A first look at the IP eXchange Ecosystem
Andra Lutu, Byunjin Jun, Fabian Bustamante, Diego Perino, Marcelo Bagnulo, Carlos Gamboa Bontje
AA first look at the IP eXchange Ecosystem
Andra Lutu
Telefonica [email protected]
Byunjin Jun
Northwestern [email protected]
Fabian Bustamante
Northwestern [email protected]
Diego Perino
Telefonica [email protected]
Marcelo Bagnulo
University Carlos III of [email protected]
Carlos Gamboa Bontje
ABSTRACT
The IPX Network interconnects about 800 Mobile NetworkOperators (MNOs) worldwide and a range of other serviceproviders (such as cloud and content providers). It formsthe core that enables global data roaming while supportingemerging applications, from VoLTE and video streaming toIoT verticals. This paper presents the first characterizationof this, so-far opaque, IPX ecosystem and a first-of-its-kindin-depth analysis of ann IPX Provider (IPX-P). The IPX Net-work is a private network formed by a small set of tightlyinterconnected IPX-Ps. We analyze an operational datasetfrom a large IPX-P that includes BGP data as well as statis-tics from signaling. We shed light on the structure of theIPX Network as well as on the temporal, structural and geo-graphic features of the IPX traffic. Our results are a first stepin understanding the IPX Network at its core, key to fullyunderstand the global mobile Internet.
International roaming is an important feature of cellularnetworks, allowing subscribers to use their devices any-where in the world as if at home. Under the IP Packet Ex-change (IPX) model [5, 6], MNOs contract the services ofthird party providers – the IPX Provider (IPX-P) – to of-fer their customers access to mobile services in any foreigncountry. No IPX-P on its own is able to provide connectionson a global basis (e.g., single-handily interworking with allMNOs). Thus, IPX-Ps peer to other IPX-Ps [17] to expandtheir geographical footprint. The resulting
IPX Network , isan isolated network that bypases the public Internet [3], toensure secure, SLA-compliant services, from video streamingand AR/VR to IoT verticals, such as connected cars.Recent years have brought a rapid growth in the numberof participants in the IPX ecosystem and the volume of traf-fic they exchanged. The growing number of internationaltravelers, reaching 1.4 billion in 2018 [1] and the “flat-rating”or elimination of international roaming charges [8, 9, 18] hasled to an exponential growth in roaming traffic, expected
Figure 1:
High level architecture of the IPX Ecosystem. to increase 32-times by 2022. At the same time, users’ QoEexpectations – when using VoIP or posting videos – hasforced content and service providers to peer close to theirusers, wherever they may roam, thus adding to a growinginterconnection ecosystem.Despite its rapid growth and increased importance as thecore of the mobile Internet, the IPX Network and its associ-ated ecosystem (Figure 1) has received little to no attentionby our community, due in part to its intrinsic opacity andseparation from the public Internet.
In this paper, we presentthe first characterization of the IPX ecosystem and a first-of-its-kind, detailed study of a large operational IPX Provider.
We contribute the first topology analysis of the IPX ecosys-tem. We analyze a private BGP routing table snapshot froman operational router that is part of the IPX Network to mapthe interconnection between IPX-Ps and Service Providers(SPs) for the data roaming service (§ 3). We build the ex-haustive list of 29 active IPX-Ps and detail their approach forpeering using three major peering exchange points (AMS-IX Amsterdam, Equinix Ashburn and Equinix Singapore). We capture the breadth and full-mesh peering fabric of theIPX Network for data roaming, which enables inter-workingbetween all ≈
800 MNOs currently active world-wide.We present a first-of-its-kind detailed analysis of a largeoperational IPX-P system, and provide the insider view into IPX is not the same as IXP, though the acronyms are similar. IPX-Ps mayrely on IXPs (such as AMS-IX) for peering with other IPX-Ps.1 a r X i v : . [ c s . N I] J u l he otherwise inaccessible IPX Network. Our study is basedon real-world traffic records for its main service (radio sig-naling for data roaming (§ 4)) over a two-week period inDecember 2019. We showcase the operational IPX-P’s sig-naling and data roaming infrastructures (§ 4.2), with notablepresence in Europe and the Americas. We study signalingtraffic patterns (for different radio access technology) anddata communications from over 22 million mobile devicesroaming in the world. In this section, we provide a detailed description of IPX anduse data roaming, one of the main services offered, to illus-trate the main IPX players and their interactions.
