Catalyst 6500 Series with Cisco IOS Software Modularity

1 EANTC TEST REPORT: Catalyst 6500 with Cisco IOS Software Modularity --- Page 1 of 6 Catalyst 6500 Series withCisco IOS Software ModularityFunctionality and Feature Verification TestsIntroductionAt the end of August 2005, Cisco Systems announcedCisco IOS Software Modularity for the Catalyst 6500 Series forwarding of mission-critical traffic isbecoming increasingly important in triple-play (voice,video, and data) networks and for new interactive userservices. In addition to high availability designs whichinclude redundant systems, the highest level of soft-ware resiliency is required everywhere in the the core and distribution to single points offailure such as the wiring closet, data center accessand WAN edge of an enterprise network, softwareresilience allows strict end-to-end service level agree-ments to be IOS Software Modularity for the Catalyst 6500exactly addresses these high availability enabling modular Cisco IOS subsystems to run inindependent processes, unplanned downtime is mini-mized through self-healing processes, softwarechanges are simplified through subsystem In-ServiceSoftware Upgrades (ISSU) and automated process-level policy control is enabled by integratingEmbedded Event Manager (EEM).

2 Venue & Test EquipmentCisco commissioned the European Advanced Networ-king Test Center (EANTC) to verify the migration pathand functionalities of the new IOS release offeringSoftware Modularity . The tests were conducted atCisco s test lab in San Jose, California in August2005. EANTC test engineers performed all tests toverify the expected feature functionality and in parallelEANTC extensively tested the first shipment ofthe Cisco IOS Software Modularity enhance-ments to the release for the Catalyst6500 Series switches. Different maintenanceactions from migrating to the new Cisco release to real life operating taskswere analyzed for ease of use and double-checked against the high availability claims ofCisco. We found a system consistent in allareas tested with the previous IOSversion and did not see, any packet lossduring our tests thereby validating Cisco sclaims for this new technology on the Catalyst6500 by2005 Test Highlights Migration to a new releasewith Cisco IOS Software Modularitywent smoothly and guaranteedoperational consistency (CLI,configurations and look & feel )with the previous IOS New installer feature to simplifyimage and patch management waseasy to use Fault transparency with zeropacket loss was observed due tothe stateful process restart featurewith memory protection and faultcontainment No service disruption withsubsystem In-Service SoftwareUpgrades (ISSU) and Non-StopForwarding (NSF)

3 Zero packet loss while applyingpatches to address recreatedvulnerabilities Intuitive patch management process that includes tagging,patch rollbacks and repackaging ofpatched images Easy to use multi-purposeEmbedded Event Manger (EEM)features for diagnostics and auto-mated policies based on TCLscriptingEANTC TEST REPORT: Catalyst 6500 with Cisco IOS Software Modularity --- Page 2 of 6continuouscontrol and dataplanetrafficforwarding. Thetesting environ-mentincludedSpirent sSmartBitsloadgeneratorincombinationwith TeraRoutingTester Test Setup and MethodologyThe test bed, seen in the following graph, consisted ofthree Cisco Catalyst 6504-E chassis. One was ourDUT (device under test) and the other two 6500s wereused to provide NSF (Non-Stop Forwarding) aware-ness in the test bed. At the start of the test all threedevices ran Cisco IOS Version (18)SXE2 which isgenerally available on Cisco Connection Online(CCO).

4 During the test, the DUT was upgraded to apre-FCS (First Customer Shipment) release (18)SXF1, the first release for the Catalyst 6500series planned to offer Cisco IOS Software 1 represented the service provider edge,receiving 180,000 E-BGP routes from the SmartBitsload generator. Switch 3 represented an enterpriseCatalyst 6500 switch and was fed by the SmartBitswith 10,000 OSPF routes. The DUT (Switch 2), config-ured between these two switches and set up for BGPand for OSPF, held a total of 190,000 routes. Bysetting up full-meshed traffic between the announcedroutes, the test bed forwarded a total of milliondifferent flows. Two GigabitEthernet ports were usedand a load of 95% was configured bi-directionally toallow for the exchange of control plane traffic,resulting in the DUT forwarding in total million64 Bytes frames per second in each direction.

5 Thesereference streams showed no packet loss and saverage end-to-end detailed configuration of the test bed was:Switch1: Supervisor Engine 720 with Policy Feature Card(PFC) 3B 24 port 1 GbE SFP, WS-X6724-SFP, with aCentralized Forwarding Card, WS-F6700-CFCS witch 2: Supervisor Engine 720 with Policy Feature Card(PFC) 3 BXL 16 port 1 GbE GBIC, WS-X6516A-GBIC, and a 48port 10/100 Mb RJ45, WS-X6348-RJ-45 Switch 3: Supervisor Engine 720 with Policy Feature Card(PFC) 3A 24 port 1 GbE SFP, WS-X6724-SFP, with aCentralized Forwarding Card, WS-F6700-CFCS witch 1 NSF AwareSwitch 3 NSF AwareSwitch 2 DUTBGPOSPFTFTPWe bSNMPDNSNTPL ogical Test SetupPhysical Test Setup in the Cisco LabsCatalyst 6500 Data PlaneNetwork Optimized MicrokernelBaseRoutingTCPUDPEEMFTPCDPINE TDetcBaseRoutingTCPUDPEEMFTPCDPINETDetcH igh Availability InfrastructureCisco IOS Software ModularityCatalyst 6500 withCisco IOS Sofware ModularityEANTC TEST REPORT.

