Вы находитесь на странице: 1из 24

I S O C O R E

T E C H N I C A L

R E P O R T

Evaluation of Alcatel-Lucent IP/MPLS Mobile Backhaul Solution

Areas of Evaluation:

GSM/UMTS/HSPA Mobile Backhaul and RAN Aggregation CDMA/EVDO Mobile Backhaul and RAN Aggregation Circuit Emulation Services over Packet Network Network Synchronization over Packet Network Resiliency and Redundancy

Isocore Internetworking Lab Isocore Technical Document Reference: ITD-13016v1.4 Version (v1.4): 06/23/2008

EXECUTIVE SUMMARY
In April 2008 Isocore was commissioned to carry out an independent, comprehensive evaluation and testing of Alcatel-Lucents IP/MPLS portfolio for mobile backhaul a part of the Alcatel-Lucent Mobile Evolution Transport Architecture (META).

Overview

We evaluated and tested Alcatel-Lucent IP/MPLS products focusing on the Alcatel-Lucent 7750 Service Router and the Alcatel-Lucent 7705 Service Aggregation Router as Systems Under Test (SUT) - for CDMA/EVDO and GSM/UMTS/HSPA backhaul, as well as for transport of Circuit Emulated Services (CES) and timing distribution over IP/MPLS. Special focus was placed on resiliency the ability to recover as a key theme for this event. For this reason, numerous failure and recovery scenarios were evaluated and tested. This report summarizes the key findings of the tests conducted.

Isocore set forth stringent test requirements for IP/MPLS Mobile Backhaul (MBH). The test requirements were based on information provided by Isocores service provider members. The tests were designed to address the actual mobile backhaul needs and real-life requirements of the mobile service providers.

Isocore validates the tested Alcatel-Lucent IP/MPLS architectures, related platforms and features a flexible, reliable and deployable end-to-end complete solution for mobile backhaul (MBH) for CDMA/EVDO and GSM/UMTS/HSPA. With successful testing of Circuit Emulation Services, adaptive timing recovery and differential timing distribution over packet switched network (PSN), Isocore also finds the Alcatel-Lucent IP/MPLS solution a reliable and deployable solution for transport of Circuit Emulation Services in real networks and the implementation of timing distribution over IP/MPLS stable and reliable.

In addition, emphasis was placed on evaluating and testing of the ability to support growing bandwidth demands from the radio access network (RAN). For this purpose, RAN scalability was introduced through scaling of the links carrying base station traffic (multiclass MLPPP for CDMA/EVDO and ATM/IMA for GSM/UMTS/HSPA).

These tests certify that todays Alcatel-Lucent IP/MPLS solutions to the cell site fully support all 2G and 3G technologies and that the solution satisfies and exceeds the requirements for resilient voice services, network timing, and general high availability mobile transport. The tests also verify and highlight the readiness of the IP/MPLS mobile backhaul solution to support tomorrows 4G, LTE, and enhanced packet core solutions as they move to complete end to end IP.

Isocore Internetworking Lab

TECHNICAL OVERVIEW
For the purpose of this testing, a simulated RAN environment for CDMA/EVDO and GSM/UMTS/HSPA was created to impose significant stress on the system under test. To verify the stability of the tested platforms and consistency of the results, multiple iterations of each test were conducted, with extended traffic tests run on a nightly basis across all CMDA/EVDO and GSM/UMTS/HSPA MBH interfaces and services. The information below captures the test bed overview and key results. 1,650 base stations aggregated across a shared CDMA/EVDO and GSM/UMTS/HSPA environment: Over 1550 base stations supported redundant MBH connections using multilink bundles (MLPPP and ATM/IMA), dual homed with multi-chassis automatic protections switching or both Averaged 2.8 DS1/E1 circuits for MBH per base station 113 channelized OC3 and OC12 ports (using Any Service Any Port cards) Total of 4,874 active DS1 and E1 circuits in the network All CDMA/EVDO and GSM/UMTS/HSPA MBH tests were performed successfully. All resiliency tests were performed successfully and demonstrated the following failover times (worst cases numbers reported): 2.7 ms - intra-chassis HA SF/CPM switchover with stateful MLPPP for CDMA/EVDO 810 ms - inter-chassis (MC-APS) switchover for MLPPP, with stateful MLPPP for CDMA/EVDO 1.2 ms - intra-chassis HA SF/CPM switchover for CES and ATM/IMA for GSM/UMTS/HSPA 154 ms - inter-chassis (MC-APS) switchover for CES and ATM/IMA All Circuit Emulation Services (CES) tests were performed successfully: Tests included multiple extended-duration hitless traffic runs with no traffic loss. 24-hour wander measurements of adaptive and differential timing for CES. Measured MTIE and TDEV results stayed well under the ANSI/T1.101 defined mask
KEY RESULT S: TEST OVERVIEW:

Isocore confirms that Alcatel-Lucent passed all tests and fulfilled all requirements assembled for this evaluation and testing event. Testing of resiliency mechanisms and features provides assurance that voice calls can be maintained even in catastrophic node failure scenarios, as the recovery mechanisms and features were able to successfully preserve integrity of these calls. The tested systems exhibited stability throughout the testing event and delivered consistent results.

Isocore Internetworking Lab 3

CONTENTS
TECHNICAL OVERVIEW .......................................................................................................................................................................................................................... 3 CDMA/EVDO IP MOBILE BACKHAUL EVALUATION ................................................................................................................................................................ 6 INTRODUCTION.................................................................................................................................................................................................................. 5 EXECUTIVE SUMMARY .......................................................................................................................................................................................................................... 2

2.1.2 3 2.1.3

2.1.1

2.1

3.3

3.2

3.1.1

3.1

LIST OF ACRONYMS USED ................................................................................................................................................................................................ 23

CONCLUSION .................................................................................................................................................................................................................. 22

SIMULATION OF BASE STATIONS (BTS, NODE-B) AND RADIO CONTROLLERS BSC/RNC ........................................................................................................ 20

EVALUATION OF NETWORK SYNCHRONIZATION TECHNIQUES.............................................................................................................................................. 17

Resiliency of GSM/UMTS/HSPA IP/MPLS MBH ................................................................................................................................................................ 16

VALIDATING IP/MPLS BACKHAUL FOR GSM/UMTS/HSPA ENVIRONMENTS ........................................................................................................................ 13

