Internet-Draft C. Tocci P. Conforto Expires: February 2001 G. Losquadro Alenia Spazio ACCORD mobile multi-segment system: IP over ATM performance and QoS trial results evaluation STATUS OF THIS DOCUMENT This document is an Internet-Draft and is in full conformance with all provisions of section 10 of RFC 2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress". The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. ABSTRACT This document reports upon the ACCORD Trial activity which aims at investigating support of TCP/IP over mobile multi-segment system. IP based applications and IP over ATM measurements have been executed in order to obtain significant information about intersegment service mobility. 1 Introduction The ACCORD project, developed in the Advanced Communications Technology and Services (ACTS) framework funded by the European Union, is an integration project of four different Access Systems and a Core Network, all coming from different ACTS projects: SECOMS (Broadband Satellite Access System), MEDIAN ATM (Broadband W-LAN Access System) SAMBA (Cellular W-ATM Access System) and EXODUS (Enhanced-DECT Access System and Mobile Broadband Core Network). Expires: February 2001 [Page 1] Internet-Draft IP over ATM August 2000 The main goal of the ACCORD project is to investigate and validate the possibility to create a Global Mobile Broadband (GMB) communication system, consisting of multiple (satellite/terrestrial) system components (or segments), able to support a wide range of multimedia services (ATM native and IP oriented applications). The project has targeted its objectives through both a theoretical and experimental activities, aiming at designing a Target Network and executing field measurements on a trial platform, respectively. The scope of ACCORD Trial is to study and measure the multi-coverage terrestrial and satellite environment. To this end appropriate measurements, at various level (Physical, ATM, TCP/IP) have been executed and the performance of the support of the ATM network to the TCP/IP applications in the ACCORD environment has been evaluated. 2 Overall Trial Strategy definition Given that the ACCORD Trial aims at gathering as much as possible significant information about the parameters involved in the InterSegment HandOver (ISHO) between a variety of radio terrestrial and the satellite radio access networks, the maximum attention has been put on selecting combinations of coverage in which the satellite access is available. Therefore the selected trial scenarios are: SECOMS/DECT trial which involves InterSegment HandOver (ISHO)/Dual Mode Terminal Roaming (DMTR) study in the context of transitions from Rural to Urban environment, from Outdoor to Dual coverage and from Broadband to Narrowband and, of course, viceversa; SECOMS/SAMBA trial which involves ISHO/DMTR study in the context of transitions from Rural to Urban environment and from Broadband to Wideband; DECT/MEDIAN trial which aims at collecting data to characterise typical transient coverage situations in the context of indoor ISHO and transition from Narrowband to Wideband. The DECT/MEDIAN trial has been selected instead of the SECOMS/MEDIAN trial which, as there is not overlapping of indoor MEDIAN and satellite SECOMS coverage, does not provide significant results. The SECOMS/MEDIAN trial can be obtained from the logical combination of the SECOMS/DECT and DECT/MEDIAN trials relying on the intermediate "bridge" coverage provided by DECT. The three trials SECOMS/DECT, SECOMS/SAMBA and DECT/MEDIAN, consist in running two parallel connections with the simultaneous execution of TCP/IP measurements on both the radio access systems involved in the trial. Expires: February 2001 [Page 2] Internet-Draft IP over ATM August 2000 In both SECOMS/DECT and SECOMS/SAMBA trials the Dual Mode Terminal is in movement along a predefined path, chosen in such a way to provide characteristics ISHO and Dual Mode Terminal conditions. The goal is to take an adequate number of measurements in significant conditions and for validation purposes. In particular the measurements have been executed at low speed (pedestrian) since the environment where the measurement execution takes place is the near building area. This environment has been chosen because it is the more interesting as for the satellite/terrestrial complementary access networks. 