跳转至

Performance Analysis

TL; DR

  • 网络差异：各运营商(MNO)因频谱、基站密度不同，覆盖(FDP)和满意度(FSP)在各省表现各异，MNO2通常领先
  • 漫游效果：国家漫游普遍提升FDP（最高可降至0）和FSP（提升0.05-0.35），资源较少的MNO3受益最明显。但FSP仍难完美，需扩容
  • 优化策略：仅共享4G接近全面漫游效果；城市漫游优于农村。共享资源越多（频谱/基站），效果越好
  • 用户选择：漫游时，多数用户（MNO3高达82%）会连接到其他运营商网络，因为MNO3实在太差了

Original

With these insights on the infrastructure of each MNO, now, let us discuss the coverage and capacity performance of these MNOs. Fig. 4 shows the FSP and FDP per province per MNO and for all MNOs together, representing the case where all operators can use each other’s network as in national roaming. Looking closer at Fig. 4a, we have the following three observations. First, in various regions, some operators fail to provide sufficient SINR to their customers, resulting in an FDP as high as 0.11 in Friesland or around 0.09 in Zeeland. Especially in these regions, national roaming provides its benefits as reflected in a significant improvement in FDP with an achieved FDP of zero. Moreover, MNO 2 consistently achieves a higher performance in almost all regions compared to MNO 1 , whose performance is in turn significantly better than MNO 3 . We attribute this superior performance of MNO2 to its higher spectrum resources (204 MHz vs. 175 MHz and 140 MHz) as we have not observed a significant difference in their BS density in Fig. 3a. The performance gap between MNO 3 and other MNOs can emerge due to the lower spectrum resources and lower BS density.When it comes to FSP, Fig. 4b again shows a superior performance of MNO 2 . However, the achieved FSP varies between 0.73 and 0.89, indicating a need for performance improvement. Comparing MNO 1 and MNO 3 , generally speaking, MNO 1 outperforms MNO 3 in terms of FSP, except in Friesland, which is one of the worst-performing regions in terms of both FSP and FDP. These low-FSP regions could be considered as initial places for investment to ensure higher user satisfaction. Also, MNO 1 and MNO 3 provide a more varying FSP across provinces compared to MNO 2 . For example, the achieved FSP for MNO 2 ranges from 0.70 − 0.95, while for MNO 1 and MNO 3 , the achieved FSP lies in [0.54, 0.89] and [0.56, 0.80], respectively. This large range in satisfaction might emerge as a result of different deployment strategies.

Second, letting users connect to every BS regardless of their MNO improves FSP consistently as expected. This improvement is, however, more significant in some regions such as Friesland or Overijssel. Third, national roaming improves FSP by around 0.05−0.35 in absolute terms (Fig. 4c). However, in contrast to FDP, it does not yet suffice to meet all rate requirements as reflected by FSP always being below 1.0. This can be considered as an indication of the need for infrastructure expansion or for more advanced schemes to provide higher throughput, e.g., expanding to higher spectrum bands with abundant bandwidth. When it comes to gains experienced by each MNO in the case of national roaming, Fig. 4c plots the performance gain in terms of FDP and FSP observed by each MNO. In line with our earlier observations, MNO 3 benefits the most from national roaming, followed by MNO 1 with a slight difference over MNO 2 . Please note that despite ∆FDP being very narrow for MNO 1 and MNO 2 , the resulting FSP gain is still remarkable also for these operators. Note that all ∆FSP and ∆FDP values in Fig. 4 are positive implying that national roaming does not lead to performance degradation and even the best-performing MNO can benefit from it, albeit less significantly compared to other MNOs with less-dense deployment. As a nation-wide implementation of national roaming requiring collaboration of all MNOs and technologies might be hard to realize, we also investigated roaming in a limited way to provide insights to the MNOs and when and where they could benefit most from national roaming. Such an analysis could also help national telecommunication regulatory bodies to develop policies enforcing national roaming for serving the underserved areas or regions benefiting the most from this mode of operation. We focused on the following questions 8 : (i) which cellular technology, i.e., 3G, 4G or 5G, should be prioritized for roaming if MNOs are eager to develop such roaming agreements? and (ii) in which areas should the MNOs consider roaming agreements, i.e., rural areas where the infrastructure is less dense or in urban areas where the population density thereby the traffic load is higher?

Fig. 5 depicts the resulting FDP and FSP when operators only share BSs of a certain technology to serve each other’s users. In this case, users access to the network of their own operator for the remaining technologies for which national roaming is not implemented. Moreover, we compare this to the scenario with sharing all technologies (‘3G, 4G and 5G’). This figure shows that only sharing the 4G LTE technology performs similar to the full national roaming approach, and even slightly better in terms of FSP for MNO 1 and MNO 3 . Only sharing 5G NR or 3G UMTS performs significantly worse, both in terms of FDP and FSP. We speculate that the superiority of 4G LTE can be explained by the large spectrum resources (Table I), as this will result in the least interference. For MNO 1 , while 36% of the cell sites are 4G LTE sites, the spectrum for these areas accounts for 60% of the spectrum used by this operator. For MNO 2 , 40% of the cell sites are 4G operating on 88% of the spectrum resources owned by MNO 2 . Lastly, for MNO 3 , 4G cell sites correspond to 54% of the cell sites and the spectrum allocated to these cell sites is 86%.For addressing (ii), we considered roaming only in UMa or RMa cells as defined in [16] as these two cells correspond to urban areas and rural areas, respectively. Fig. 6 illustrates FDP and FSP when sharing is applied only in certain areas.

This figure shows that sharing infrastructure in the UMa areas outperforms sharing in RMa areas significantly. We attribute this result to the number of resources: for all operators, there are 5-6 times more BSs classified as UMa compared to RMa. Moreover, the population is concentrated in urban areas resulting in higher traffic load. A clear recommendation on whether MNOs should prioritize sharing a certain technology or sharing in certain areas is not straightforward as our results suggest that sharing more resources (either spectrum or BS locations) are expected to result in better FDP and FSP.User association: Fig. 7 shows the percentage of users that is being roamed to another MNO in case of national roaming. This figure shows that under national roaming, user association changes significantly and at most half of the users of an MNO will connect to that MNO, while the other users connect to the BSs of another MNO. For MNO 3 , this percentage is even lower: only 17.9 % of their users connect to the BSs of their own MNO. These results are in line with the results shown in Fig. 4: since MNO 3 provides less coverage to their users, they benefit most from national roaming and therefore many users will use the network of MNO 1 and MNO 2 .Takeaway — Our analysis shows that FDP and FSP vary across MNOs and geographic regions. National roaming can consistently offer benefits; up to 13% improvement in FDP and up to 55% in FSP. However, the observed benefits vary across MNOs, technologies and regions. Thus, for internet equity, national roaming can be a solution in those regions not meeting a desired level of FDP and FSP to achieve a certain target level. Moreover, the simulations show that only sharing a certain technology or in a certain area results in better FDP and FSP, although amount of resources (available spectrum or number of BSs) in this case is the most important factor: sharing more resources generally results in better FDP and FSP. Similarly, MNOs can consider these areas for their own network expansion with newer technologies, including mid-band 5G frequency usage for enhancing FDP and high-band 5G frequencies for enhancing FSP.