At a high-level, the IPX ecosystem (Figure 1) includes ServiceProviders (SP) and networked IPX-Providers (IPX-Ps). IPX-Psare third-party interconnection providers to SPs (e.g., MNOs,Internet of Things (IoT) providers). IPX-Ps peer with otherIPX-Ps to extend their footprint worldwide.While IP-based, the resulting IPX Network is a privatenetwork, separate from the public Internet, that meshes to-gether the infrastructures of the IPX-Ps. It guarantees trafficseparation between IPX services from the rest of the Internet.The IPX Network enables the transport of global roamingdata between networks, with interoperability of differentimplementations and standards.SPs require a single connection and agreement with oneIPX-P in order to connect to the IPX Network, and intercon-nect with partner SPs world-wide. For instance, to enabledata roaming, two MNOs must each have an agreement withan IPX-P in order to interconnect. For redundancy, a SP couldestablish connections to more than one IPX-P. Dependingon the footprint of the IPX-P’s infrastructure, SPs can useone or more Point of Presences (PoPs) of the IPX-P.
To establish roaming, roaming partner MNOs must have afunctioning commercial agreement, implement their roam-ing technical solutions, establish inter-working and deploytheir billing function. We use data roaming to illustrate themain IPX ecosystem players and their interactions.In terms of business agreement solutions, the legacy op-tion for MNOs is a standard bilateral agreement where thetwo parties involved define terms and conditions of their co-operation. These bilateral roaming agreements for roamingand inter-working are costly and generally of lower value Although direct interconnection between SPs through leased lines orVirtual Private Network (VPN) is possible, it is outside the scope of ouranalysis. today, something that has served as additional motivationfor MNOs to adopt the IPX model.Under the IPX model, operators connect to an IPX-P togain access to many roaming partners world-wide, external-izing the inter-working establishment to the IPX-P offeringthe service. IPX-Ps are then peering with each-other to ex-pand their international footprint through the IPX Network.This IPX hubbing solution does not preclude the existence ofbilateral agreements between MNOs, which can be viewedas a complementary roaming model.Once a commercial agreement has been created, the IPX-Psets up the technical roaming solution, including coordina-tion over the signaling platform, and establishes the IPX-P in-terconnectivity. After MNOs establish roaming inter-working,they deploy the billing service, which is key to recoveringroaming revenue. The roaming partners must each recordthe activity of roaming users in a given Visited Mobile Net-work Operator (VMNO). Then, by exchanging and compar-ing these records, the VMNO can claim revenue from thepartner Home Mobile Network Operator (HMNO).When a mobile device is at home, the subscriber’s trafficwill take a short path inside the network to reach a suitablePacket Data Network Gateway (PGW) to the Internet. Wheninter-working exists between two MNOs, there are severalnetwork configurations the IPX Network supports to enableroaming.
The IPX-P’s main function is to build the commu-nication tunnel between the Serving Gateway (SGW) and thePGW, enabling traffic to flow to and from the roaming mobiledevice.
The traffic of a roaming mobile device is directed toan egress PGW whose location depends on the roaming con-figuration. Different configurations for roaming over the IPXNetwork are available – home-routed roaming (HR), localbreakout (LBO) and IPX hub breakout (IHBO). Prior workfound that the default roaming configuration majority MNOscurrently use in Europe is the HR roaming [14]. In the case ofHR, the mobile device receives the IP address from its homeMNO and the roaming traffic is then routed over a tunnelbetween the SGW in the VMNO and a PGW in the HMNO.