GSM/UMTS/HSPA IP/MPLS MOBILE BACKHAUL EVALUATION............................................................................................................................................ 13

Multiclass MLPPP with 4 QoS (MC-4)................................................................................................................................................................................ 11

Verification of stateful MLPPP resiliency with multichassis APS (MC-APS) in case of switching fabric/ control processing module failures (high availability)...................................................................................................................................................... 11

Verification of stateful MLPPP resiliency with multichassis APS (MC-APS) in case of link, card and node failures................................................................ 8

VALIDATING A DEPLOYABLE COLLAPSED CDMA IPBH ARCHITECTURE .................................................................................................................................... 6

Isocore Internetworking Lab

This report provides a summary of Isocores independent evaluation of Alcatel-Lucents IP/MPLS mobile backhaul (MBH) solutions for CMDA/EVDO and GSM/UMTS/HSPA environments. In order to optimize the available hardware resources, a single testbed was used to evaluate the two mobile technologies simultaneously. All testing was conducted under conditions that emulated large-scale real-world radio access network (RAN) environments. Under all conditions, all services were operational simultaneously. The key areas of evaluation included: Basic tests of MBH for CDMA/EVDO and GSM/UMTS/HSPA MBH scalability tests: Large scale RAN IP/MPLS backhaul for GSM/UMTS/HSPA deployments Large scale RAN IP backhaul with MLPPP for CDMA and EVDO deployments Large scale aggregation of base station traffic using multilink bundles (ATM/IMA for UMTS/HSPA and MLPPP for CDMA/EVDO) MBH resiliency tests: Inter-nodal failover with multichassis APS for CES, ATM (IP/MPLS) and MLPPP (IP) Included multiple link, card and node fault convergence scenarios Intra-nodal high availability SF/CPM failover for all active services Pseudowire redundancy protection for IP/MPLS MBH CES, ATM and ATM/IMA over redundant pseudowires (RPW) Protection of MLPPP states across different nodes using MC-APS for IP BH Evaluation of MBH architectures related to IP and IP/MPLS MBH Evaluation of synchronization options and timing distribution over packet-based network Adaptive clock recovery (ACR) with extended wander measurements for CES Differential timing with extended wander measurements for CES MBH latency and impact on the services MBH quality of service (QoS) Testing of multiple QoS classes with multi-class 4 and MLPPP for CDMA/EVDO Isocore placed emphasis on evaluating the end-to-end network availability of the MBH solutions presented by Alcatel-Lucent. Numerous test cases were conducted on the end-to-end resiliency of the platforms and architecture. Multiple iterations of each case were executed to assure consistency of the measurements.

1 Introduction

A single 7750 SR-12 was used to emulate all BTS and Node-B sites as well as Base Station Controller/Radio Network Controller (BSC/RNC) functionality for GSM/UMTS/HSPA. To emulate mobile subscriber traffic and to generate multi-class traffic flows for high/low priority data, voice, and control traffic, Isocore utilized the Agilent N2X multi-service test solution. Agilent N2X offered comprehensive measurements of individual streams (latency, jitter, packet loss), which were critical for the validation of the Alcatel-Lucent IP/MPLS MBH solution.
Isocore Internetworking Lab
5

The architectures evaluated in this evaluation and testing event were those presented by Alcatel-Lucent for their IP and IP/MPLS MBH deployments. The primary systems under test (SUT) were the 7750 Service Router and the 7705 Service Aggregation Router. The following variants for these product lines were used: the 7750 SR-12, the 7750 SR-7 and the 7705 SAR-8.

Another key requirement in MBH deployments is the ability to support the increased bandwidth requirements imposed on the mobile aggregation layer by the traffic (new, bandwidth-intensive mobile services). Multiple DS1/E1 circuits are bound together using MLPPP or ATM/IMA to support the bandwidth requirements of next generation MBH deployments. The ability to support large scale aggregation of multilink bundles was considered a key requirement for this event.

2 CDMA/EVDO IP MOBILE BACKHAUL EVALUATION


This section provides details of Isocores evaluation and testing of Alcatel-Lucents RAN backhaul architecture for CDMA and EVDO for IP mobile backhaul (IP MBH). Typically, CDMA deployments involve use of IP directly from the base station (BTS). The use of legacy transport options has resulted in a requirement to send IP using PPP (over SONET type of transport), and MLPPP has been used in order to increase the bandwidth of this mode of transport by bundling several member DS1 links into the aggregated MLPPP group. This approach calls for the deployment of separate aggregation routers (AR), which aggregate the traffic from a base transceiver station (BTS), and multilayer switches (MLS) a Layer 2 switch used to connect to the essential sub-elements of a CDMA Mobile Switching Center (MSC) Radio Network Controller (RNC), Mobility Manager (MM) and packet switch (PS). As CDMA mobile technology has evolved towards EVDO, the aggregation devices have also varied, requiring multiple devices used for aggregation and switching.

Isocore placed emphasis on evaluating the deployability (applicability in real-life deployments) of this collapsed CDMA/EVDO IPBH architecture, the involved platforms and their related feature set. Considerable time was spent evaluating the resiliency of the architecture and platforms in simulated large scale RAN environment.
2 .1 V A L I D A T I N G A D E P L O Y A B L E C O L L A P S E D C D M A I P B H A R C H I T E C T U R E

The architecture Alcatel-Lucent presented for evaluation and testing used a simplified CDMA RAN architecture in which one element the 7750 Service Router performs the combined functionality of the aggregation router (AR) and Multi-Layer Switch (MLS). Typically, aggregation routers and multilayer switches are deployed in redundant pairs, and for the same reason the 7750 SR routers were deployed in a redundant pair. In this report, we will be referring to this architecture as collapsed AR+MLS architecture or collapsed CDMA/EVDO IPBH architecture.

Alcatel-Lucent defines the stateful protection of the MLPPP groups as follows: in the case of a failure of the physical link carrying the traffic with active MLPPP sessions, the mechanism of protection (which involves APS or MC-APS) is such that it ensures end integrity of MLPPP groups (terminated on the aggregation router pair), so it is not required to re-initiate MLPPP sessions (allocation of member links to MLPPP groups and synchronization between endpoints). This implementation results in much shorter convergence times and preservation of voice calls in progress. This stateful protection covers multiple failure scenarios such as: link, card, port and Switching Fabric / Central Processing Module (SF/CPM) failures. For these tests, working (protected) and protection links were equally split across the two SUTs. This setup represented a large-scale real-world deployment for CDMA and EVDO IP MBH.