3 TCP/IP level measurement approach The goal of the trial measurement campaign execution is to find out what applications could be provided over a mobile multi-access system. With reference to the OSI protocol stack the point for measurement is on the User-Plane at application level. This is the basic step to be done to identify the features of the applications that can run over the system. The followed approach foresees that the measurements are executed with a spatial and temporal correlation. The spatial correlation (measurements executed on same points/path) is intrinsically provided by the Dual Mode Terminal emulation (co-location of two mobile terminals) and the temporal correlation is guaranteed by executing the two access system measurements at same time. This approach not only studies what of the currently available applications can be used over a Multi-mode/Multi-access platform, but also allows to start considering applications specifically designed for such platform. These applications are capable to run over multiple user-plane belonging to different networks and take full advantage as for user QoS both for bandwidth, mobility and cost. 4 Measurement Tests executed over the ACCORD platform In order to select a significant set of measurements to be done during the ACCORD trials and in order to set up and tune a number of tools to be used, preliminary tests, where only fixed systems are involved in communication chain and no radio access are included, have been executed. The performance parameters selected for trial measurements are: Throughput and Round Trip Delay at TCP/UDP/ICMP level, the tools used are the netperf program and the ping program respectively. 5 SECOMS/DECT joint-trial execution SECOMS/DECT Trial aims at validating a mobile environment scenario where an open area / urban area (near building) coverage is assured by a dual-mode access network consisting of the Broadband Satellite and of the DECT segments. Expires: February 2001 [Page 3] Internet-Draft IP over ATM August 2000 The equipment for the SECOMS/DECT Trial consists of a dual mode ACCORD Mobile Terminal (DECT and SECOMS), the corresponding Base Stations, the ACCORD Core Network (ATM switch) and two EXODUS Fixed Broadband Terminals (FBTs). The ACCORD Mobile Terminal consists of two co-located subsystems: a SECOMS Terminal and a DECT Portable Part. The Base Stations for SECOMS and DECT are connected via ATM to the ACCORD Core Network. The SECOMS/DECT joint-trial execution mainly aims at evaluating the impact that different radio coverage environments have on ATM connection, IP and TCP/UDP connection throughput and transfer delay. The ACCORD mobile Terminal is mounted on a van and two simultaneous calls towards the two different EXODUS FBTs are established: one through the SECOMS access and the other one through the DECT access. The ATM switch is properly set in order to support two ATM permanent cross-connections both identified by the couple VPI/VCI=10/10 (as a required by the ATM card installed in the EXODUS FBT). The same IP applications run on the two connections. Once the two connections (one between the SECOMS Terminal and FBT1 and one between the DECT Portable Part and the FBT2) are set-up and the two applications started up, the van is moved along a predefined path which includes three different coverage areas: area A where the sole satellite coverage is available (outdoor out of buildings shadow area); area B where there is SECOMS/DECT coverage overlapping; area C where the sole DECT coverage is available. At each step data concerning TCP/IP are stored. Tests to obtain throughput data at TCP/IP level are executed while the van is actually mobile. The first run (slow moving test) is done as slowly as possible and the second run (fast moving test) is performed at about 15 - 20 km/h. DECT worked well for both runs and we got a few results from SECOMS for the second run. Data as for DECT have been collected both at the Inter Working Unit (IWU) and at FBT in the Laboratory. 