The core players of the IPX ecosystem are IPX-Ps and SPs.IPX-Ps provide the interconnection between SPs directlythrough their network or via peering with other IPX-Ps (seeFigure 1). The resulting ecosystem has a layered topology:a core of tightly interconnected IPX-Ps in a full mesh (theIPX Network) and the edge of diverse SPs that interconnectthrough the IPX Network, either through a single IPX-P ormulti-homed through multiple IPX-Ps. For this, SPs may usetheir own access network (e.g., fixed and mobile networkoperators) or use a local provider to connect to the PoP ofthe IPX-P (e.g., Application Service Providers (ASP)). SPs riginate and/or terminate traffic for one or several services;they do not transport traffic. IPX-Ps establish interconnection either through private bi-lateral interconnections or through an Internet ExchangePoint (IXP) (Figure 1). The benefits of peering are well knownamong Internet Service Providers (ISPs) and Content Deliv-ery Networks (CDNs), particularly when it comes to publicpeering via an IXP [2]. In recent years, we have seen moreefforts to expand the peering culture to the mobile ecosys-tem [17], by establishing new mobile peering infrastructureworld-wide. This is (slowly) propagating to the mobile in-dustry, and the interconnection of IPX-Ps via mobile peeringat specific IXPs, which offer this service, is becoming com-monplace. Currently, the two major IXPs offering the mobilepeering service are AMS-IX and Equinix, with three locationsoverall (Amsterdam, Ashburn and Singapore).IPX-Ps dynamically exchange routing information withother IPX-Ps using the BGP routing protocol. According toGSMA reccomendations [6], IPX-Ps should not act as a tran-sit IPX-P (i.e., there can only be a maximum of two IPX-Psbetween two partner SPs). Therefore, when an IPX-P has acustomer SP who requires a connection to a customer SP ofanother IPX-P, the two IPX-Ps should peer, either through(direct) private peering or peering points. For example, inFigure 1, the path between MNO1 and MNO2 should only tra-verse two IPx-Ps (i.e., MNO1 -> IPX-P1 -> IPX-P2 -> MNO2).Also, IPX-P3 should never transit traffic for neither of its twopeers, IPX-P1 and IPX-P2. To ensure this, network routesIPX-P3 receives either over private peering or over a peeringpoint should not be re-advertised to other IPX-P peeringpartners. These recommendations result in a tightly inter-connected IPX Network, with a theoretical diameter of twoentities between any pair of SPs. (a) IPv4 prefix length. (b) AS-Path length.
Figure 2:
Analysis of the prefixes within the routing tablesnapshot from an operational IPX-P: (a) distribution on pre-fix length; (b) distribution on AS-Path length (without AS-Path prepending).
Given the opaque nature of the IPX ecosystem, we cannotcapture its characteristics from the public Internet, not byusing public Internet routing data nor with active end-to-endmeasurements (e.g., traceroute). In order to shed light on theinterconnection fabric between IPX-Ps and SPs in the IPXecosystem, we analyze three different datasets: (i) routingdataset: a private BGP routing table snapshot for data roam-ing that one of the largest operational IPX-Ps provided us;(ii) peering dataset: the list of AMS-IX members that connectto the mobile peering services for data roaming service only,together with the full internal list of IPX-P peers from theoperational IPX-P providing us the routing dataset; (iii) sur-vey dataset: market surveys and reports [13, 16] from thirdparties. In the routing data, we capture one snapshot of therouting table on the 30th of January 2020, which providesone view of the IPX-P’s relationships for the data roamingservice. We note that this information, though descriptiveof the ecosystem, might be incomplete (i.e., there might beinformation not included here, but present in snapshots atother vantage points).Specifically, the IPX-P routing table snapshot includesreachability information for a total of 10,418 IPv4 prefixesadvertised by 59 different entities (neighbors), which includethe peer IPX-Ps and the customer MNOs of the IPX-P weanalyze. By checking the originating AS of the prefixes in therouting dataset, we find that within the IPX ecosystem thereare a total of 824 different service providers for data roaming(i.e., MNOs, MVNOs or M2M platforms). This number is con-sistent with the total number of MNOs active world-widethat register with the GSMA. These are public IPv4 prefixesthat IPX-Ps do not announce in the global BGP routing ta-bles. Hence, they are not reachable from the public Internet.From Figure 2a, we note that the median prefix length inthe routing dataset is /29, and there is a large number of /32prefixes that MNOs originate. These likely represent specificelements (e.g., Home Location Registry (HLR) or MobilityManagement Entity (MME)) within the MNO infrastructure,which are involved in procedures for data roaming.