Figure 1 shows the detailed setup used to evaluate Alcatel-Lucents collapsed CDMA/EVDO IPBH architecture. The primary systems under test (SUTs) were two 7750 SR-7s. The SUT aggregated the traffic coming from 604 fully protected base stations (BTS). Each BTS was configured to use MLPPP bundling and the MLPPP sessions were terminated on L3 routed interfaces on the aggregation router (7750 SR-7). A total of 2600 DS1s coming from BTS sites were configured as member links to MLPPP bundles. The 604 MLPPP groups were configured with one, two, four or eight DS1 member links per bundle. On the aggregation routers a total of 32 OC3 ports were configured for the CDMA/EVDO test environment. Multichassis APS (MC-APS) protection was enabled in order to achieve stateful preservation of the MLPPP groups and member links between two adjacent aggregation routers.

Isocore Internetworking Lab

In our tests, a 7750 SR-12, shown on the left side in figure 1 was used to simulate the combined functionality of large numbers of base stations (BTSs) and a digital cross-connect system (DACS). This 7750 SR-12 terminated all 32 working and protection OC3 channelized circuits, carrying 1208 combined active and standby MLPPP bundles, with 2600 DS1 member links. ARP mediation was required between the 10 Gigabit Ethernet attached N2X test solution and each of the protected MLPPP bundle pairs. In order to perform this functionality, 604 local ARP-mediation pseudowires were configured on the 7750 SR-12 and verified per draft-ietf-l2vpn-arp-mediation.

Figure 1: Testing Alcatel-Lucent collapsed CDMA/EVDO IPBH architecture

The QoS requirements can differ for CDMA and EVDO IPBH. For this test all 1208 (604 fully protected pairs) CDMA/EVDO MLPPP BH terminations were configured with support of multi-class MLPPP, (MC-4), allowing each MLPPP bundle to have four (4) unique traffic classes (from now on referred to as forwarding classes (FCs). Separate FCs were used for control traffic, voice, high-priority data and low-priority data. Each FC had a separate queue and scheduling rate. Alcatel-Lucents MC-4 implementation can dynamically adjust the bandwidth (BW) allocation across FCs when a member link(s) (DS1s) is lost.

Isocore Internetworking Lab

The setup included evaluation of basic functionality proper termination of MLPPP sessions. The tests were conducted and latency for different types of traffic was measured. During these basic tests with all CDMA/EVDO and GSM/UMTS/HSPA traffic running concurrently the voice latency averaged 2.0 ms and even latency of low-priority best effort (BE) data traffic stayed below 3 ms in an uncongested state. Latency measurements for the CDMA/EVDO were measured across a cascaded setup involving three nodes: the 7750 SR12 (used for BTS emulation), a 7750 SR-7 (deployed as RAN aggregation router), and a 7450 Ethernet Service Switch (ESS-1) which was used to simulate the MSC environment MM, PS and RNC, which are typically connected in Layer 2 switched environment (to an MLS). During these tests, obtained latency measurements were consistent and even under severe congestion voice latency remained below the required 5 ms range for critical voice and control traffic. Latency under congested states is covered in the next sections (see under MC-4 QoS testing). Throughout the event, numerous iterations of link, card, SF/CPM and catastrophic node-level failures were conducted. Throughout the test event all traffic (CDMA/EVDO and GSM/UMTS/HSPA) was run overnight to validate the stability of the deployment. Traffic was run for these long durations without loss. Even with the considerable traffic stress, all systems remained stable throughout the event. Table 1 summarizes the scaling numbers used for the CDMA/EVDO IP BH test.

Table 1: Summary of CDMA and EVDO IP MBH configuration


DETAILS OF NETWORK AND NODAL ELEMENTS VERIFIED FOR CDMA IP MBH

604 MLPPP protected bundles on each 7750 SR-7 (AR) (with MC-APS ) 1300 total DS1 member links configured per 7750 SR-7 (AR) 1208 MLPPP bundles terminated on 7750 SR12 (BTS) 2600 DS1s member links terminated on 7750 SR12 (BTS)

CDMA Network Details:

A mix of MLPPP groups (bundles) using 1, 2, 4 and 8 DS1s member links

604 L3 Interfaces per 7750 SR-7 (AR) used for MLPPP aggregation from BTS 604 multiclass MLPPP (MC-4) QoS profiles per 7750 SR-7 (AR) 32 total OC3 links supported with MC-APS on the 7750 SRs Traffic integrity verification of voice, control traffic, high-priority data and low-priority data

All failure scenarios were carried out using a large scale of concurrent MLPPP sessions (i.e., the tests were not done on one individual MLPPP sessions running at the time). To validate the functionality and performance of Alcatel-Lucent stateful MLPPP with MC-APS implementation, the following tests were conducted: Link-layer fault (fiber cut and port shutdown scenarios) Multiple port shutdowns Card-level failures (card pull) Catastrophic node failures (bringing down the whole node) High availability validation of SF/CPM (switching fabric/central processing module) failures.

The tested Alcatel-Lucent CDMA/EVDO IP MBH architecture provides protection against link, card, SF/CPM, and node failures. In accordance with explanation made to Isocore by Alcatel-Lucent, it was expected that regardless of the network fault, the BTS would not require renegotiation of MLPPP sessions. Also, in order to assure that voice calls are not being dropped by the Mobile Switching Center (MSC), traffic interruption (or loss) should never exceed 5 seconds.

2.1.1 VERIFIC ATION OF STATEFUL MLPPP RESILIENCY WITH MULTICHASSIS APS (MC- APS) IN C ASE OF LINK, C ARD AND NODE FAILURES

Isocore Internetworking Lab

Figure 2 illustrates the setup that was used for failing the working (protected) multilink groups from one aggregation router (7750 SR-7) to another. The logical MC-APS link included a working (protected) SONET link on one node and a protection link on the adjacent neighbor AR. The working and protection links were equally split across the SUTs. Each of the 604 protected MLPPP bundle pairs were protected by MC-APS. The MC-APS was configured to revert back to the working facility link one minute after a fault was cleared. Revert was used for this event to maintain the location of the 32 OC3 working and protect circuits throughout the event and to demonstrate the recovery times associated with revertive behavior. During all failure scenarios the MLPPP states representing the BTS sites were monitored on the 7750 SR-12 (BTS) to verify if they remained up and active. The traffic flows originating from the Agilent N2X platform used the same traffic profiles as the basic setup. Traffic loss was measured separately from the aggregation network towards the BTS and from the BTS towards the aggregation network.

