MONROE-ROAMING AND MEASUREMENT SETUP¶
In this section, we present the hardware platform we built for roaming measurements, and the manner in which we orchestrate it to collect our data.
本章节中,我们将介绍为漫游测量构建的硬件平台,以及我们协同安排该平台以收集数据的方式
MONROE-Roaming Platform¶
We design and build MONROE-Roaming, a dedicated platform for roaming measurements in Europe. MONROE-Roaming integrates several components that we depict in Fig. 3. The main blocks include measurement nodes distributed in six different EU countries, the backend system, several measurement servers and a scheduler, all of which we detail next. To build the MONROE-Roaming platform we adapted the open source software provided by MONROE [8], an open measurement platform. 2
MONROE-Roaming nodes: Each MONROE-Roaming node is equipped with an APU board from PC Engines with two 3G/4G MC7455 LTE CAT6 miniPCI express modems. Because of the high cost of nodes and subscriptions, and the complexity of the coordination effort required (see § 2.3), we have set up a platform with a total of 12 MONROE-Roaming nodes dedicated for roaming measurements.
MONROE-Roaming backend: Upon completion of each measurement, MONROE-Roaming nodes transfer the measurement results to a central server for further analysis.
Measurement servers: We have deployed one measurement server in each country as measurement responders and also to capture traffic traces.
MONROE-Roaming scheduler: The scheduler allows the user to query for resources, select nodes and launch different tests in the platform simultaneously. We used the open source MONROE scheduler as a basis for the MONROERoaming scheduler. Each test is designed and implemented in a Docker container [21].
我们设计并构建了MONROE-Roaming,一个专用于欧洲漫游测量的平台。MONROE-Roaming集成了多个组件,如图3所示。其主要模块包括:分布于六个不同欧盟国家的测量节点、后端系统、若干测量服务器以及一个调度器,下文将对这些模块进行详细阐述。为构建MONROE-Roaming平台,我们改编了MONROE [8](一个开放测量平台)提供的开源软件
MONROE-Roaming节点: 每个MONROE-Roaming节点均配备一块来自PC Engines的APU主板及两枚3G/4G MC7455 LTE CAT6 miniPCI Express调制解调器。鉴于节点和SIM卡订阅的高昂成本以及所需协调工作的复杂性(参见§2.3),我们搭建了一个共包含12个专用于漫游测量的MONROE-Roaming节点的平台
MONROE-Roaming后端: 每次测量完成后,MONROE-Roaming节点会将测量结果传输至一个中央服务器,以供进一步分析
测量服务器: 我们在每个国家部署了一台测量服务器,作为测量响应器,同时也用于捕获流量轨迹
MONROE-Roaming调度器: 该调度器允许用户查询资源、选择节点,并在平台上同时启动不同的测试。我们以开源的MONROE调度器为基础构建了MONROE-Roaming调度器。每个测试均在Docker容器[21]中设计和实现
Experimental Setup¶
To understand the roaming ecosystem in Europe, we focus on diversity of the MNOs. In other words, we aim to cover a large number of SIMs rather than running measurements from a large number of vantage points. To this end, we deployed two MONROE-Roaming nodes in each of the six European countries to measure a total of 16 MNOs that operate their own network, as illustrated in Fig. 2.
For each MNO, we bought six SIMs that support roaming in Europe and we distributed one SIM in each of the countries we cover. For example, in Germany, we bought six Vodafone DE SIMs that support roaming. We kept one Vodafone DE SIM as the home SIM in the home country (i.e., Germany). Then, we distributed five roaming SIMs from Vodafone DE to the other five countries (i.e, Sweden, Norway, UK, Italy and Spain). Each roaming SIM connects to (or camps on) a local roaming partner (or visited network) native to the visited country. For example, Vodafone DE in Germany is a roaming partner of Telenor SE in Sweden. Therefore, Telenor SE serves Vodafone DE’s customers roaming in Sweden by allowing Vodafone DE users to camp on Telenor SE’s network. For each roaming SIM, we identify the corresponding visited network (e.g., Telenor SE in Sweden for Vodafone DE) and, when available, activate the corresponding native SIM from the visited network (which we hereinafter denote by visited SIM). We illustrate this configuration in the experimental setup in Fig. 3. We also describe the terminology in Table 1.