6 Catalyst 6500 with Cisco IOS Software Modularity --- Page 3 of 61) Migration to IOS with SoftwareModularityTest ObjectiveThe aim of this test was to show the seamless migration path from aprevious release to a new Cisco IOS release withSoftware Modularity , to verify CLI (command line interface) andconfiguration consistency of the new softwarerelease, to present the installer feature, which is used fornew operations of Cisco IOS with SoftwareModularity and to show the new Cisco IOS enhancements in theform of modularized run-time MethodologyWe started this test by downloading the new softwarevia FTP to the Supervisor Engine of the DUT. Afterreconfiguring the bootstring to point to the new image,the DUT was reloaded to run the new Cisco IOS in thewell known binary mode, which is a single imagethat is not expanded into a file system.

7 At this time, afirst verification of the DUT configuration and CLIconsistency was order to be able to apply patches to softwaresubsystems, we then changed from the binary to the installed mode configuration. In the installed mode, the system runs from a file system, either boot-disk: or disk0/1: . By using some show commands ( show install running or showinstall disk0:/sys ) we verified the new (DUT)#sh install runningSoftware running on card installed at locations72033_rp - Slot 1 :B/P C State Filename--- - -------- -------- B Active disk0:/sys/s72033_rp/base/DRACO2_MPSoftw are running on card installed at locations72033 - Slot 1 :B/P C State Filename--- - -------- -------- B Active disk0:/sys/s72033/base/s72033-adventerpr isek9-vm( (20050810:214544))LEGEND:-------:B/P - (B)ase image or (P)atch C - (C)ommittedPruned - This file has been pruned from the systemActive - This file is active in the systemPendInst - This file is set to be made available torun on the system after next - This file is set to be rolled back afternext - This file will run on the system afternext reloadRollPRel - This file will be removed from thesystem after next reloadRPRPndIn - This file is both rolled back pending areload, and pending installation.

8 On reload, thisfile will not run and will move to PendInst install activate is done before reload,pending removal and install cancel each other andfile simply remains activeIPRPndRo - This file is both installed pending areload, and pending rollback. If the card reloads,it will be active on the system pending a roll-back. If install activate is done before areload, the pending install and removal withcancel each other and the file will simply beremovedTo analyze the new process architecture, we issuedcommands which were enhanced compared to theprevious Cisco IOS release, like show process and show process cpu , and we observed more than 20modular the different stages of these first tests, likedownloading or installing the OS, we verified thepacket forwarding performance to show zero packetloss and no increase in Test AnalysisAs expected, only a few small steps were necessary tomigrate to the new release with Cisco IOSS oftware Modularity on the Catalyst 6500 .

9 The testedCisco IOS Software came up with the same look & feel(CLI) as the previous Cisco IOS version and the switchconfiguration remained exactly the same during thischange (apart from the boot string). With the excep-tion of the system restart to load the new image, nopacket loss or increased latency could be Highlights Seamless migration to the new releasewith Cisco IOS Software Modularity Consistency with previous Cisco IOS versions(CLI, configurations and look & feel ) Easy to use installer featureEANTC TEST REPORT: Catalyst 6500 with Cisco IOS Software Modularity --- Page 4 of 62) Restarting Processes and ShowingNSF FunctionalityTest ObjectiveThe aim of this test was to show that modularprocesses like the routing process, covering BGP andall IGPs such as OSPF, can be restarted without packetloss in the data MethodologyAs a precondition of this test, the BGP and OSPF neighbors of the DUT had to be aware of the NSFprotocol, avoiding route flaps due to the failing and after the manual restart of the routingprocess , we observed the detailedprocess status with the show process.

10 In parallel to the process restart,we sent the reference test traffic to observe correctNSF (DUT)#sh processes Job Id: 69 PID: 24619 Executable name: Executable Path: Instance ID: 1 Respawn: ON Respawn count: 1 Respawn since last patch: 1 Max. spawns per minute: 30 Last started: Thu Aug 18 17:11:32 2005 Process state: Run Feature name: iprouting Core: SHAREDMEM MAINMEM Max. core: 0 Level: 100 Mandatory: ON Last restart userid: Related Processes:PID TID Stack pri state BlkedHR:MM:SS:MSEC FLAGS NAME24619 1 32K 10 Receive 10:00:13:0248 00000000 2 32K 10 Receive 10:00:00:0000 00000000 3 32K 10 Receive 10:00:26:0344 00000000 4 32K 11 Nanosleep0:00:00:0000 00000000 5 32K 10 Receive 10:00:35:0248 00000000 6 32K 10 Receive 10:00:00:0000 00000000 Test AnalysisAs expected, we observed no packet loss during theprocess restart.