product hiper ring vs. rstp - interlog-it.com · product analysis: hiper ring vs. rstp the...

PRODUCT

ANALYSISHiPER Ring vs. RSTP

Redundancy Process withHirschmann Switches

Product Analysis:HiPER Ring vs. RSTP

Redundancy Process withHirschmann Switches

Markus SchaubHartmut Kell

Product Analysis: HiPER Ring vs. RSTP

Copyright © 2003 by ComConsult Technologie Information GmbH

All rights reserved. No part of this report may be reproduced or transmitted in any form or by any means, electronic or

mechanical, including photocopying, recording, or by any information storage and retrieval system, without written permission

from the publisher, except for the inclusion of brief quotations in a review.

Marketing and translation into other languages are subject to prior written permission by ComConsult Technologie Information

GmbH

The authors and the publisher have made every effort to make this document error-free and as accurate as possible. No

warranty or fitness is implied.

The information is provided on an “as is” basis. The authors and the publisher shall have neither liability nor responsibility

to any person or entity with respect to any loss or damage arising from the information contained in this book.

Reproduction of the product names, brand names, trademarks, service marks, trade descriptions, etc., even without special

marking, should not be regarded as affecting the validity of such names or marks. The authors and the publisher adapt

their way of writing in major part to that of the producer, but cannot attest to the accuracy of this information.

ComConsult Research represents the product and technology test laboratory of ComConsultTechnologie Information GmbH.

Copyright of the English edition: © 2003 by ComConsult Technologie Information GmbH.

First edition dated 11th November, 2003

Feel free to send your comments and questions to:

ComConsult Technologie Information GmbHPascalstraße 25 52076 Aachen

+49 (2408) 955 400 +49 (2408) 955 399

E-Mail: [email protected]


Copyright © 2003 by ComConsult Technologie Information GmbH

MOTIVATION AND SUMMARY 1–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

HIRSCHMANN PRODUCTS IN THE TEST 3–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

RAIL-Switches 3

MICE-Switches 3

MACH 3000 Switches 4

INDUSTRIAL SUITABILITY 5–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

INSTALLATION AND MANAGEMENT 9–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Menu-driven User Interface 9

WEB Interface 9

Network Management Tool HiVision 10

Unit Exchange with the Auto-Configuration Adapter (ACA) 12

Modularity: Addition/Exchange in Operation 12

Help Functions 13

Summary 13

REDUNDANCY PROCESS 14–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

RSTP 14

Hirschmann Process 14

TEST OF THE REDUNDANCY PROCESSES 19–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Test Processes Applied 19

HiPER-Ring 19

Ring coupling 24

Dual Homing 25

HiRRP 26

HiPER Ring vs. RSTP 27

SUMMARY AND DISCUSSION OF THE RESULTS 31–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––


Copyright © 2003 by ComConsult Technologie Information GmbH Page i

Index

For more than 10 years, Ethernet has beenestablished in master computer systems bothin production and industry environments.Nowadays the availability of modern switchsystems for direct production environmentsmore and more often tends to generate pro-jects where the traditional bus systems in-stalled underneath the master level, such asthe profi bus, are replaced by Ethernet aswell.

This product test is designed to analysewhether Ethernet switch systems are in factsuitable for this application case. As testproducts the Ethernet switch products ofMessrs. Hirschmann Electronics GmbH & Co.KG are used. For more than 15 years, thenetwork products produced by Hirschmannhave successfully been in use in this domain.Still, the question remains to be answeredwhether these products enable an applicationdirectly involved in data processing of thiskind.

The test is focussed on the following majorissues:

1. Suitability for industrial surroundings

2. Easy handling.

3. Reaction in cases of failure in the network.

The third item is of special significance heresince Hirschmann applies a prioritary methodby installing the HiPER Ring. The HiPER Ringis not only limited to Hirschmann componentsbut presents (almost) industrial standardwhich is also applied by Siemens, SchneiderElectric and Phoenix Contact. The reason forthis divergence from the usual standardizedRapid Reconfiguration Spanning Tree is itsbetter suitability regarding technical andproduction requirements. Exactly here will ourtest set in by drawing a comparison betweenthe performances of the HiPER Ring and theRSTP method when located within the typicalscenario.


