Should Deutsche Telekom be allowed to deploy vectoring?
After some intense detective work we are ready to revisit the regulatory issues with regard to rollouts of VDSL vectoring. This is a hugely important issue: Rene Obermann, Deutsche Telekom CEO, has indicated that the operator will increase its superfast broadband coverage in Germany to 24 million homes by 2016, double the current 12 million VDSL homes passed, if it is allowed to deploy vectoring. The technology would allow download speeds for these households of 100Mbps. This could entail billions of Euros in spending for vendors and has the ability to reshape the German fixed broadband market, Europe’s largest, where the incumbent has been losing out to cable operators’ aggressively priced high speed offers.
But the big if in Rene’s proposal is whether regulation will permit the deployment of vectoring. The issue is one we have covered before: whether sub loop unbundling (SLU) and vectoring can coexist. Obermann’s position is that SLU should be banned, as he notes has already occurred in Belgium, because the vectoring rollout in Germany would be incompatible with SLU. So let’s examine that argument and also the underlying premise which is that SLU should be banned because it is not sufficiently useful to be allowed to stop the incumbent’s vectoring plans. This article will concentrate on whether cross DSLAM level vectoring allows SLU and vectoring to coexist. An earlier blog covers the issues of using DSM level 2 to ensure the coexistence of SLU and vectoring.
Part 1: cross DSLAM vectoring: a solution?
Cross DSLAM vectoring refers to a situation where each operator deploys its own DSLAM but lines are vectored not just across each DSLAM but also between DSLAMs. In this way, at least in theory, SLU may be possible and both operators could enjoy the benefits of vectoring. Whilst that is the theory the practical reality is somewhat different.
Case A: cross DSLAM vectoring with DSLAMs from the same vendor
Inter-DSLAM bandwidth requirements
There are a number of drawbacks here. The DSLAMs need to exchange a lot of information with regard to crosstalk cancellation and there are also quite strict latency requirements for this information exchange. Practically, this limits the distance between the operators’ DSLAMs. For the time being the DSLAMs would likely need to be installed on the same rack to ensure the communication bandwidth and latency requirements could be met, although there may not be sufficient space in a cabinet for this to be possible in practice.
But what is the situation with regard to the current SLU offer in Germany? Alternative operators in practice may have to build their own street cabinet next to that of the incumbent which could make appropriate cabling solutions difficult because the distance is too great between DSLAMs. Where the DSLAMs are separated by an appreciable physical distance, for example, some tens of metres, a timeline of 2015 may be possible, in an optimistic scenario, for cross DSLAM vectoring with DSLAMs from the same vendor.
Computation and communications bandwidth tradeoffs
Another issue is the requirements that would be placed on the vectoring computation engine (VCE), the hardware/software subsystem that calculates the cross talk within a vectoring group and then instructs the modems as to crosstalk mitigation. But if each operator in an SLU scenario has deployed hundreds of lines then a single VCE faces very challenging computational requirements which might mean that full crosstalk cancellation is not possible and so only partial vectoring, without the full benefit for consumers, would occur.
One potential way out of this problem might be to use multiple VCEs, thereby lessening the computational load per VCE, but this would mean more interconnections between the engines to exchange information on disturbers, and therefore worsen the communications complexity issues already mentioned. Bottom line if there are a high number of lines then there might be tradeoffs between issues of computational load and communications bandwidth, requiring suboptimal solutions that cancel only portions of the crosstalk on each line with attendant performance reductions. These issues might be surmountable, however, as incumbents and alternative operators may prefer to deploy only say 96 lines each to begin with, adopting a more pay as you grow approach, and in such a scenario a single VCE might be able to do the computational work required.
Who decides on what?
In a scenario where full cross talk cancellation is not possible, for example, because there are too many lines operators may opt for partial cancellation. But this would create problems as to who would decide on which crosstalkers would be cancelled: one operator may stand to benefit more than the other. Even with full crosstalk cancellation there would still be the question as to who would actually control the vectoring processor performing the calculations to cancel the crosstalk. In addition what would happen if the cross-connect cables to the VCE failed or the VCE itself stopped working, who would be responsible for resolving these issues?
Another issue is that whilst operators may initially deploy DSLAMs from the same vendor there may be differences of opinion about when to activate software upgrades. One more cost conscious operator may decide it wants to skip a software release whereas another may want to activate it, making agreement difficult. Maintenance may also become a factor. One operator may prefer to perform maintenance during normal working hours, therefore reducing its costs, but this would have the consequence of disrupting the other operator’s vectoring performance. As a result getting the operators to agree could prove difficult.
For operators isn’t LLU about having a choice of vendor?
One of the benefits, of course not the only benefit, of an ADSL LLU model is that the alternative operator can choose whichever DSLAM vendor it prefers. Even if cross DSLAM vectoring using the same vendor proved practically possible some of this flexibility of choosing a more or less expensive vendor would be lost because both operators would have to agree to deploy DSLAMs from the same vendor.
