International Civil Aviation Organization
WORKING PAPER / ACP-WG-S/Web Meeting 6IP-02
29/8/2014
AERONAUTICAL COMMUNICATIONS PANEL (ACP)
SixthWeb Meeting of the Surface Datalink Working Group WG-S
Web Meeting
Agenda Item XX:AeroMACS VoIP Validation
Prepared and presented by Bruce Eckstein
SUMMARYAeroMACS SARPs recommend that implementations support Voice applications but do not contain specific requirements for the RF link to support Voice. This paper notes AeroMACS capabilities with respect to Voice implementation; it is not intended to present or validate requirements. This paper contains proposed material for inclusion in the validation report that supports using VoIP over AeroMACS. The paper reiterates prior testing results of data packet performance characteristics obtained at NASA-Exelis Cleveland AeroMACS test bed located at the Cleveland Hopkins Airport.
ACTION
The AeroMACS Working Group is invited to consider the information for inclusion in the AeroMACS Validation Report.
Contents
1.Introduction: AeroMACS VoIP Data Packet Testing
2.Segregation of Different Services in CLE Test Bed
2.1.Service Prioritization – Capability Area #1
2.2.Segment Differentiation – Capability Area #2
2.3.Mixed Traffic Types – Capability Area #3
2.4.Preemption of Services – Capability Area #4
List of Tables
Table 1. AeroMACS QoS Properties
Table 2. CLE Test Descriptions
Table 3. CLE Service Prioritization Tests
Table 4. CLE Segment Differentiation Tests
Table 5. CLE Mixed Traffic Type Tests
Table 6. Preemption of Services Test Summary
List of Figures
Figure 1 CLE Test Configuration
Figure 2. VLAN Data Segregation Test
1.Introduction: AeroMACS VoIP Data Packet Testing
The AeroMACS SARPs recommends providing voice services for the aeronautical community. The voice services that AeroMACS provides would be the transport of Voice over Internet Protocol (VoIP) data packets. VoIP encoding and decoding is performed via processes outside of the AeroMACS standard and the resulting data from the VoIP encoded packets would be provided to the AeroMACS for transfer to the VoIP application for processing. VoIP data packets could be for transfer from one Mobile Station (MS) to another MS within the AeroMACS system or an AeroMACSMS with a ground network connected VoIP subscriber station. The AeroMACS standards recommendation for voice does not identify any specific performance requirements for AeroMACS to meet. International Civil Aviation Organization (ICAO) has not developed any VoIP standards for Air/Ground Communications and the ICAO VoIP standards for Ground/Ground do not indicate performance requirements such aslatency or dropped packet requirements typically important to VoIP systems and applicable asAeroMACS allocated requirements. Voice quality is mainly a function of external systems processing of data packets and as a result latency, dropped data packets and integrity of packet information are the important parameters for AeroMACS. As such, the information presented in this paper for data packet performance reflects AeroMACS characteristics for support of VoIP data packet transfer.
AeroMACS provides Quality of Service (QoS) capability for the transfer of data packets. Proper configuration of QoS in AeroMACS is needed to minimize the key characteristics of data packet latency (including data packet jitter) and dropped packets associated with VoIP. The tests that were performed were specifically to show the separation of data streams with different priorities. This information is also applicable to VoIP as the results show the characteristics associated with data packet transfer.
The tests were performed between AeroMACS MS and Base Stations (BS) on the surface of Cleveland Airport and identified the packet jitter (which includes the latency of the data packets in these test) and dropped packets response of the system. The prototype system did not have full mobility management implemented and therefore tests of these parameters during MS handoff from BS to BS handoff did not occur.
Testing was performed using VLANs (an optional capability within the AeroMACS standards) to support quantification of data transfer characteristics of an AeroMACS system between a BS and MS. To provide voice data packet performance separation from other types of data packet transfer to a single MS, either Internet Protocol Convergence Sublayer (ip.cs) or Ethernet Convergence Sublayer (eth.cs) can perform this function to support VoIP within AeroMACS. Implementation and configuration of IP Differentiated Services Code Point(DSCP) within the network may be required for use with ip.csconfigurations of AeroMACS. The tests identify the segregation of data packet characteristics by use of the notation of Air Traffic Control (ATC), Airline Operational Control (AOC), and Logistics Control Traffic. The tests showed that the appropriate selection of QoS services achieved low jitter and low loss of data packets when the total demand on the channel did not exceed the capacity of the 5 MHz channel bandwidth (approximately 8 Mbit per second) at the Cleveland Airport installation. The tests also showed data packets at one QoS service performance could be protected at the expense of data packets transferred at a lower QoS service when the capacity of the link was exceeded. These tests indicate the ability of the AeroMACS to provide the VoIP data packets with a quality of delivery for use by a VoIP application assuming proper configuration of the QoS system. The tests did not explore MS to MS characteristics (which would be through the BS) or MS to BS to different BS to MS as might be a typical airport surface VoIP operation. The following sections from the Exelis report SE2020 TO 0008 (TORP 1240) Task 3.1 describe the applicable tests and results.