IPX Network:
We merge and corroborate the informa-tion we extract from the above-mentioned datasets to builda list of IPX-Ps that currently form the IPX Network. InAnnex B, we detail the full list of these providers and themethodology we found we compile this list. We specificallynote that, in light of the growing popularity and worldwidefootprint of IPX services, many heavyweight telecoms areparticipating in the IPX environment, leveraging their un-derlying extensive infrastructure.We verify that these IPX-Ps appear in the routing datasetas active peers of the IPX-P providing us the routing table napshot. The total number of entities advertising reacha-bility information to the IPX-P is equal to 59 different ASes.Out of these, we separate the ones that advertise prefixeswith an AS-Path length longer than one (i.e., they advertisetheir SPs). We check the overlap with the list of IPX-Ps webuilt, and find a set of 23 different ASes. (a) Customer SPs per IPX-Ps. (b)
IPX Providers per SP.
Figure 3:
Interconnection between IPX-Ps and SPs: (a) dis-tribution of number of customers per IPX-P; (b) distributionof number of IPX providers per SPs.
IPX Interconnections:
The set of 29 IPX-Ps we foundmust provide interconnection services to the over 800+ MNOscurrently active world-wide [4]. The number of PoPs indi-cates the number of locations where an IPX-P can cross-connect with SPs, giving insight into its world-wide geo-graphical footprint. For example, large players such as TeliaSonera, TATA Communications, Orange, Vodafone, TelecomItalia, Telefonica or Telekom Austria offer an infrastructurewith more than 100 PoPs world-wide each. The average num-ber of PoPs among the 18 IPXs publicly disclosing this infor-mation is 116, hinting the breadth of the system we capture.We further use the routing dataset in order to charac-terize the interconnection between IPX-Ps and SPs in theecosystem. For the IPX-Ps present in the routing dataset, weverify the different number of SPs for which they advertisereachability information (Figure 3a).
We find that four major players within the IPX Network(namely, Syniverse, BICS, Orange and Comfone) together giveservices to a total of more than 600 MNOs (out of the total 800).
Inversely, we also verify the popularity of multi-homingamong SPs. In other words, we quantify the number of IPX-Ps that advertise reachability information for the same SP.Figure 3b shows that 80% of SPs are single-homed (i.e., theyonly connect to one IPX-P), while for the rest we observe upto seven different IPX-Ps. In particular, we note that multipleM2M platform providers use at least four different IPX-Ps,which is intuitive due to their reliance on roaming.Finally, in order to characterize the interconnection be-tween IPX-Ps, we analyze the AS-Path of the prefixes in therouting dataset. To comply with the recommendation of afully connected IPX Network, there can be no more than two The remaining 36 ASes observed in the BGP routing table are SPs that arecustomers of the analyzed IPX-P. different IPX-Ps involved in the communication betweentwo different SPs. Figure 2b shows the distribution of pre-fixes on AS-Path length. Note that we eliminated AS-Pathprepending (used by IPX-Ps and SPs for traffic engineering).
We show that, indeed, majority prefixes advertised by peerIPX-Ps have a path length equal to two, confirming the tightinterconnection required in the IPX Network.
The paths longerthan two ASes represent the result of MNOs working to-gether with their parent networks (e.g., national MNOs con-necting to their parent carrier) or third party network providers(e.g., for MVNOs, M2M platforms) to achieve a broader ge-ographical footprint. Thus, sibling ASes that belong to thesame organization appear in the same AS-Path.We also observe that ASes in the IPX network use standardtechniques for traffic engineering. In particular, we observea heavy use of AS prepending. AS path prepending makesa route less preferred to receive traffic by making the ASpath length for the route artificially long by repeating ASesin the AS path attribute. We observe 1,331 routes from 131different origin ASes where prepending was used. We alsoobserve extensive use of Multi Exit Discriminator (MED),a BGP attribute that serves to express preference betweendifferent links between two ASes. We observed 1,583 prefixesthat contained the MED attribute.