Figure 2: Verification of 7750 SR stateful MLPPP and MC-APS implementation

As expected, traffic from the MSC to the BTS exhibits a higher loss due to the higher number of in-flight packets impacted in the outbound direction (towards BTS). Multiple iterations of each test case were conducted to validate consistency of the measurements and worst case numbers are presented in the respective tables. Basic traffic flows of control traffic, voice, high-priority data and low-priority data were used for all measurements.

Table 2 shows the results of failure scenarios executed by command line interface (CLI), and table 3 shows results of physical interruptions caused by fiber cuts or card pulls.

Isocore Internetworking Lab

Table 2: Results summarizing the stateful MLPPP link recovery through MC-APS (case of link failures)
Single port failure with MC-APS Failing link between AR and BTS with 21 MLPPP bundles Recovery times from BTS to MSC 80ms Revert times from BTS to MSC 130ms Revert times from MSC to BTS 403ms Recovery times from MSC to BTS 810ms
RECOVERY TIMES FOR ADMINISTRATIVELY INDUCED FAILURES (USING MC-APS)

Single port failure with MC-APS Recovering link between AR and BTS with 21 MLPPP bundles Four port failure with MC-APS Failing link on RAN side (towards BTS) with 176 MLPPP bundles Recovery times from MSC to BTS 917ms Recovery times from BTS to MSC 210ms Revert times from BTS to MSC 640ms Revert times from MSC to BTS 967ms Four port failure with MC-APS Recovering link between AR and BTS with 176 MLPPP bundles

The worst-case node-level failure resulted in traffic loss for 2.07 seconds, which was well below the target of 5 seconds a requirement set forth by Isocore based on input from mobile providers.
Table 3: Results summarizing the MLPPP link recovery through MC-APS (case of fiber cuts)
Single fiber pull-out with MC-APS failing link between AR and BTS with 21 MLPPP bundles Recovery times towards MSC 100ms Revert times towards MSC 100ms Revert times towards BTS 576ms Recovery times towards BTS 576ms
RECOVERY TIMES FOR PHYSICAL DISRUPTIONS INCLUDING FIBER CUTS AND CARD PULL-OUTS (USING MC-APS)

It is important to note that during all node-level failures, the entire end-to-end IP BH architecture was being tested, including the L2 switch (7450 ESS-12), configured with spanning tree protocol (STP), which simulated the MSC environment.

During the node-level failures, 16 OC3 SONET links were interrupted, including 8 working OC3 MC-APS links carrying traffic from 302 base stations (MLPPP bundles) with 750 DS1 member links in total.

Node-level failure scenarios of the aggregation routers were also conducted. The AR (7750 SR-7) was rebooted while the MLPPP states were monitored at the BTS (7750 SR-12), and traffic loss was measured in each direction.

Multiple fiber cuts were performed by scripting the CLI shutdown of multiple ports simultaneously. Card failures consisted of pulling out the 4-port OC3 Any Service Any Port (ASAP) MDAs which failed multiple ports simultaneously. The results show that regardless of the number of ports failed during port and card level failovers the worst-case traffic loss was below 1 second.

Single fiber insertion (link up) with MC-APS recovering link between AR and BTS with 21 MLPPP bundles Four (4) fiber pull-outs (MDA pull-out) with MC-APS failing link between AR and BTS with 176 MLPPP bundles MDA insertion with MC-APS recovering link between AR and BTS with 176 MLPPP bundles Node failure on AR 1 (reboot) 302 MLPPP bundles Recovery times towards MSC 1635ms Recovery times towards BTS 2072ms Revert times towards MSC 270ms Revert times towards BTS 588ms Recovery times towards MSC 1000ms Recovery times towards BTS 1800ms

10

Isocore Internetworking Lab

Based on the results presented in table 2 and 3, Isocore confirms the Alcatel-Lucent implementation of MC-APS supports the stateful protection of the MLPPP sessions and that this implementation is robust and scalable.
2 .1. 2 V E R I F I C A T I O N O F S T A T E F U L M L P P P R E S I L I E N C Y W I T H M U L T I C H A S S I S A P S ( M C - A P S ) IN C ASE OF SWITCHING FABRIC / CONTROL PROCESSING MODULE FAILURES (HIGH AVAIL ABILIT Y)

Administratively-induced and physical failures were repeatedly executed in order to stress the configured protection in form of MC-APS. During all these repeated tests no instance of MLPPP sessions was lost. The tests confirmed that traffic which is sent from the MSC to all wireless subscribers was minimally impacted. The MC-APS recovery time was consistent for each of the failure scenarios and at no point was any system instability detected.

The tests were performed with all CDMA/EVDO traffic being simultaneously active. To measure HA performance, the same basic traffic flows were exercised to all 604 BTS. Similarly, all HA tests performed on 7750 SR12 (simulating a large number of BTSs) included the combined case of concurrently active CDMA/EVDO and GSM/UMTS/HSPA traffic.

Our assessment also included the evaluation and testing of Alcatel-Lucent 7750 SR high availability (HA) feature set full redundancy of the control complex which supports non-stop routing and non-stop services, including stateful MLPPP sessions and MC-APS functionality. The HA feature set was verified during the tests by failing the SF/CPM and measuring the impact to forwarded traffic while monitoring the active MLPPP states. The tests included failing the SF/CPM on the aggregation routers (7750 SR-7) and also on the BTS (7750 SR-12) node.

