Management Entity) is the call processing entity, S-GW (signaling Gateway) acts as Local Mobility Anchor while P-GW acts as IP Anchor responsible for allocation of IP addresses and management of IP connections. One important requisite that 3GPP has defined is the interoperability with 2G/3G networks, since it is anticipated that LTE is not likely to see uniform deployment, and will exist as island in the heavy data zones.
This could mean multiple ways of deployment as shown in the exhibits above:
wherein, as primary option, as a stepped migration, simply the GGSN is replaced with the P-GW, connectivity still being the UMTS interface (Gn) running on GTPver 1. Second option being pure LTE option, with SGSN connecting to S-GW and MME on LTE Interfaces (S-4 and S3) while P-GW connects to S-GW on S5. finally the last option would be direct tunnel option which is derived from the option presented in 3GPP rel6/7, which attempts to bypass the SGSN by connecting the RNC directly to S-GW on S12 interface. The essential requirement for the option 2 is that the Iu and Gn should be in the same network.
Key concern in LTE:
Being a pure IP system, (E-UTRAN-EPC), which means it is not designed to carry voice as circuit switch. There are three propositions to handle Voice in LTE
1. As pure VoIP(as is envisaged in LTE) this would be controlled by IMS Core, meaning VoIP in LTE and PS- CN in UMTS.
2. As a combination of CS and VoIP, where traffic in UMTS will be carried as single radio VCC that is Packet switching for signalling and Circuit switching for Voice, and in Pure LTE environment, voice will travel as VoIP. VoLGA (Voice over LTE via Generic Access) is derived out of this.
3. The last option( actually the first in precedence of appearance) is Circuit switch fall back (CS Fall Back) where in while the data traffic uses the LTE components, when it comes to Voice, it goes purely on Circuit switching. The LTE variation of MME-MSC interface (SGs) is used to register the user on UMTS network which pages the device in case of incoming call and connects the device through RNC. For outgoing call, calls are directly routed through the RNC-MSC route, bypassing the MME. One big issue is that immediately at the moment the device falls into UMTS domain with ongoing data transfer getting reduced from around 100 Mbps of LTE to 14 Mbps of UMTS. We will evaluate the issues in following posts.
























Monday, November 16, 2009

International Long Distance: Opportunities and Drivers


 Government has so far given 23 ILD licences since the opening up of the sector in the year 2002 as on March 2009. The annual license fee has been reduced to 6% of the Adjusted Gross revenue (including the USOF), since 2006 and the roll out obligation for the operator is to set up on ILD gateway within a period of three years. While such policy initiatives mark high growth era in terms of usage, it further puts more emphasis on the increased competition, especially with big ticket telcos like Singtel, BT and Orange jumping into the fray. The great difficulty is that this is one business segment which can not be driven by price fall with controlled ARPU only, rather multiplicity of the carriers push for a tough tightrope walk. Earlier the difficulty was more on bilateral termination arrangement with US operators which had low rate for subscriber usage in their own and were bleeding under reciprocity on account of high rates in India.

 

Market Distribution:

As by end of 2008-09, the market still is largely held by Tata Communications, at 54.6 % market share, followed by Reliance communications at 11.5% and BSNL at 7%.

 

While the market size is quite big (pegged at INR 15000 Crores in 2008-09), with a sharp growth from flat-ish 2006-07 to 2007-08 (from INR 11506 crores to 11532 Crores), with Vodafone, BT and BSNL showing the highest growth (450%, 166.7% and 120% respectively), the pie irrespective of division is big and attractive enough to push every one on the ILD bandwagon. Eventual differentiation will be based on the capability to grow the market, with gaining the market share, and capability to sustain competition on the basis of well oiled global alliances. This is a market battle which will in great part be played outside the country. TRAI has raised the call termination charges for calls terminating on Indian operator’s network from 30 ps per minute to 40 paise per minute to bring in benefit for Indian long distance operators, while for the moment keeping the termination charges for 2G and 3G at the same level based on the inputs collected from the operators as per the notification effective from 1st of April, 2009.