We continue our analysis of the IPX ecosystem and zoom intothe operations of one of the largest IPX-Ps in the ecosystem.The IPX-P we dissect is a Tier-1 Internet Service Provider op-erating one of the largest backbone networks world-wide. Aspart of its interconnectivity products, the carrier operates anIPX infrastructure that runs on top of its vast MultiprotocolLabel Switching (MPLS) network. The IPX-P infrastructureintegrates more than 100 PoPs in 40+ countries with a particu-larly strong presence in America and Europe.
In terms of network connectivity, the IPX-P offers twotypes of interfaces, namely the IPX Access for clients (ser-vice providers) and the IPX Exchange for peering with otherIPX-Ps. The main mobile peering points the IPX-P uses arethose in Singapore, Ashburn and Amsterdam. By peeringwith other large Tier-1 carriers, the IPX-P extends its foot-print world-wide to geographic regions where it does notown infrastructure (§ 3.2). The IPX-P serves clients in multi-ple countries in Europe (Germany, Spain, UK) and the Ameri-cas (including US, Mexico, Brazil, Argentina, Colombia, Peru,Chile, and Ecuador). An IPX-P requires access to an underlying backbone network. The IPX-Pmay own its own MPLS network or alternatively, it might lease capacityon MPLS networks on which they deployed the infrastructure needed todeliver and manage inter-operable cross-network services. igure 4: High level architecture of the IPX-P’s signalingplatform.
We now describe the IPX-P infrastructure, the monitoringmethodology and the datasets we collected to characterizeits operational system and services. We monitor the IPX-P in-frastructure corresponding to the two main services – SCCPGlobal Signaling, LTE Diameter Exchange – for two weeks,from December 1st to December 14th 2019.
Overall, the total number of IMSIs we capture is of morethan 22M active daily in 2G/3G and more than 2M active dailyin 4G/LTE. . They correspond to 215 home countries and 210visited countries. The IPX-P’s customers are active within 19countries and include MNOs, IoT/M2M connectivity providersand cloud service providers.SCCP Global Signaling:
This service provides access to theIPX-P’s SS7 signaling network, satisfying the 2G/3G inter-connection needs for international roaming of MNOs. TheSCCP Signaling network of this particular IPX-P has a redun-dant configuration with four international Signaling TransferPoints (STPs) located in North America (Miami, Puerto Rico)and Europe (Frankfurt, Madrid) as depicted in Figure 4.To capture clients’ activity across this signaling platform,we monitor the Mobile Application Protocol (MAP) protocol,which supports end-user mobility and is used by devices tocommunicate with the major network elements, includingthe HLR, Visiting Location Registry (VLR) or the MobileSwitching Center (MSC). We collect traffic correspondingto the following procedures of each device belonging toone of the IPX-P’s clients (outbound roaming) or to foreigndevices that connect to the network of one of the IPX-P’sclients (inbound roaming): i) location management (updatelocation, update GPRS location, cancel location, purge mobiledevice); ii) authentication and security (send authenticationinformation); iii) fault recovery.
LTE Diameter Exchange:
This service provides the Diam-eter signaling capabilities necessary to enable 4G roamingfor customers. The infrastructure of this particular IPX-P Note that there might be an overlap between these two sets. However, weaim to show here the load on the two different signaling infrastructures includes four Diameter Routing Agents (DRAs) meant toforward Diameter messages and simplify interworking be-tween different network elements. The LTE Diameter serviceintegrates value added services, including Welcome SMS,Steering of Roaming or Sponsored Roaming.To monitor the activity of the IPX-P’s customers acrossthis platform, we monitor traffic across the geo-redundantsignaling network with four DRAs located two in Europe(Frankfurt, Madrid) and two in North America (Miami, BocaRaton). The infrastructure is similar to the one in Fig. 4.
Figure 5 shows signaling activity of roaming mobile sub-scribers during the observation period of December 2019. Welook at both MAP and Diameter signaling procedures. MAPis the most important application protocol in the SignalingSystem No. 7 (SS7) stack, and handles the roamers’ mobil-ity between countries. Although this is still the most usedprotocol for mobile interconnection application messages,the more recent Diameter [12] signaling protocol has beengrowing with the adoption of LTE. Figure 5a focuses on thetotal signaling traffic and on the number of different mobilesubscriber devices that generate this traffic.