Our test will arrive at the following results:

1. The requirements to be met by networkproducts in industrial environments maysignificantly vary from the conditions ofoffice environments. Even if in practiceonly part of the production is faced withtough environmental conditions the pre-requisite of uniform installations pre-scribes the use of products which arecapable of fulfilling such requirements.The Hirschmann products tested herehave been designed for this special ap-plication and they outdo the perform-ances of comparable products used inoffice environments by far.

2. Both the web operation surfaces of thetested products and the ManagementTool HiVision have been structuredclearly and they have been provided withvery good documentation. Even some-one not working with it every day will befamiliar with it very soon and the helpfunction will offer extensive explanationson the configuration para-meters.

3. Under certain conditions, the HiPER-Ring method is by far superior to thestandardized RSTP method. Here theimplementation used in the test in com-parison with the RSTP method byCISCO must be regarded as a very goodimplementation. In the lab, other imple-mentations were tested which gave con-siderably worse results. The Hirschmannprocess has always got an advantagewhen a chain of systems has to beswitched to it in a row or in a circle. Thefailover time is independent of thenumber of switches used, while, with theRSTP method, down time and number ofswitches are closely interrelated.

Copyright © 2003 by ComConsult Technologie Information GmbH Page 1

Motivation and Summary

Picture 1: Increase of the failover time in relation with number of switches

Conclusion:

In many cases, network products which are designed for office environments will not meet therequirements set by industrial environments. The Hirschmann products tested in this testreveal the differences between industry and office products. They fulfil the highly demandingrequirements of data processing in production environments.



0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

3 4 5 6 7 8

HiPER Ring RSTP



Hirschmann Products in the TestRAIL-Switches

Picture 2: RAIL Switch

MICE-Switches

Picture 3: MICE Switches

The case of the RAIL switch series consistsof compact and firmly installed devices forindustrial application. There is a great numberof different systems, regarding both thenumber as well as the type of ports available.

The switches are supplied with 24 V DC and,for redundancy purposes, can be configuredwith two independent power supplies. TheRAIL switches support both the HiPER Ring(for this purpose two ports have been fore-seen) and the redundant ring interconnection.For the latter, a fixed standby port has alsobeen pre-configured which cannot be used forthe data packets either.

The functions Ring Manager and StandbyManager are configured by DIP switches onthe front face of the unit.

The latest generation of Hirschmannindustrial switches is the MICE series. Incontrast to the RAIL systems, these switcheshave a modular structure. They consist of aback-plane which, alike the RAIL switches, ismounted on top hat rails, a basic unit andseveral slots for modules. All modules can becombined with all backplane variants, even ifthe construction height of the modules andthe basic unit is varying.

Alike the RAIL systems, the MICE switchesare also supplied with 24 V DC and can beconnected redundantly. In contrast to theRAIL series, however, the network connec-tions are located at the bottom side of theunit.

As redundancy functions, the HiPER-Ringand the redundant ring interconnection areavailable.

The switches of the MACH series consist ofa chassis with a passive backplane and oneor more basic boards. Each basic board isbuilt with a modular structure and thus is anindependent switch so that a central switchmodule can be dropped. This will enhancefailure security significantly since one unit willstill be able to function even if individual mo-dules stop working.



MACH 3000 Switches

Picture 4: MACH 3002

The power can either be supplied via powerpacks with 220 V or via 24 V. It goes withoutsaying that the power supply of the MACHunit is also designed on the principle of re-dundancy.

By way of one or several router modules it ispossible to change a MACH switch into aLayer-3 switch. The Layer-3 techniques thusmade available are static routing entries andHiRRP is the router redundancy solution.

On Layer 2, the MACH switches support theHiPER Ring for Fast and Gigabit-Ethernet aswell as Dual Homing.

By the end of 2003, RSTP will also be sup-ported by MACH, MICE and RS2 so as tooffer compatibility to other manufacturers.

Industrial environments differ largely fromoffice environments. In particular, in the caseof special environment conditions the appli-cation of switch products, as have been usedin local networks for years, is doubtful here.However, this question cannot be answered ingeneral since, firstly, the environmentconditions are varying considerably from caseto case and, secondly, a difference must bemade between environment condition andinstallation situation. In each case, a project-specific individual decision will be taken whichwill depend on the requirements of therespective underlying situation.

Some manufacturers offer switch systemswhich are specially designed for the industrialarea. Here the definition of suitability is notup to the manufacturer but is stipulatedclearly by various standardization guidelines.Accordingly, the buyer of industrial switch sys-tems should take these standards as a basisof decision. Since most of the propertiesstipulated in these standards cannot be veri-fied by the end customer himself the manu-facturers are also called upon to prove ad-herence to the respective standard criteria ina verifiable way.