6 SECOMS/SAMBA joint-trial execution SECOMS / SAMBA Trial aims at validating a mobile environment scenario where an open area / urban area (near and far building) coverage is assured by a dual-mode access network consisting of the Broadband Satellite and of the Digital Cellular W-ATM segments. The equipment for the SECOMS/SAMBA Trial consists of a dual mode ACCORD Mobile Terminal (SAMBA and SECOMS), the corresponding Base Stations, the ACCORD Core Network (ATM switch) and two EXODUS Fixed Broadband Terminals (FBTs). The ACCORD Mobile Terminal consists of two co-located subsystems: a SECOMS MBT (Mobile Broadband Terminal) and a SAMBA MBT. The Base Stations for SECOMS and SAMBA are connected via ATM to the ACCORD Core Network. The dual-mode terminal consisting of both the SECOMS MBT and the SAMBA MBT is mounted on a van. The SECOMS MBT consists Expires: February 2001 [Page 4] Internet-Draft IP over ATM August 2000 of an EXODUS FBT (Fixed Broadband Terminal) interconnected to the SECOMS IWU connected to a CommQuest TDMA modem, model CQM-248A. The SAMBA MBT consists of an EXODUS FBT interconnected with the SAMBA MTA (Mobile Terminal Adapter). The two EXODUS FBTs can be implemented either by 133 MHz PC's running Windows NT 4.0 Operating System (OS) and containing NTUA ATM card or by SUN workstations running UNIX OS and containing commercial ATM card. Two simultaneous calls towards two different EXODUS FBTs (network side) are established: one through the SECOMS access and the other one through the SAMBA access. Once the two connections (one between the SECOMS MBT and FBT1 and the other one between the SAMBA MBT and the FBT2) are set-up and the two IP-based applications started up, the van is moved along a predefined path . At each step data concerning TCP/IP are stored. The van circulates in a predetermined path were 3 different coverage areas are available: SAMBA only, SAMBA and SECOMS simultaneously, SECOMS only. 7 DECT/MEDIAN joint-trial execution The DECT/MEDIAN trial considers the indoor nature of MEDIAN coverage and aims at testing it with the DECT indoor coverage. The trial layout consists of a MEDIAN Terminal, a DECT Portable Part, the corresponding Base Stations, the ACCORD Core Network (ATM switch) and two Fixed Terminals: a SUN Workstation running UNIX Operating System (OS), an EXODUS Fixed Broadband Terminal (FBT) running Windows NT 4.0 OS. The Base Stations for MEDIAN and DECT are connected via ATM to the ACCORD Core Network. One of the objective of this configuration is to evaluate the TCP throughputs of two simultaneous calls evolving in parallel on the two access systems. Moreover, various test configurations have been used to characterise the performance of the MEDIAN demonstrator at TCP/IP level, both in stand-alone mode and in combination with EXODUS equipment. The comparison of the retrieved information with the information collected on the other access systems (especially for SECOMS) are very useful to investigate service mobility through different access segments. The configuration used for MEDIAN (stand-alone) TCP/IP level testing includes two HP PCs which serve as User Terminals, a MEDIAN Portable Station, a MEDIAN Base Station, an ATM switch and an Internet Advisor inserted between the HP PC and the MEDIAN Portable Station. The MEDIAN Base Station is connected via ATM to the ATM switch. The PCs that serve as User Terminals are equipped with ForeRunner PCA-200E ATM adapter cards. The driver software for these cards allows the specification of Peak Cell Rates on a per-PVC basis. Expires: February 2001 [Page 5] Internet-Draft IP over ATM August 2000 The operation of Internet protocols over ATM is enabled on the basis of the 'Classical IP' approach (PVC's) as specified in RFC1577, using LLC/SNAP encapsulation of the IP packets as per RFC1483. The Internet Advisor is put in monitor mode, thus making it possible to inspect the bi-directional traffic flow at ATM and higher levels. 'Ping' and 'Netperf' were used as tools for TCP/IP level testing. Applications that were run over this configuration include a Web browser / Web server with video content, FTP and NetMeeting. The MEDIAN/EXODUS test configuration consists of two SUN Workstations which serve as User Terminals, a MEDIAN Portable Station, a MEDIAN Base Station and a LightStream ATM switch. The MEDIAN Base Station is connected via ATM to the LightStream ATM switch. This configuration was used for TCP/IP level testing, using 'Ping' and 'Netperf' as test tools. The 'Netperf' version running on the SUN workstations made it possible to run UDP_STREAM tests as well as TCP_STREAM tests. (Using the MEDIAN configuration in stand-alone mode attempts to run UDP_STREAM tests only produced error messages, despite the fact that NetMeeting (UDP) worked). The overall transmit rate of the SUN workstations had to be reduced to 40 Mbit/s to prevent cell buffer overflows in the MEDIAN system. 