We find that the number of devices using 2G/3G infrastruc-ture (extracted from MAP traffic) is an order of magnitudehigher than those using 4G infrastructure (based on the Di-ameter traffic). The volume of signaling traffic in the SCCPinfrastructure is, correspondingly, more significant in terms oftotal volume than in the Diameter infrastructure.
We also notethe typical daily and weekly traffic patterns on mobile sub-scribers’ activities. For instance, December 1st was a Sunday,showing the expected decreasing trend in signaling trafficactivity which can be seen again the following weekend(December 7-8th).Each record in this dataset represents a signaling pro-cedure that a network element triggers, corresponding todifferent standard routines. For instance, from the MAP in-terface we capture mobility management routines, includinglocation management and authentication. Figure 5b showsthe time series of signaling traffic broken down by type ofsignaling procedure, including Update Location (UL), CancelLocation (CL) and Send Authentication Information (SAI)messages. The latter, SAI, represents the highest fraction ofMAP signaling traffic. Indeed, according to the GSM stan-dard definition, the Serving GPRS Support Node (SGSN) inthe visited network triggers the authentication of subscriberprocedure upon IMSI attach, location update or before start-ing data communication, thus explaining the large volumeof SAI messages.Figure 5c shows the average number of records per IMSIcalculated over all the IMSIs we observe in each one-hour a) Signaling traffic volume and number of devices in eachplatform (SCCP Signaling and Diameter Signaling). (b)
SCCP signaling traffic time series; breakdown per type ofsignalling procedure. (c)
Average and standard deviation of the number of SCCPmessages and Diameter messages per IMSI per hour.
Figure 5:
Signaling traffic time series for the observation period of December 2019.(a)
SCCP global signaling. (b)
Diameter signaling.
Figure 6:
Signaling traffic load (total number of records) perinfrastructure element per day for SCCP platform and Di-ameter. The boxplots capture the first two weeks in Decem-ber 2019. interval (continuous line) during the observation period, aswell as the standard deviation of the number of records perIMSI calculated over all the IMSIs active in the same one hourinterval (shaded area). We observe both the MAP proceduresfor 2G/3G (red color) and the similar Diameter procedures for4G/LTE (green color). While Diameter and MAP are differentprotocols, the underlying functional requirements (e.g., au-thenticating the user to set up a data communication) havemany similarities in terms of the messages used for Diameterand the SS7 MAP protocol implementation. We note that theload in terms of average signaling records per IMSI is in thesame order of magnitude (the continuous lines on the plot),regardless of the infrastructure the devices use; yet, thereare significantly more messages for MAP, as Diameter is amore efficient protocol than MAP [12, 19].We further investigate the traffic load on the signalinginfrastructure, both for the SCCP (with STPs in Frankfurt,Madrid, Miami and Puerto Rico) and the Diameter signalinginfrastructure (using DRAs in Frankfurt, Madrid, Miami andBoca Raton). Figure 6 shows the aggregated signaling traffic(number of messages) per infrastructure point per day overthe same observation period. We note that the redundantdeployment of infrastructure points in each geographicalarea allows the IPX-P to load-balance the signaling trafficacross them (e.g., for 4G between Frankfurt and Madrid inEurope and between Miami and Boca Raton in America).Signaling traffic in Europe through the signaling points in Frankfurt and Madrid is considerably higher than that flow-ing through the signaling points in America (Miami and BocaRaton/Puerto Rico). This is true for both the SCCP infras-tructure (Fig. 6a) and the Diameter infrastructure (Fig. 6b).When comparing the two signaling services (i.e., SSCP andDiameter), we also note that the traffic volume in the SCCPsignaling infrastructure is approximately twice larger thanthe traffic volume flowing through the Diameter signaling in-frastructure. This proves the continued popularity of 2G/3Gservices in the region served by the IPX-P.