Isocore confirmed that during HA failovers, the MLPPP sessions stayed active and no transitions were noted on any of the MC-APS ports or member DS1 links. The impact to traffic was minimal, with worst case of 2.7 ms for traffic loss. The tests confirmed that stateful MLPPP implementation by Alcatel-Lucent is fully supported by the observations of this test. Table 4 summarizes the test results of the HA testing.
Table 4: Results summarizing the recovery times for HA MLPPP failover
MC-APS with MLPPP high availability on 7750 SR-12 (SF/CPM card pull-out) No loss of MLPPP bundles observed No loss of MLPPP bundles observed Worst case traffic loss 2.7ms Worst case traffic loss after multiple SF/CPM failovers 3.5 s
RECOVERY TIMES FOR HIGH AVAILABILITY FAILURE SCENARIOS (FAILURE OF SF/CPM WHEN USING MC-APS)

MC-APS with MLPPP high availability on 7750 SR-7 Aggregation Router 1 (SF/CPM pull-out)

2 .1. 3 M U L T I C L A S S M L P P P W I T H 4 Q O S ( M C - 4 )

The test focused on ten (10) MLPPP bundles, each configured with four (4) DS1 member links. The MLPPP bundles carried voice, control traffic, high-priority data and low-priority data. Table 5 shows the forwarding classes defined for each of the forwarding classes. The bandwidth percentage shown in the table represents the bandwidth allocated to a specific forwarding class, and is expressed as maximum allowed. Under congestion, voice and control traffic are scheduled with a strict priority over the data classes. The allocation of bandwidth to
Isocore Internetworking Lab
11

As presented in the basic setup in section 2.1.1, all of the CDMA/EVDO MLPPP bundles were configured with MC-4 QoS to support multiple classes of traffic (forwarding classes). This section provides an overview of the evaluation testing of the Alcatel-Lucent MC-4 implementation under conditions of traffic congestion and also the evaluation and testing of the ability to dynamically adjust the bandwidth (BW) allocation if a DS1 member link in an MLPPP bundle is lost.

The data rates shown represent the total bandwidth transmitted by the Agilent N2X in the egress direction to drive each forwarding class across all 10 MLPPP bundles. The table also shows the latency measured with traffic levels near saturation for the 4xDS1 member links.
Table 5: Traffic types (forwarding classes) used for the multiclass QoS evaluation
TRAFFIC TYPE

four classes stayed in effect in cases of failures the stated percentages were also supposed to stay in effect for the adjusted bandwidth (after DS1 member links were lost), and this is shown in table 6.

Network control Voice

FORWARDING CLASS

High-priority data Low-priority data

Expedited Forwarding Assured Forwarding Best Effort

Network Control

PERCENTAGE OF BW

66% 33%

85%

max

DATA RATE

200 Kbps 1.5 Mbps 3 Mbps 1.5 Mbps

LATENCY

2.66 ms 2.67 ms 2.74 ms

2.46 ms

Table 6 summarizes the results for three different test cases:

Test case 1 introduced congestion by increasing the BE low-priority data from 3 Mbps to 4.5 Mbps.

Test case 2 used the same traffic profiles but introduced a shutdown of a single DS1 member link on each of the 10 MLPPP bundles, reducing the bandwidth available to each bundle. Test case 3 shut down a second DS1 member link on each MLPPP bundle, further reducing the available bandwidth to each MLPPP bundle.

The table shows worst case latency values. In all three cases, the latency of critical-priority voice traffic remained well below 5 ms. Test case 3 represented a case of severe network traffic congestion, and both high-priority and low-priority data traffic was lost. However, high-priority data had significantly more throughput even in the case of this severe traffic congestion.
Table 6: MC-4 test case results
TEST CASE 1 BE CONGESTION Total traffic BW available 7.7 Mbps TEST CASE 2 SINGLE DS1 DOWN WITH BE CONGESTION 7.7 Mbps TEST CASE 3 TWO DS1S DOWN WITH BE CONGESTION 7.7 Mbps

Network control Voice High-priority data Low-priority data

Traffic loss No No No

6.144 Mbps

Latency

4.08 ms high

4.28 ms 4.36 ms

Traffic loss No No No

4.608 Mbps

Latency 4.26

Traffic loss No No 5.8% high

3.072 Mbps

Latency 94 ms high

4.23 high

4.40 ms

32%

high

4.58

4.36 ms

12

Isocore Internetworking Lab

This section presents the results of the evaluation and testing of Alcatel-Lucents architecture, platforms and related feature set for IP/MPLS MBH for GSM/UMTS/HSPA mobile services. The GSM/UMTS/HSPA IP/MPLS BH was tested simultaneously with the CDMA and EV-DO IP MBH (i.e. common test equipment was used for MBH for both GSM/UMTS/HSPA and CDMA/EVDO and all the tests were done with GSM/UMTS/HSPA and CDMA/EVDO simultaneously present). Emphasis in these tests was placed on the deployability of the architecture, platforms and related features specific to GSM/UMTS/HSPA, as well as to MBH resiliency. Figure 3 shows the physical topology of the network used for the evaluation and testing of Alcatel-Lucent IP/MPLS MBH solution for GSM/UMTS/HSPA. The testbed was composed of base station simulation layer, mobile aggregation layer, MPLS transport layer and the base station controller/radio network controller (BSC/RNC) layer. The base station simulation included the Agilent N2X multi-service test solution attached to the 7750 SR-12 using 10 Gigabit Ethernet interfaces. The radio access network (RAN) aggregation layer aggregated the simulated base stations (Node-B sites) carrying ATM VCs, ATM IMA bundles and CES services. The MPLS transport used LDP MPLS signaling and transported ATM/IMA and CES based services over redundant pseudowires (RPWs). The BSC/RNC layer was configured with multichassis APS protection. All CES, ATM/IMA and ATM VCs MBH services originating on the 7750 SRs in the aggregation layer were protected with RPWs.
Figure 3: Test architecture for Alcatel-Lucent IP/MPLS backhaul for GSM/UMTS