While the interconnection charges are broadly defined in the tenets of WTO agreement (GATT), 1997, policy wide tweaking is generally permitted in order to ensure that the providers in the country could make profit out of international operations (and passing off some of it towards USOF) and at the same time ensuring that the charges are conducive the overall market environment. IUC consists of Origination, termination charges, Carriage charge and transit charge. Termination charge has been fixed at 0.30 INR per minute. In case of non-availability of direct connectivity, the charge for carrying the traffic through another NLDO is termed as Transit charge which again is left to mutual workability with the ceiling of INR 0.20 per minutes. Carriage charge has been left to forebearance (mutual workability) between the ILDO and Foreign operators. It is the responsibility of TRAI to ensure while by reducing the termination charges it makes connectivity to India easier from overseas, it needs to raised to a level that connectivity to foreign shore from India does not end up as a loss making enterprise for operators (some charging as high as INR 3.00 per minute for termination). In view of this, while this is an extremely attractive market, it is driven much by global market forces and policy directives and the ability to understand and forge better alliances will spell the success.

 

Monday, August 24, 2009

The Case for MPLS Backbone

It is a known fact that most of the existing operators have long distance backbone built on ATM/SDH based network, but then, so is the fact that almost all major operators have either moved or are contemplating move towards MPLS based backbone. The wave started as Overlay network (separate SDH and MPLS based network) is giving way to the converged MPLS core. Although independent reports like Infonetics cite operator response to near 100 percent as Ayes in favor of MPLS, there are still apprehension in traditional operators, looking at the cost implications of MPLS vis-a-vis a layer 2 Optical network. The services which needs to be considered while deciding on the move can be categorized into existing services and aspired services. Basic voice services together with leased line, frame relay will fall in the existing services (requiring separate physical layout for each customer), while high speed internet access, VoIP (or IMS based services), Video (specially those requiring multicast like e-Education and IPTV) and IP VPN services can be looked up as aspired services which decidedly require MPLS.
The growth rate of traditional data services seems stagnating at close to 8% (reference IEEE Report on Quantifying the Value Proposition of MPLS Evolution, Bell Lab report), and the growth rate of new data services are growing at 50-75%. Looking at the figures, writing on the wall seems to be quite clear. This interest in the new data services is evident from the amount of interest with which all operators with long term view of the market, seem to be looking at 3G and LTE, both referring to increase in the amount of video(with a significant amount of multicast traffic) and data traffic, which subscribers are demanding and for which operators are gearing up, looking at the amount of money people are willing to spend on it. The IEEE study report mentioned above puts the Present Value saving (taking into account new revenue figures) of 23% over the ATM based network and 10% over overlay model. The biggest source of PV saving is new IP VPN services. Another advantage of this model is that is wonderfully complements the SDH network, which it uses as undelying transport network at the same time using features like MPLS TE-FRR to enhance the redundancy by working out the backup paths between to nodes of connectivity.
This redundancy however, can not be taken as the de-factor prime mover for establishing an MPLS network, rather this enhanced redundancy is a default benefit arising out of the attempt of the operator to re-invent its service offering with focus on high growth areas like the non-traditional data services. There are cheaper ways of enhancing the SDH redundancy (for instance x-ing the links by two) if one is only concerned with making the currently available basic voice and conventional data services fail-proof, which already is plagued by falling ARPU, low subscriber interest and high cost of customer acquisition, but if the eye is on the future and the intent is to remain "Forever Young", MPLS is the way to go.