Case B: cross DSLAM level vectoring with DSLAMs from different vendors
Standards issues to the fore
For this scenario all the problems that are there with case A are also present, to which we can add a number of other drawbacks. At the moment vendors’ vectoring solutions, in particular the way data is exchanged within a DSLAM or between DSLAMs, are proprietary and not compatible so there is no standard for data exchange between different vendors’ offerings. Not only this but there is also no ongoing work to ensure standardisation between different vendors’ systems. Furthermore there is little incentive for vendors to engage in such activity having spent so much time, effort and intelligence in developing their systems.
Even if such standards work occurred there would still be an issue around compatibility between different vectoring chipsets from vendors such as Lantiq, Ikanos and Broadcom, which each use different vectoring algorithms. To ensure compatibility these algorithms would require modification and this would have the result that the performance achieved would likely be less than that in the single vendor case.
All of this means that even in the most optimistic scenario cross DSLAM vectoring will only be available by 2017 at the earliest, and quite possibly never.
Part 2: is SLU really worth preserving?
Limitations with SLU in Germany
So we have established that cross DSLAM vectoring may, with very significant caveats, allow the coexistence of SLU and vectoring. The second point then is whether removing an SLU obligation is justified. Whilst there is currently some SLU in Germany this is very limited, standing at only around 60,000 lines. I am also not aware of any plans from major LLU players to deploy SLU as the incumbent rolls out vectoring. Of course alternative operators in other countries might take a different view on SLU viability, for example, in Italy Fastweb has announced plans for SLU. But the lack of SLU plans in Germany is significant because using DSM level 2, a technology already standardised and commercially available, to ensure operator coexistence with vectoring and SLU would be particularly useful if 2 operators deployed vectored VDSL in the same area at the same time.
The current SLU offer in Germany is also severely limited because alternative operators may in practice need to build their own cabinet, entailing more CAPEX and OPEX. The incumbent has tended to overbuild its existing passive cabinets and place its DSLAMs in these larger structures. These cabinets may only have the space for one other small DSLAM from an alternative operator and even this may not fit because that DSLAM will require powering. This means that if the alternative operator has to build its own cabinet to house its DSLAM then around 100m of cross-connect cabling may be required to the MDF in the Telekom Deutschland cabinet. This is because it may simply not be possible to build a new cabinet next to the incumbent’s existing cabinet because of, for example, restrictive planning regulations. The resultant increase in the length of cross-connect cabling has the consequence of reducing bandwidths for the SLU player when compared to the incumbent, rendering SLU unattractive.
Germany’s LLU players struggle against cable
To these issues we can add the fact that Germany’s LLU players are under even more pressure than the incumbent from the march of the cable operators. A quick rollout by the incumbent of vectoring and use of virtual unbundling could be the boost they need, and would certainly be less CAPEX intensive for the local loop unbundler than an SLU rollout. The strength of cable also means that LLU players have a lower market share than would otherwise be the case and this is significant because SLU really requires a high concentration of customers within a cabinet area to be viable.
Regional LLU players prefer to invest in FTTH/B
It is true, however, that there are regional LLU players with very high market shares in particular areas. But some players like this, such as Net Cologne, have decided not to pursue SLU and have instead rolled out FTTH/B. The rationale for this is that such players do not want to pay any rental fees to Telekom Deutschland and now that they have rolled out their own FTTH/B infrastructure ironically are arguing for higher LLU fees. Operators like Net Cologne have been able to roll out their own FTTH/B infrastructure because of the stakes that energy companies, whose ducts can be used for fibre rollout, hold in these companies. But Germany’s national LLU players, such as Vodafone, likely lack the local concentration of customers in a particular area to justify an SLU rollout.
Contingent bitstream offers attract interest
Removing the SLU obligation may also not be a problem because the incumbent has developed a contingent model for bitstream over VDSL. The concept of the model, approved by the national regulator and the European Commission, is that those operators that make an upfront commitment to a certain number of VDSL lines can receive a discount. This allows the incumbent to spread some of the risk of its NGA rollout and also allows larger local loop unbundlers to replicate some of the economies of scale they can benefit from when rolling out LLU. Nationwide LLU player 1&1 has already signed a contract with the incumbent to use the contingent model.
Vectoring a way to meet policy targets
The quick rollout of vectoring would also please politicians at both European and national level, helping to meet superfast broadband rollout targets, a point Rene was not shy in making. This does leave the question of whether 100Mbps could be achieved in other ways with fewer regulatory consequences, perhaps with VDSL2 pair bonding? Most households in Germany do have more than one pair, in fact more than two in most cases, in the final section between the distribution point and subscribers’ homes. However, the problem is not all these pairs are connected to the street cabinet so VDSL2 pair bonding from the street cabinet will not always be possible.
Conclusion: Rene is right
Ultimately I believe that banning SLU in Germany is the best, if not perfect, course of action. This is not intended to be a general conclusion applicable to other markets, rather a reflection of the particular German market circumstances. In other markets where alternative operator and incumbent plan to launch vectoring at the same time DSM level 2 could be an attractive solution to preserve infrastructure competition, although subscribers would not receive the full vectoring benefit. In any case any vectoring rollout in Germany should of course not be an excuse for the incumbent to squeeze out LLU players.