3 GSM/UMTS/HSPA IP/MPLS MOBILE BACKHAUL EVALUATION

3 .1 V A L I D A T I N G I P / M P L S B A C K H A U L F O R G S M / U M T S / H S P A E N V I R O N M E N T S

Isocore Internetworking Lab

13

Since the GSM/UMTS/HSPA architecture focused on IP/MPLS BH with RPWs, CES was also tested in this environment. CES transport is not unique to GSM/UMTS/HSPA MBH and can also be deployed in CDMA/EVDO environments. An OC3 CES port was configured on AR 1 (with 84 DS1 CES services per OC3). To stay consistent with MBH applications, structured transport of TDM DS1 circuits was used, implemented as per RFC 5086. 56 of the CES DS1 based services originating on AR1 were transported to the BSC/RNC edge, where they were protected by MC-APS and RPWs. The remaining 28 CES DS1s were transported over pseudowires to the two 7705 SAR deployed as ARs (14 to the each 7705 SAR). The CES services on the 7750 SR-7s used the external Stratum 1 primary reference clock (PRC) applied to the systems BITS input. The 7705 SARs derived their timing from the adaptive clock received on the CES circuit. To further stress the adaptive clock recovery, the CES services destined to the 7705 SARs were routed within the packet-switched network (PSN) using multiple hops. The handoff from the MPLS transport to the BSC/RNC used a pair of 7750 SR-7 routers. Typically, these edge routers support many ARs using an MPLS transport. For this reason, these 7750 SR nodes were lightly equipped with a mix of clear-channel OC12 ATM ports and OC3 channelized ASAP ports, and OC3 CES ports. Inter-node redundancy was provided by MC-APS and RPWs for the CES, ATM and ATM IMA MBH services.

All CES, ATM-VC and ATM/IMA MBH services originating on the 7750 SRs were protected within the MPLS network and at the BSC/RNC by RPWs and MC-APS. The RPWs were configured to track the MC-APS edge protection across the RAN AR and BSC/RNC nodes respectively.

The ATM/IMA bundles provided redundancy to the Node-B site with two E1 or DS1 member links per IMA bundle. The ATM VCs were dual homed to both aggregation routers (ARs) and protected by MC-APS and RPWs at the RAN AR. This provided two ATM E1 terminations for the redundant Node-B aggregation.

Figure 3 shows the logical representation where the 7750 SR-12 was used to simulate the Node-B (base station) sites (shown on the left) and also the BSC/RNC functionality (as shown on the right). The 7750 SR-12 was configured with thousands of local ATM-to-Ethernet interworking pseudowires to support the Ethernet-to-ATM handoff between the 10 Gigabit Ethernet attached Agilent N2X tester and the ATM based GSM/UMTS/HSPA environment. The 7750 SR-12 had a total of 2274 E1 and DS1 terminations from the combined AR and BSC/RNC handoffs related to the GSM/UMTS/HSPA setup. Table 7 summarizes the scaling numbers verified during the test for the GSM/UMTS/HSPA environment.

14

Isocore Internetworking Lab

Table 7: Baseline scaling setup for GSM/UMTS/HSPA IP/MPLS backhaul and aggregation
NETWORK AND NODE TOTALS VERIFIED FOR UMTS RAN AGGREGATION GSM/UMTS/HSPA Network Overview

1046 combined Node-B IP/MPLS MBH sites 646 IMA-based IP/MPLS MBH sites 316 ATM VC IP/MPLS MBH sites 84 CES IP/MPLS MBH sits

2274 E1 and DS1 terminations used for GSM/UMTS/HSPA MBH 630 protected ATM/IMA based IIP/MPLS MBH sites 252 7750 SR-7 AR 1 378 7750 SR-7 AR 2

GSM/UMTS/HSPA 7750 SR-7 Aggregation Router Overview:

252 protected ATM VCs IP/MPLS MBH sites 84 CES IP/MPLS MBH sites on AR 1

All ATM/IMA MBHs protected by RPWs and MC-APS at the BSC/RNC 252 protected ATM VC dual homed to both ARs using MC-APS Protected at RAN and BSC/RNC by MC-APS and RPW 56 CES IP/MPLS BHs protected RPWs and MC-APS at the BSC/RNC

GSM/UMTS/HSPA 7705 SAR-8 Aggregation Router Overview:

28 CES IP/MPLS BHs transported to 7705 SARs ARs

28 CES DS1s based IP/MPLS BH 64 ATM VC IP/MPLS BH 16 ATM IMA IP/MPLS BH

Trafc verication of all services included voice, control, high-priority data and low-priority data trafc

Overnight traffic runs were conducted across all GSM/UMTS/HSPA and CDMA/EVDO services simultaneously. The traffic was active for all traffic types (and forwarding classes configured), including voice, control traffic, high-priority data and low-priority data traffic. Repeated overnight runs were executed without traffic loss or any other signs of performance degradation. Latency measurements were made across four nodes for the GSM/UMTS/HSPA environment this included two hops across the 7750 SR12 for BTS and BSC/RNC simulation, one hop for the RAN AR and one hop for the BSC/RNC router. Voice averaged 3 ms of latency for the ATM-based GSM/UMTS/HSPA traffic. Latency remained consistent and low across the network regardless of the test cases. All CES services showed stability throughout the event, and CES tests included multiple overnight runs without any detected/recorded traffic loss. All of the ATM VC, ATM/IMA and CES traffic on the 7705 SARs used the adaptive clock recovery using CES throughout the event. The tests were conducted over the course of multiple extended traffic runs, totaling more than 50 hours and no packet loss was encountered on the 7705 SARs. This is a good indication that the adaptive clock recovery implementation on the 7705 SAR is stable. Further evaluation of the CES timing implementation is covered below (see dedicated section). Throughout the event considerable stress was placed on the nodes by repeatedly failing different elements such as links, cards and central processing modules, and also by starting and stopping traffic randomly. At no point during the tests was there any sign of instability on any of the systems tested. Isocore is comfortable stating the Alcatel-Lucent GSM/UMTS/HSPA IP/MPLS MBH was successfully evaluated and tested during this event.

Isocore Internetworking Lab

15

Multiple resiliency test cases were conducted on the GSM/UMTS/HSPA setup including link, node, and high availability SF/CPM test cases. All test cases except the HA SF/CPM test case were focused on inter-node convergence scenarios where traffic converged to the adjacent node. Figure 4 shows the redundancy verification setup for the GSM/UMTS/HSPA tests. The GSM/UMTS/HSPA test cases included the use of redundant pseudowires in the IP/MPLS transport layer. When the MC-APS becomes engaged, failure of the node results in the fail-over switching to the protection circuit on the adjacent node, and the RPW converges with the edge. LDP was used as the MPLS signaling protocol to demonstrate that there was no influence on the RPWs by RSVP fast reroute (FRR) protection. Traffic loss was measured in each direction and all services were tested. Table 8 provides the results of the various resiliency test cases.
Figure 4: GSM/UMTS/HSPA redundancy for PWs for CES, ATM-VC and ATM IMA

3 .1.1 R E S I L I E N C Y O F G S M / U M T S / H S P A I P / M P L S M B H

Isocore exercised the resiliency of the each component in this setup to ensure that end-to-end resiliency is functional and consistent, and that the RPW protection tracks the edge convergence. The access redundancy was verified by failing the working MC-APS circuit and ensuring the functionality of the PW redundancy. Its important to note that the CES services were also supported by MC-APS and RPWs. In the case of APS failovers the worst case loss was 154 ms core to edge, measured at the BSC/RNC for CES. Table 8 summarizes the results for the GSMUMTS/HSPA resiliency test cases. In all cases multiple iterations were conducted and worst case numbers are shown.