Tuesday, August 4, 2009

Carrier Ethernet-Choices- Present and Future

Ethernet, conceptualized and initiated as a technology to connect nodes within a small, limited area, administered by common policy, in a defined physical space brought with it a distinct advantage, working on Collision Detection. This allowed physical reuse, no waiting period resource wastage and high bandwidth together with highly granular bandwidth control. Then with the opening of the dotcom floodgate with Yahoo!! and Googles of the world, substantiated by newer application like online trading, Instant Messaging etc. made it very common for people to stay connected on the net for the entire of their waking hours, although their talk time stayed limited to six to seven hours for high usage subscribers. This exemplifies the importance that data traffic has gained in the last decade. While the voice volume was more or less not increasing per sub, or at best increasing at very slow pace, the exponential growth of data as a tool of communication, empowerment and entertainment attracted the traditional service providers to stake claim in then tempting market segment. The large behemoths walked in to kill the smaller ISPs, positioning themselves as converged service provider. Immediate access requirments were met with by putting in place a parallel data network, carrying date on the IP devices, while voice continued to move on the physical layer, with a distinct segregation of Circuit switched (Voice) and Packet Switched (Data) traffic. This did impose some demands on the data networks as it came out of the protected LAN environment to the large public space, specifically in terms of High availability, High capacity and scalability. While this did bring in business, it came with its own share of trouble. There was a need to ensure the upkeep and growth of two separate networks, both needing different skill sets. The pre-existing voice carrying network was already on the last legs, with PSTN equipments getting phased out, OEMs pulling out the support and discontinuing the products. Dual cost and worry of the future, pushed operators and vendors alike to look for the means to put in place a single network which could carry diverse kind of traffic. This put additional demand on the existing carrier ethernet network, to ensure voice grade traffic on the network, support for large number of subscribers (identified in IP world by MAC), Mechanism to manage the limitations of MAC addresses and VLAN, technical ability of the network to provide, what is termed as Voice-grade 50ms protection.
The initial struggle between RPR(Resilient Packet Ring, IEEE802.17) and VPLS as a mechanism to provide for 50ms protection finally ended in favor of VPLS (which does the protection based on TE-FRR- Traffic Engineering - Fast Reroute). The key reason for this outcome, apart from industry support, was the traffic engineering benefit that VPLS brought with it. While both VPLS and RPR based solutions used Q-in-Q for the VLAN scalability, with the technology question for the backbone already settled in favor of MPLS long time back, it made sense to extend the same technology to the aggregation and access with VPLS (or Layer 3 MPLS, if need be so). MAC scalability still remained an issue, further the network with Labels being distributed on DoD (Downstream on Demand) was not as deterministic as we would want it to be in order to be able to phase out the SDH based network completely. Physical layer did not give up however, backed by IEEE it came back with technology called PBB (Provider Backbone) to address the MAC scalability based on IEEE 802.1ah. This proposition put forward my ethernet proponents like Nortel, based MAC scaling on the idea similar to QinQ, that is encapsulating customer MAC in Provider MAC. While it used Ethernet OAM to provide for providing reliability, it left the QoS issue unaddressed with only basic ethernet features like IEEE 802.1p and 802.1q (priority and queuing) for QoS. Which was a feeble attempt against strong traffic engineeting offered by MPLS. MPLS forum responded to the MAC challenge with VPLS-PBB integration (Draft-Balus), while IETF got together with ITU-T, leading to the scrapping of earlier ITU-T proposition for Aggregation, T-MPLS, forming a Joint Working team (JWT) in February, 2008, giving way to MPLS- Transport Profile, which unlike its predecessor could interwork with MPLS on the backbone. IEEE on the other hand responded to the QoS challenge by statically provisioning multiple paths with deterministic QoS, based on IEEE 802.1Qay. Both MPLS-TP and IEEE802.1Qay are currently under standardizationa and are likely to be vetted by early 2010. Although industry majors like Alcatel-Lucent, Ericsson, NSN and Huawei have thrown their weight behind MPLS-TP, wholeheartedly since BT abandoned its PBB plans, it is more of wait and watch situation. Both the standards will bring with them the promise of a network with Quality of Service, Capacity and High availability numbers at par with or better than the traditional network that they seek to replace. A converged network would also mean a single network on which entire traffic including mobile traffic will flow, requiring support for features like Synchronous Ethernet and 1588v2 etc. which are leading the standardization work covered in MMBI(MPLS Mobile Backbone Initiative) undertaked by MPLS forum. The latter will be an interesting subject in itself, while the key summary points at the present are:
1. VPLS is technology of choice in the aggregation.
2. VLAN scalilibility is addressed in Metro or Carrier Ethernet scenario by QinQ, and MAC Scalability with IEEE 802.1ah on access or by PBB-VPLS integration on the aggregation.
3. Future choices are split between PBB-TE and MPLS-TP, both being in pre-standard stage.