2.Segregation of Different Services in CLE Test Bed
The ability for AeroMACS to segregate and prioritize traffic was evaluated using the Cleveland prototype network in a series of tests. The tests evaluated the following capability areas:
1)Service Prioritization - control scheduling and network priority for traffic from multiple services
2)Segment Differentiation - segregate and maintain traffic segregation from multiple operational segments like ATC and AOC operations
3)Mixed Traffic Types – simultaneous handling of multiple traffic types (continuous, burst, time-critical) using the UDP and TCP protocols
4)Preemption of Services – use of QoS and network priority settings to assure that low priority traffic (Best Effort) will be sacrificed to assure on-time delivery of higher-priority traffic in congested network conditions
The AeroMACS prototype network built at the NASA Glenn campus and the Cleveland Hopkins airport (CLE) is a full data network containing two base stations and eight fixed-site subscriber stations. The prototype network uses IEEE802.16e based system components. Central servers for CSN functions contain a secure network router, Network Management System (NMS) and Authentication, Authorization, and Accounting (AAA) functions.
Traffic segregation tests used a base station and two subscriber stations as illustrated inFigure 1. The AeroMACS network was configured to operate with three VLANS to evaluate a method for traffic isolation. Each VLAN was assigned a segment of traffic to represent ATC, AOC, and Control traffic as follows:
- ATC traffic assigned to VLAN 90
- AOC traffic assigned to VLAN 80
- Logistics control traffic assigned to VLAN 54
The AeroMACS test network configuration and VLAN assignments are illustrated inFigure 1. Three Single Board Computers (SBC’s) are used to generate test traffic. A SBC located at Building 110 (B110) receives traffic that represents logistics control. A network switch at the B110 Subscriber Station (SS) is set up for port tagging the traffic for VLAN 54.
Two SBCs at the Consolidated Maintenance Facility building (CMF) receive test traffic representing ATC and AOC traffic carried over VLANS 90 (ATC) and 80 (AOC). Traffic in the three VLANS pass through Aircraft and Firefighter’s building (ARFF) BS sector 2-3.
Network switches at the SSs and in B110 are configured for port tagging to establish traffic routing in the three VLANS. With this configuration, two VLANS (90 and 80) are carried simultaneously over the CMF-to-BS air link, and the ARFF BS sector 2-1 carries traffic from three VLANS simultaneously.
The AeroMACS prototype network is set to operate as Layer 2 (Ethernet IP) for these tests because VLANS are Layer 2 constructs. Operating as Layer 2 implements the Ethernet Convergence Sublayer (eth.cs) in the AeroMACS radio Media Access Control (MAC) layer.
Figure 1 CLE Test Configuration
In addition to the establishment of VLANS, three service flows were established within the AeroMACS service to establish QoS and priority settings for the simulated traffic from ATC, AOC, and Logistics services. These settings are summarized in Table 1.
Table 1.AeroMACSQoS Properties
TrafficService / VLAN Channel / Assigned QoS / Assigned Network Bandwidth
ATC / 90 / nrtPS / 6-54 Mbps
AOC / 80 / nrtPS / 1.5-7 Mbps
Logistics / 54 / BE / 0-1 Mbps
The network tests listed in Table 2 were performed to evaluate the four listed traffic segregation and priority criteria. A purpose and description is provided in the table for each test.