16

Isocore Internetworking Lab

Table 8: Resiliency and PW redundancy failure recovery measures


RECOVERY TIMES WITH PW REDUNDANCY AND MC-APS FOR ATM VCS AND IMA RAN AR protection link failure of MC-APS carrying 252 ATM VC each with RPWs

Single port fiber pull: Failing APS link carrying ATM traffic between AR 1 and Node-B Traffic loss from Node-B edge to MPLS core 52ms Traffic loss from MPLS core to Node-B edge 220ms

BSC/RNC protection link failure of MC-APS carrying 56 DS1 based CES each with RPWs

Single port shutdown: Failing APS link carrying ATM traffic at the BSC/RNC Traffic loss from BSC/RNC towards MPLS core 100ms Traffic loss from MPLS core to BSC/RNC 154ms Traffic loss from BSC/RNC to MPLS core 60ms

BSC/RNC protection link failure of MC-APS carrying 252 ATM VC each with RPWs

Single port fiber pull: Failing APS link carrying ATM traffic at the BSC/RNC edge Traffic loss from MPLS core to BSC/RNC 188ms

BSC/RNC protection link failure of MC-APS carrying 384 ATM IMA each with RPWs

Single port shutdown: Failing APS link carrying ATM traffic at the BSC/RNC Traffic loss from MPLS core to BSC/RNC - 105ms Traffic loss from BSC/RNC to MPLS core- 100ms

BSC/RNC node failure for UMTS/GSM MBH 636 ATM/IMA, 252 ATM VC and 56 CES each with RPWs

Traffic loss from MPLS core to BSC/RNC 1.24 sec Worst case traffic loss 1.2ms

HA SF/CPM failover for UMTS/GSM RAN AR ATM/IMA, ATM VC and CES

Traffic loss from BSC/RNC towards MPLS core 0.42 sec

The setup for the adaptive timing test is shown in figure 5. A Stratum 1 clock was used as a reference clock into SUT-A for CES packet generation and into the ANT-20 for comparison to the adaptive CES clock. SUT-B was configured to use the adaptive clock recovered from the CES for the as its central clock reference. The network included four IP/MPLS hops across the PSN. The ANT-20 tester was set to run a wander measurement of the adaptive timing at SUT-B. The adaptive clock wander test measurements included time interval error (TIE), maximum time interval error (MTIE), and time deviation (TDEV), as defined by the ITU-T G.810 recommendation.

This section describes the evaluation of Alcatel-Lucents adaptive clock recovery and differential timing implementation for CES services. Using the CES setup as a part of the GSM/UMTS/HSPA evaluation described as above, a JDSU ANT-20 advanced network tester was used to perform 24 hour wander measurements on separate adaptive and differential timing setups.

3.2 EVALUATION OF NET WORK SYNCHRONIZATION TECHNIQUES

Isocore Internetworking Lab

17

Figure 5: Multi-hop CESoPSN PWs setup used for adaptive clocking wander measurements

Figure 6 below shows the results of the wander measurements on the adaptive recovered timing (clock). In order to pass the test, the MTIE and TDEV measurements are expected to stay below the respective MTIE and TDEV mask shown in the chart. The actual test results showed that the measured MTIE and TDEV results stayed well under the ANSI/T1.101 mask. This comprehensive performance measure assures proper clock recovery and synchronization of the network when deploying for adaptive clock recovery using CES across a packet-switched network (PSN).
Figure 6: MTIE analysis for recovered adaptive clock

18

Isocore Internetworking Lab

The setup for the wander measurements of differential clocking is shown in figure 7. For this test a TDM multiplexer was added to provide an offset between the STM1 and embedded E1 as is expected in real SDH/SONET networks. The differential test included a 7710 SR-c12 as SUT A and a 7750 SR7 as SUT B. Each node was configured with an OC3/STM1 CES port with SDH framing and E1 channelization. The MUX, SUT A and SUT B used an external Stratum 1 reference clock and ANT20 used an offset from this Stratum 1 reference clock. SUT A extracts the embedded E1 from the STM1 and applies a timestamp to the RTP header which includes the difference between the reference clock (GPS) and the received E1 clock (GPS+offset). SUT B retrieves the timestamp from the RTP header and, using the offset and its local reference clock (GPS) regenerates the original clocking from the ANT-20.
Figure 7: Setup used for the differential clock wander measurement

The differential wander test was run for 24 hours and MTIE was measured. The results graph in figure 8 shows that the MTIE stayed well below the mask and successfully passed the ITU-T G.8261 criteria. This comprehensive performance measure verifies that the differential timing implementation using CES was able to adjust the differential clock source to match the reference clock at SUT-B. This is essential for CES services to be support across endpoints with different timing references.

Isocore Internetworking Lab

19

Figure 8: MTIE plot against ITU-T G.8261 Case 1

The Agilent N2X was a critical tool used to support the analysis requirements for this event. It wasnt enough just to spot-measure the latency or packet loss on a subset of streams this event required being able to monitor and measure latency and packet loss for any of the more than 10,000 traffic streams instantaneously present during the testing. The ability to sort them all by peak latency or loss was critical to the extensive real-time analysis performed during this event.

The 7750 SR-12 was a part of every convergence and performance measurement made during the evaluation and testing. The scalability and performance of the system was nothing less than impressive.