为深入了解欧洲的漫游生态系统,我们重点关注移动网络运营商(MNO)的多样性。换言之,我们的目标是覆盖大量SIM卡,而非从大量观测点进行测量。为此,我们在六个欧洲国家各部署了两个MONROE-Roaming节点,用以测量总共16个运营自有网络的MNO,如图2所示
针对每个MNO,我们购买了六张支持欧洲漫游的SIM卡,并在我们覆盖的每个国家各分配一张SIM卡。
例如,在德国,我们购买了六张支持漫游的Vodafone DE SIM卡。
- 我们将一张Vodafone DE SIM卡作为归属SIM卡保留在其归属国(即德国)
- 然后,我们将五张来自Vodafone DE的漫游SIM卡分配到其他五个国家(即瑞典、挪威、英国、意大利和西班牙)
- 每张漫游SIM卡会连接到(或驻留于)到访国本地的一个本地漫游合作伙伴(即到访网络)
例如,德国的Vodafone DE是瑞典Telenor SE的漫游合作伙伴。因此,Telenor SE通过允许Vodafone DE用户驻留于Telenor SE的网络,为在瑞典漫游的Vodafone DE客户提供服务。
对于每张漫游SIM卡,我们识别其对应的到访网络(例如,Vodafone DE在瑞典的到访网络是Telenor SE),并在可用时,激活来自该到访网络的相应本地SIM卡(下文我们称之为到访SIM卡)。我们在图3的实验设置中阐释了此配置。我们也在表1中描述了相关术语
Measurement Coordination¶
Each MNO-specific measurement campaign involves 11 SIMs and 6 nodes: (i) one node with the home SIM and (ii) five nodes with both the roaming SIM and the corresponding visited SIM, as illustrated in Fig. 3. This enables us to capture performance metrics for the roaming SIM, but also to compare those with the local performance of the home network and the visited network (when possible).
Before running the set of measurements (see § 3 and § 4), we first need to configure the nodes by activating and deploying the SIMs. For each MNO, we perform the measurements at the same time from all six countries and coordinate the configuration of the experimental setup in two steps:
Home and Roaming User Activation: To measure a MNO, we first insert the SIM into the first SIM slot in each node in all six deployment locations. For the SIM active in its home country, this step triggers the home user activation (by inserting the SIM in the measurement node). For the rest of the nodes, this step triggers the roaming user activation.
Visited User Activation: Once we complete the home and roaming user activation, we check which visited network the roaming SIM uses in each of the five visited countries. Then, we insert the SIM of each partner MNO (when available) into the second slot of each corresponding node.
Using the MONROE-Roaming scheduler, we orchestrate the execution of the measurements so that they run in parallel on all nodes. The measurement coordination effort was a significant part of the process. In each country, at least one person was dedicated to carry out the physical experimental setup configuration for each MNO in a timely manner. Given that we deploy two nodes per country, we could measure two MNOs and (maximum) 22 SIMs in parallel. We coordinated the SIM changes over email. Furthermore, before the change of the next pair of SIMs, we double-checked the measurement results we had collected to ensure correctness and completeness of the dataset. Each round lasted one week, over a total period of more than four months of experiments.
针对每个MNO的特定测量活动涉及11张SIM卡和6个节点:(i)一个带有归属SIM卡的节点,以及(ii)五个同时带有漫游SIM卡和相应到访SIM卡的节点,如图3所示。这使我们能够捕获漫游SIM卡的性能指标,并且(在可能的情况下)将其与归属网络和到访网络的本地性能进行比较
在运行系列测量(参见§3和§4)之前,我们首先需要通过激活和部署SIM卡来配置节点。对于每个MNO,我们从所有六个国家同时进行测量,并分两步协调实验设置的配置:
归属和漫游用户激活: 为测量某个MNO,我们首先在所有六个部署地点的每个节点中,将该MNO的SIM卡插入第一个SIM卡槽。对于在其归属国激活的SIM卡,此步骤将触发归属用户激活(通过将SIM卡插入测量节点实现)。对于其余节点,此步骤将触发漫游用户激活
到访用户激活: 一旦完成归属和漫游用户的激活,我们会检查漫游SIM卡在五个到访国家各使用的是哪个到访网络。然后,我们将每个合作MNO的SIM卡(若可用)插入对应节点的第二个卡槽中
利用MONROE-Roaming调度器,我们协同安排测量的执行,使其在所有节点上并行运行。测量协调工作是整个过程中的重要组成部分。在每个国家,至少有一人专门负责及时为每个MNO进行物理实验设置的配置。鉴于我们在每个国家部署两个节点,我们可以并行测量两个MNO以及(最多)22张SIM卡。我们通过电子邮件协调SIM卡的更换。此外,在更换下一对SIM卡之前,我们会仔细复核已收集的测量结果,以确保数据集的正确性和完整性。每轮测量持续一周,整个实验周期总计超过四个月