Table 2. CLE Test Descriptions
Purpose / DescriptionSegmentDifferentiation / TestsVLANprivacy byplacingtrafficsimulatingATC,AOC,and ControltrafficonVLANS90,80,and54individually
MixedTrafficTypes / MixedUDPandATCprotocoltraffictypesinhigh-priorityATCandAOCVLANS(90and80)
MixedTrafficTypes / MixedUDP andATCprotocoltraffictypesinhigh-andlow-priority ATCandControlVLANS(90and54)
ServicePrioritizationMixedTrafficTypes / Trafficprioritizationin congestednetworkconditions withmixedUDPandTCPtraffic
ServicePrioritizationMixedTrafficTypes / Test #4withhigh-priorityVLANS90and80differentiatedwithnrtPSbandwidthdefinitions
PreemptionofServicesMixedTrafficTypes / Congestednetworktestswithhigh-andlow-priorityATCandControlVLANS(90and54)
PreemptionofServicesMixedTrafficTypes / ReferencetestsfortrafficinsingleVLAN
PreemptionofServicesMixedTrafficTypes / Trafficprioritizationin congestednetworkconditions withmixedUDPandTCPtrafficusingallthreeVLANS
PreemptionofServicesMixedTrafficTypes / CongestednetworktestswithATC, AOCandControlVLANS(90,80and
54)withmixedtraffictypes
SegmentDifferentiation / TestsVLANprivacy byplacingtrafficsimulatingATC,AOC,and Controltrafficon VLANS90,80,and54simultaneously
The above tests were conducted in the four capability areas with the CLE prototype AeroMACS network configured according toFigure 1. Highlights of test results for each capability area are discussed below.
2.1.Service Prioritization – Capability Area #1
Description: Based on Test 10A, described in Table 3. Three traffic streams were established simultaneously in three VLANS for the purpose of testing correct service prioritization operation. The aggregate traffic rate of 7 Mbps is below the channel capacity of the service BS sector 2-3. Table 3 lists resulting link performance showing that high-quality links are supported for all three levels of service prioritization, indicated by low levels of percent dropped packets, out of order packets, and jitter. The delivered payloads are commensurate with the expected average traffic rate for the test time period. As expected, higher jitter statistics occurred for low-priority traffic using QoS of BE on VLAN 54. Overall, the results show proper operation for a BS sector channel operating below capacity with three concurrent levels of traffic priority and three established VLANs.
Table 3.CLE Service Prioritization Tests
2.2.Segment Differentiation – Capability Area #2
Description: Based on Test #10B, described in Table 4. Three traffic streams were established simultaneously in three VLANS for the purpose of testing traffic segregation. The test was designed to show that traffic stream content is visible only within their assigned VLAN. Properties of the test traffic streams are described below and in Table 4.
- Test traffic stream #1 sent through VLAN 90 to represent ATC traffic using UDP protocol at 4.5 Mbps and QoS of non-real-time poling service (nrtPS).
- Test stream #2 sent through VLAN 54 to represent Logistics Control traffic using UDP protocol at 0.5 Mbps and Best Effort (BE) QoS.
- Test stream #3 sent through VLAN 80 to represent AOC traffic using UDP protocol at 2 Mbps and nrtPSQoS
Table 4. CLE Segment Differentiation Tests
Execution: Traffic flow through the three VLANs identified in Figure 1was verified. ATC traffic was started first and set to run for 5 minutes. The AOC and Logistics traffic streams were started at the one-minute mark after the ATC data started. Both were set to run for 3 minutes. The performance data was recorded.
The next step that verified traffic segregation within a VLAN was completed by establishing traffic through one VLAN route at a time to verify that this traffic could not be observed outside of the established VLAN connection. The three Iperf client traffic sources to the right of the secure router in Building 110, shown in Figure 2, were activated one at a time. For each activated VLAN, a laptop computer located in the CMF Building was sequentially connected to all active switch ports to test for connectivity with the three Iperf Clients in B110.
Figure 2. VLAN Data Segregation Test
Results: The original test with three VLANS activated with traffic, operated as expected with traffic delivered to the remote end points through AeroMACS. Correct VLAN operation was verified when each VLAN was activated individually with traffic and all switch ports in the CMF building were probed. The remote laptop PC was able to establish a connection and a traffic flow when switch ports of the same VLAN were probed, and no connection or traffic flow occurred when mismatched VLAN channels were probed.
2.3.Mixed Traffic Types – Capability Area #3
Description: Based on Test #8described inTable 5. Three traffic streams are established to flow simultaneously with one stream per VLAN. The established flows are a mixture of QoS and IP protocol. Test results show that AeroMACS channel capacity and quality of link is maintained when carrying mixed traffic types.