The Alcatel-Lucent 7750 SR-12 and Agilent N2X were used to simulate all of the Node-B and BTS equipment, DACS equipment, BSC/RNC functional blocks and all related traffic load and analysis for this event. Figure 9 represents the setup of these platforms and presents the scaling numbers that were applied to the 7750 SR-12. Although references to the mobile elements (base stations, controllers, switches) were used throughout the report, Isocore feels it is important to present the details of what was involved to perform the simulation for this evaluation. Although the 7750 SR-12 was not considered a primary SUT and the same can be said about the 7450 ESS, their performance was vital to the success of the tests.

3.3 SIMUL ATION OF BASE STATIONS (BT S, NODE-B) AND RADIO CONTROLLERS BSC/RNC

20

Isocore Internetworking Lab

Figure 9: Alcatel-Lucent 7750 SR-12 simulation node and Agilent N2X

Isocore Internetworking Lab

21

4 CONCLUSION
Isocores evaluation of Alcatel-Lucents IP/MPLS MBH solution is the first of its kind in the industry and the most comprehensive third party validation of IP/MPLS mobile backhaul technologies ever conducted. The event is significant in the scope and breadth of technologies that Alcatel-Lucent was able to support on a single test bed. The platforms under test supported two highly-scaled MBH environments CDMA/EVDO and GSM/UMTS/HSPA concurrently. The fact that the evaluation considered many technologies including pseudowire redundancy, multilink bundles, ATM/IMA, ATM/VC, QoS for MLPPP bundles, in parallel to scaled IP MBH, IP/MPLS MBH and CES MBH on a single product set is commendable. Isocore put the entire test environment and the many technologies tested under considerable scrutiny to ensure the accuracy of observations made.

The high scalability of the Alcatel-Lucent 7750 was confirmed by the system under test successfully aggregating traffic from 1650 MBH sites (base stations). But more imperative were the number of DS1/E1 circuits supported per site. Each of the 1650 emulated base stations had an average of 2.8 DS1/E1s circuits per site. This is significant in the amount of bandwidth that can be allocated per base station in real-world deployments and in meeting a key requirement of next generation MBH networks. Even though the 7750 SR12 was not considered a primary SUT, the scalability of the platform was crucial in simulating the entire Node-B, BTS equipment, and BSC/RNC functional blocks. With over 4800 DS1 and E1 terminations on a single 7750 SR12, the performance and stability of this node was vital to the success of the test event. Through the execution of large numbers of failure scenarios and consistently low recovery times Isocore concludes that Alcatel-Lucent resiliency implementation is comprehensive and is end-to-end. The ability to protect the stateful MLPPP sessions and implementation of pseudowire redundancy supporting the MC-APS for faster convergence was distinctive. Similar to the scalability results, the findings of resiliency tests were consistent and passed all the test limits set forth by Isocore.

With the consistency of results observed throughout the entire effort, Isocore concludes that the tested Alcatel-Lucent META architecture, related platforms and feature-set is reliable and the solution is deployable for supporting end-to-end mobile backhaul services for CDMA/EVDO and GSM/UMTS/HSPA. The scope of the evaluation and breadth of technologies verified in this validation confirms the readiness of the solution for carrier-class deployments. Isocore feels confident to say that the Alcatel-Lucent platforms and features related to IP/MPLS MBH are complete, scalable and robust.

Throughout the entire test cycle, Isocore placed emphasis on assuring that all executed test cases and the evaluated feature-set were relevant to real-world deployments. To fulfill this requirement, many tests were repeatedly carried out to ensure the consistency of the results. Using the same methodology, many of the multiclass traffic-verification tests were conducted for long duration, with several overnight runs to monitor the traffic loss and the system stability. In all instances, the observed results were consistent. Concerted effort was placed on verifying deployable circuit emulation services by extensively evaluating network synchronization using adaptive and differential clock recovery implementations. There was no observed traffic loss in any of the tests performed for measuring the clocking accuracy for adaptive and differential clock recovery. Throughout the tests, the Alcatel-Lucent CES showed stable synchronous, adaptive, and differential timing.

22

Isocore Internetworking Lab

5 LIST OF ACRONYMS USED


3G1x ACR AR BSC BTS BW CDMA CES DS1 DUT EV-DO GSM HA HSPA IMA IOM L2 L3 MBH MC-4 MC-APS MLPPP MLS MM MPLS MSC MTIE Node-B PPP PS RAN RNC SR SAR SF/CPM SUT TDEV TIE UMTS VC Third Generation Wireless Technology Adaptive Clock Recovery Aggregation Router Base Station Controller Base Transceiver Station Bandwidth Code Division Multiplexed Access Circuit Emulation Services Digital Signal 1 Device Under Test Evolution Data Optimized Global System for Mobile communication High Availability High Speed Packet Access Inverse multiplex for ATM Interface Layer 2 Technologies Network Layer 3 Technologies Mobile Backhaul Multi Class 4 Multi-Chassis Automatic Protection Switching Multilink PPP Multi Layer Switch Mobility Manager Multi-Protocol Label Switching Mobile Switching Center Maximum Time Interval Error BTS in UMTS Environment Point to Point Protocol Packet Switch Radio Access Network Radio Network Controller Service Router Service Aggregation Router Switch Fabric/Control Processor Module System Under Test Time Deviation Time Interval Error Universal Mobile Telecommunication Systems Virtual Circuit

Isocore Internetworking Lab

23

I S O C O R E

T E C H N I C A L

R E P O R T

About Isocore
Isocore provides technology validation, certication and product evaluation services in emerging and next generation Internet and wireless technologies. Isocore is leading validation and interoperability of novel technologies including MPLS, Carrier Ethernet, IPv6, IP Optical Integration, wireless backhauling and Layer 2/3 Virtual Private Networks (VPNs), IPTV service deployment architecture validation, and certication of other provider backbone technologies. Major router and switch vendors, Service Providers, and test equipment suppliers participate in Isocore activities. Isocore has major ofces in the USA (the Washington DC area), Europe (Paris, France) and Asia (Tokyo, Japan).

12359 Sunrise Valley Drive, Suite 100 Reston, VA 20190 Phone: 703.860.1777 Fax: 703.860.1778 www.isocore.com
Isocore. All rights reserved For more information about the testing capabilities of the Internetworking lab, please email testing@isocore.com. This document should not be reproduced as a certification or a validation reference from Isocore. Reproduction of this document to a third party requires written approval from Isocore Corporation.

Вам также может понравиться