In general, when talking about the industrialsuitability of switches the following require-ments are drawn up to fulfil these suitabilitycriteria:

• Protection against water, liquids and dust

• Enhanced thermic resistance

• Enhanced mechanic resistance

• Enhanced EMV stability

• DC voltage supply

• Data connections suitable in the industry


The requirement regarding enhanced „tight-ness” against liquids and dust is seldom,however, on the other side, with a view tomarket and product situation a k.o.-criterion.Standard IP67 provides for the highestprotection class for a completely closed unitcasing which means highest requirementsregarding both the construction of the caseitself and the thermic resistance. However, inmost application cases, a switch can rela-tively easily be mounted in a completelydense casing. The market situation vis-a-vissuch requirements does not leave anyquestions here. More than 99% of all switchsystems offered are IP20 and they aredesigned for office environments. Hirschmannis among the few manufacturers able to offeran IP67 switch.

However, mounting a switch in a closedcabinet housing will rise another, in fact verysig-nificant problem: The heat produced by theswitch can only with difficulty escape from adense housing which will cause a rise intemperature in the cabinet and in the switchit-self. When do we talk of enhanced thermicload? Here some typical operation tempera-ture ranges:

• "office suitable" switch systems can beoperated at temperatures ranging be-tween –5° C and +45° C.

• The IAONA cabling guideline defines therange between 0° C and +55° C.

• Twisted-pair installation cables can beoperated at temperatures between –20° Cand +60° C.

• Light wave conductor cables can beoperated at temperatures between –5° Cand +70° C.


Industrial Suitability

All industry suitable systems of Hirschmanntolerate an ambient operation temperature of0° C to +60° C which is significantly higherthan the switches “only” suitable for officeenvironment. If a switch would be applied inan environment with higher temperatures onewould note that standard installation cablesare no longer applicable in this environmenteither. But Hirschmann also supplies EEC(Extended Environmental Conditions) typeswith extended temperature ranges of -40°C to+70°C.

The majority of Ethernet switches or hubsinstalled today in hall operation mode is


located outside machine cabinets in field level.Here cranes, fork lifts or even the personnelwould rather present a “mechanical” dangerfor the components. This particularly appliesto wall cabinets and standing cabinets whichare often installed in unfavourable positions.In a closed distribution cabinets or in amachine mounted in field level one has to takeinto account both massive mechanicaldamage and strong and high-frequencyvibrations. Especially affected by suchdamage are plug connections, adapter cardsand optionally mounted switches orhubs/converters.


Table 1: Definition of IP-Protection Class

Water Protection

No particular protection

Against vertically fallingdripping water

Against diagonally fallingdripping water (up to 15°deviation from theperpendicular line

Against water spray (anydirection with up to 60°deviation from theperpendicular line)

Against water spray fromall directions

Against water jets from anozzle from all directions

Against flooding

Against submersion

Against total immersion

SecondIndex

0

1

2

3

4

5

6

7

8

Foreign Bodies Protection

Big foreign bodiesdiameter > 50 mm

Medium-sized foreign bodiesdiameter > 12 mm

Small foreign bodiesdiameter > 2,5 mm

Grain-shaped foreign bodiesdiameter > 1 mm

Dust-protected, dustdeposits are tolerable buttheir quantity must notimpede the function of theunit

Complete dust-protected

Contact Protection

Against large bodysurfaces

Against fingers orobjects of similarsize

Against tools, wiresand similar objectsof a thickness > 2,5mm

Complete protection

Complete protection

Complete protection

FirstIndex

0

1

2

3

4

5

6

No particular protection

When specifying active components manymanufacturers of industrial switches refer, forinstance, to the following two standards (Unit“g“ is a measure for the gravity):

• EN 60068-2-6 A (requirement to vibrationresistance with 5 g at 10 to 150 Hz)

• EN 60068-2-27 (requirement to shockresistance at 15 g to 11 ms)

If you compare the market and, in particular,the Hirschmann products with thesespecifications you will find that thisrequirement is also “met“ by Hirschmann. If,for instance, in the machine environment,higher loads arise their value must be knownprecisely, and tests in cooperation with themanufacturers of the components areindispensable.