Test #8 uses the following three traffic flows:
- Test traffic stream #1 sent through VLAN 90 to represent ATC traffic using UDP protocol at 6 Mbps and QoS of non-real-time poling service (nrtPS).
- Test stream #2 sent through VLAN 54 to represent Logistics Control traffic using TCP protocol and Best Effort (BE) QoS.
- Test stream #3 sent through VLAN 80 to represent AOC traffic using TCP protocol and nrtPSQoS
AOC and Logistics Control traffic are assigned to use TCP protocol. TCP is a guaranteed-delivery protocol so it will transfer at the fastest rate that it can achieve within the limits set by the services granted. The ATC traffic is assigned the nrtPSQoS that provides higher-priority scheduling than the BE Logistics Control traffic.
ATC and AOC traffic are both assigned nrtPSQoS. However, the traffic is differentiated by the committed bandwidth setting; ATC with 6.0 to 54 Mbps and AOC with 1.5 to 7.0 Mbps bandwidth. Traffic is also differentiated through use of UDP protocol for ATC and TCP protocol for AOC traffic.
Channel capacity for traffic rate will not be exceeded because the ATC traffic with UDP protocol is assigned a rate below the expected 8 Mbps channel capacity and the two additional traffic flows are assigned TCP protocol.
Table 5. CLE Mixed Traffic Type Tests
Execution: The ATC test traffic is launched first and is set to run for 5 minutes (300 seconds). AOC traffic is launched at the 1 minute mark, closely followed within approximately 2 seconds by start of the Logistics Control traffic, with both set to run for 3 minutes (180 seconds).
Results: High-priority ATC traffic is reliably delivered across AeroMACS while the lower-priority traffic is adjusted according to the total channel capacity of 8.13 Mbps. ATC traffic is reliably delivered using UDP protocol in the presence of two simultaneous TCP streams.
2.4.Preemption of Services – Capability Area #4
Description: Based on Test #6described inTable 6.Test traffic stream #1 is sent through VLAN 90 to represent ATC traffic using UDP protocol initially at 6 Mbps rate and non-real-time polling service (nrtPS) QoS. Test stream #2 is sent through VLAN 54, representing Logistics Control traffic, using TCP protocol initially at 1 Mbps and Best Effort (BE) QoS.
Simultaneous use of UDP and TCP protocols tests the behavior of mixed data types. Test trials with increasing traffic rates on the VLANS explore the effects of QoS on preemption of services. It is expected that higher-priority ATC traffic, set to nrtPSQoS, will be delivered with priority over Logistics traffic that is assigned BE QoS.
Execution: ATC traffic is set to run for 5 minutes (300 seconds) and begins first. The Logistics traffic is set to run for 3 minutes (180 seconds) and starts 1 minute after ATC traffic starts. Therefore, ATC traffic has no competition for network resources for the first minute. ATC data continues for the final minute after Logistics traffic finishes. This test sequence is performed at three traffic rates to test prioritization.
Results: Table 6summarizes the Preemption of Services test conditions and results. Tests are listed in three groups as 6A, 6B, and 6C. Test 6A used an aggregate traffic rate below the AeroMACS link capacity, while 6B and 6C used progressively higher rates in congested network conditions where the test traffic rate exceeds the AeroMACS link capacity.
Table 6. Preemption of Services Test Summary
Test 6A results: ATC data transferred at an average rate of 6.0 Mbps, matching the set rate for the test. The total data transfer was 214 Mbytes with a packet loss rate of .07%, which is under the 1% threshold considered to be the limit for a well-performing UDP channel. Logistics data transferred at an average of .96 Mbps (maximum rate was set for 1 Mbps). The total Logistics data transferred was 20.6 Mbytes. These rates and total transfers are consistent with a channel that is operating below its capacity.
Test 6B results: Tests are repeated with ATC traffic rate increased to 8 Mbps while the Logistics traffic rate remained at 1 Mbps. ATC traffic at 8 Mbps is near channel capacity by itself. These test trials resulted in ATC traffic throughput rate increasing to 7.99 Mbps and a total payload transfer of 286 Mbytes. The packet loss rate increased slightly to 0.17%, which is still an acceptable rate for a UDP link. However, the Logistics traffic with BE QoS was not able to sustain the 1 Mbps rate; averaging 0.14 Mbps instead. The total payload transfer over the length of the test was limited to 3 Mbytes. These results clearly show that high-priority ATC traffic was transferred without impact, while lower-priority Logistics data rate was sacrificed for the ATC traffic.