8 ATM Level Measurement execution for SECOMS, SAMBA and MEDIAN Access Networks In the framework of the ACCORD trial activity, besides to evaluate the network performance at TCP/IP level, also ATM level measurements have been executed for satellite SECOMS, Digital Cellular W-ATM SAMBA and W-LAN MEDIAN access segments. As matter of fact the above mentioned access networks can be seen as mobile (SECOMS and SAMBA) and portable (MEDIAN) wireless ATM networks which access an Internet core network. In order to evaluate the ATM connectivity performance that each access segment is able to provide a set of ATM performance parameters has been selected for trial measurements. These parameters are: Cell Transfer Delay (CTD), Cell Delay Variation (CDV), Cell Loss Ratio (CLR), Cell Error Ratio (CER). The performance evaluation at ATM level of the different access networks includes the verification of the ACCORD network compliance with the ITU-T network performance objectives defined in the ITU-T Recommendation I.356. Therefore in the ACCORD trial activities only the ATM user information transfer performance parameters are measured, moreover no international backbone capabilities are taken into account. The SECOMS, SAMBA and MEDIAN test-bench configurations set to execute the ATM measurements have been implemented following the same approach: each test-bench configuration consists of the relevant access segment integrated with the ACCORD ATM switch Expires: February 2001 [Page 6] Internet-Draft IP over ATM August 2000 which in turn is connected to a Protocol Analyzer (ATM Measurement Equipment) configured in the loop-back mode. The protocol analyzer performs the twofold task of generating ATM traffic and analyzing the network performance. 9 IP over ATM performance and QoS trial results evaluation On the basis of the measurements executed in the above described trials, ATM and TCP/IP over ATM performance in multi-segment application can be evaluated. Moreover both subjective and objective QoS evaluation can be derived. As far as the DECT system is concerned the TCP Throughput measurements executed on the DECT access show that the actual rate available at TCP level is about 15 Kbps for single slot DECT available in the demonstrator. This is likely to be enough for signalling and control in the frame of Multi-access Multi-media network. Anyway limitation to single slot DECT was only for the demonstrator and the rate can be easily increased using Multi-slot DECT in the Target System. The available throughput over DECT is always quite good even with the terminal in motion. There can be anyway drop in the throughput and this implies that it has to foresee buffering and re-transmission in the data or signalling application to cope with instantaneous gaps in the throughput, and it has to limit the actual speed of the mobile. The second issue is likely not to be a problem in the actual context of complementary terrestrial satellite coverage that is near building. Also the use of multi-slots could significantly improve the behaviour of DECT. Moreover the throughput of the DECT despite being low has a high availability also where no Line of Sight is present, this suggests the use of DECT as preferred application control channel in the design of multi-segment/multi-access specific applications. As far as the SECOMS system is concerned the data gathered and processed show that the delay is a significant issue for satellite since it is higher than terrestrial links. The Round Trip Delay is not dependent on either the terminal position or the ICMP packet size, it is a constant delay of 510-530 ms. This implies that ATM transport does not affect the delay figures. Also the same is for IP if we do not exceed the MTU size (here MTU size at IP level was 1500 bytes). Anyway the effect of IP fragmentation on Round Trip Delay is not very impressive (consider also that the path is run two times). As a hint on Target Network signalling this means that the signalling message used on satellite will suffer significant delay (impact on signalling state machine timers). The effect of satellite link delay actually sums up for each fragment but is not dramatic (that is ~ 10 msec over 520 msec that is 2%). It is estimated Expires: February 2001 [Page 7] Internet-Draft IP over ATM August 2000 that also fragmentation in 2 or 3 fragment can be supported (packet size exceeding 2 times the MTU size). The ATM level measurements executed for the SECOMS access segment show that the CTD parameter is 252 ms, the CDV is 0.584 ms, the CLR is 2.41E-7, the CER is 1.3E-6. The measurements on SAMBA were targeted to have information for the case in