The growing demand for global, mobile broadband accessand a shift to all-IP-based services (from broadband last mileto VoIP) have brought new impetus to the old idea of the IPX,first proposed by the GSMA in 2007 to replace the traditional,bilateral-agreement model for international roaming [6]. De-spite the continuous technical development by IPX-Ps andthe related parties [7, 20, 21] there has been few academicworks on the topic. Takaaki [15] provides an early surveyof IPX and its technical requirements, but there have beenno in-depth analysis of IPX since due, in part, to its closednature. Our work presents the first in-depth analysis of theIPX Network and the associated ecosystem.
In this paper, we provided both a qualitative and quantitativedescription of the IPX ecosystem using information collectedfrom the one of the largest IPX providers. We believe thatunderstanding the IPX network is cornerstone to understandand evolve the mobile Internet, and that it will become morerelevant as new services emerge. For example, relying onIPX services, novel technologies such as eSIM that allowremote provisioning of mobile devices permit Global MobileNetwork providers (e.g., Truphone) to emerge and offer novelconnectivity options to end-users.
REFERENCES [1] 2019.
International Tourism Highlights, 2019 Edition . World TourismOrganization, Madrid, Spain. https://doi.org/10.18111/9789284421152
2] Bernhard Ager, Nikolaos Chatzis, Anja Feldmann, Nadi Sarrar, SteveUhlig, and Walter Willinger. 2012. Anatomy of a large European IXP.In
Proceedings of the ACM SIGCOMM 2012 conference on Applications,technologies, architectures, and protocols for computer communication .163–174.[3] Mostafa Ammar. 2018. Ex Uno Pluria: The Service-InfrastructureCycle, Ossification, and the Fragmentation of the Internet.
SIGCOMMComput. Commun. Rev.
Proceedings of the 24th Annual International Conference onMobile Computing and Networking . 179–189.[15] Takaaki Moriya. 2013. Survey of IPX (IP eXchange) as an EmergingInternational Interconnection between Telecommunication Networks.
IEICE Transactions on Communications ETHICAL CONSIDERATIONS
Data collection and retention at network middle-boxes arein accordance with the terms and conditions of the IPX-Pand the local regulations, and only with the specific purposeof providing and managing the IPX service. The terms alsoinclude data processing for monitoring and reporting as al-lowed usages of collected data. Data processing only extractsaggregated information and we do we not have access to anypersonally identifiable information. We nevertheless con-sulted with the Institutional Review Board (IRB) office at ourinstitution who confirmed that no IRB review was necessaryas the study relies on the analysis of de-identified data.
B IPX PROVIDERS
We compiled a list of 29 IPX-Ps that, at the time of writing,interconnect to form the IPX Network.
Table 1 lists the IPX-Pswe identified as active in the IPX Network, their (public)peering policy, number of PoPs and the type of AutonomousSystem (AS) number (private/public). We built the list ofIPX-Ps and the information we show in the table by manu-ally exploring the peering dataset. Specifically, at AMS-IXwe found 27 customers for the mobile peering service withIPX/GRX tags [11]. In addition, we found 24 of the total 27IPX-Ps peering at AMS-IX also present at Equinix IXPs. Ac-cording to GSMA, the majority of IPX-Ps connect to thesetwo IXPs [6]. Additionally, we checked the top 10 largestIXPs and some global IXPs and found no additional IXPsoffering the mobile peering service. We complete this listwith two additional IPX-Ps we observe in the internal listof peers from the operational IPX-P, thus bringing the totalnumber of players in the IPX Network to 29, which we listin Table 1. We confirm this list and the IPX-Ps’ identitiesby checking several commercial market surveys and reportsabout IPX-Ps from diverse parties [10, 13, 16].We also find that among the set of currently active IPX-Ps, there are several which focus on interconnecting SPswithin a specific region (e.g., Telin Indonesia, SAP). We do not include these in our analysis, but instead focus on thelist we show in Table 1.The number of PoPs in Table 1 indicates the number oflocations where an IPX-P can cross-connect with SPs, givinginsight into its world-wide geographical footprint. The aver-age number of PoPs among the 18 IPXs publicly disclosingthis information is 116.
Table 1: List of active IPX Providers.
IPX-P Name BGP Peering Policy PoPs