which a complete access network inclusive of its own ATM switch interfaces with a Global Core Network supporting multi-segment mobility. The above allowed to study the impact of multi access radio when more than a network and operator is involved. This pointed out that problems can arise in ATM buffer management (ATM switches owned by different players). In turn this indicated that high level protocols need to have packet handling as TCP while datagram protocols as UDP do not behave well unless a fine tuning of all the resources on the transmission path is performed. As for round trip delay the executed measurement clearly show that for the SAMBA access segment the mobile terminal position and the ICMP packet size have significant impact on round trip delay. The SAMBA round trip delay figures are all not exceeding 250 ms (half of the average on SECOMS access); this implies that the delay impairment at SECOMS/SAMBA radio access change is foreseen to be a major requirement to be handled in design of both multi-segment/multi-access application as well as in design of target network signalling. The TCP throughput of SAMBA is good whenever good coverage is provided. The measurements show that the actual value for TCP throughput is 5 Mbps nevertheless when the radio condition worsened the actual average TCP throughput figure dropped to 2 Mbps. The ATM level measurements executed for the SAMBA access segment show that the CTD parameter is 1.1 ms, the CDV is 1.37 ms, the CLR is 1.32E-4, the CER is 1.95E-4. As far as the MEDIAN system is concerned the TCP Throughput measurements executed on the MEDIAN/EXODUS test configuration show that the actual rate available at TCP level is about 30 Mbps in downlink and 14 Mbps in uplink direction respectively. The ATM level measurements executed for the MEDIAN access segment show that the CTD parameter is 1.4 ms, the CDV is 1.7 ms, the CLR is 7.4E-6 and it includes also the errored cells. >From the obtained results of the ATM level measurements executed for broadband satellite, W-LAN and cellular W-ATM access segments, it can be deduced that the network performances guaranteed by the satellite SECOMS Expires: February 2001 [Page 8] Internet-Draft IP over ATM August 2000 and W-LAN MEDIAN access segments are compliant with the ITU-T network performance objectives defining the QoS class 1 and class 2. This implies that a network configuration for portable services obtained complementing the broadband satellite with the W-LAN segment, can support, on both access networks, services requiring QoS class 1 and 2. Service mobility between open and indoor environment is therefore assured. On the contrary SAMBA access segment performances are not compliant with the QoS class 1 and class 2 recommended figures. This implies that service mobility in a network configuration for mobile services based on the broadband satellite and cellular W-ATM segment, is not guaranteed. A possible reason of this situation can be found in the still prototypal nature of the SAMBA demonstrator and in the hostile atmospheric conditions at the moment of measurements. The above described ACCORD trial execution and the relevant obtained results along with the ACCORD Target Network design represent a first step toward the definition and the implementation of an advanced network supporting mobile IP QoS sensitive services. The ACCORD project results will be further expanded in the framework of the IST SUITED project which aims at developing and demonstrating an Internet based multi-segment (satellite/ terrestrial) network for business class and mobile services. 10 References [1] Gianfranco Ferrari, Lino Moretti, Paolo Conforto, Vincenzo Schena, Stefan Wolters, Steve Pitchers, Eliseu Macedo, Manuel Dinis, Nikolas Georgantas, "Trial Strategy Description", CEC Deliverable Number AC348-IT-ITL-DS-DS-R-031. [2] AC348 ACCORD Annual Project Review Report (Parts A1 and A2), November 1999. [3] ITU-T Recommendation series I.350, "B-ISDN ATM layer cell transfer performance". [4] M. Laubach, "Classical IP and ARP over ATM", RFC 1577, Hewlett-Packard Laboratories, January 1994. [5] J. Heinanen, "Multiprotocol Encapsulation over ATM Adaptation Layer 5", RFC 1483, Telecom Finland, July 1993. Expires: February 2001 [Page 9] Internet-Draft IP over ATM August 2000 11 Contact Clementina Tocci Alenia Spazio Via G. V. Bona, 85 00156 - Roma Phone: +39/06/ 4151-275 Fax: +39/06/ 4151-297 E-mail: c.tocci@rmmail.alespazio.it Expires: February 2001 [Page 10]