The difference of the requirements consideredto date are loads produced in the environmentby electro-magnetic disturbances (EMI). Theymust be expected at any place within theindustrial surrounding, not just in the vicinityof the machine, and there are also heavydisturbances which are difficult to calculate. Formost of the users/palners of an industrialenvironment, it is difficult to estimate what EMIdisturbances precisely have to be expected andwhat way they will take. The requirement thatcompoments should offer a high EMCresistance especially refers to minimumdisturbance resistance and maximumdisturbance emission. Disturbance compatibilityand disturbance sensibility of a communicationsystem are described by two “core standards“:

• Disturbance emission according to EN55022

• Disturbance resistance according to EN55024


EN 55022 defines the application ranges In-dustry (radiation class A) and Living/Office(radiation class B). When planning and, inparticular, when deciding on a cabling sys-tem, however, it is in most cases recom-mendable to adhere to the more “severe”radiation class B. In analogy to EN 55022,there is also a similar subdivision for EN55024 which, in most cases, demandsadherence to the more severe part 2. Thereis quite often some confusion when referenceis made to the former standards EN 50081and 50082. These are standards whichstipulate a specification on EMC resistance forall electric components in the living or in-dustrial environment. The above cited stan-dards EN 550xy were developed from thesetwo standards. However, in the domain of thedisturbance resistance a weakening of therequirements was observed.

When choosing active components one has todistinguish between those which must bemounted “EMC unprotected“ and those to beinstalled inside an additional EMC-resistantcabinet. In the first case, an adherence to theabove described standards is demanded asobligatory and to be confirmed by the manu-facturers. In the product documentation,Hirschmann confirms adherence to bothstandards.

Particularly in the immediate vicinity of themachine installation there is no 230 Valternating voltage available but 24 V directvoltage so that, in principle, space-consuming power packs could be dispensedwith. This infrastructure is also used byindustrial switches and it is nowadaysgeneral practice and standard that switchesfor industrial applications do no longer havea power pack but a terminal block forconnection to such direct power sources.


If such voltage supply is missing at the placeof installation the switch manufacturers offertop hat rail mountable power packs with differentpower per-formances. Hirschmann equip theirsystems with connections for 24 Volt directcurrent and offers compatible network partswhich can also supply a switch redundantly.

The possibility of using a data connection insidean industrial surrounding considerably dependson two prerequisites which have already beendescribed above:

• Will a switch remain dense with the dataplug connected?

• What is the reaction of the connectionslike when exposed to heavy mechanicalimpacts?

Since, as already mentioned above, switcheshardly offer a higher IP protection class thanIP20 there is little sense in demanding anespecially “dense” plug connection for suchtypes. As a consequence, the few IP67switches are exclusively manufactured withspecial data connections. For instance, in thecase of the Hirschmann IP67 switch this is aM12 plug.


When considering the mechanical require-ments one would rather expect that, in caseof vibrations or even falls, the internal switchhardware such as mother board, internalplugs, skirting board for RAM and the likewould rather be affected instead of a RJ45plug which, once plugged-in and firmlyseated, will not be dislodged by such loadsand impacts.

In the end, an analysis regarding thequestion what plug should be used will leadto the re-sult that the extremely high numberof indi-vidual cases will try to cope with theplug technology traditionally used in officesur-roundings such as RJ45, ST or SC . If, inspecial cases, particularly high requirementsshould require “better” plugs, in view oftoday’s state of standardization proprietarysolutions will be indispensable. Hirschmannis one of the very few manufacturers who areable to offer IP67-suitable connection tech-niques with improved mechanical propertiesfor a certain type (FE-IP67-4TX with fourFast-Ethernet suitable copper connections).

In the near future, Hirschmann will launchanother IP67-switch on the market, i. e. theOCTOPUS 5TX with five Fast-Ethernet suit-able standard-compatible M12 connections.




Installation and ManagementMenu-driven User Interface

The putting into operation of all testedHirschmann components was at first effectedin the same way. By means of a specialcable the serial COM-Port of the computer

is connected with a V.24 of the respectiveswitch. Subsequently, it is possible to set upa first connection by means of a hyper-terminal programme.

The user interface is completely menu-driven, and, unlike products of many othermanu-facturers, there is no command lineinterface. It is primarily designed for the firstconfigura-tion since not all functionalities canbe con-figured from here.

WEB Interface

JEach of the systems tested is equipped withan integrated http server by means of whichconfiguration of the switches via Browser caneasily be effected. Precondition for anaccess to the Browser is that an extra Java

Picture 5: Configuration via hyper-terminal

environment of SUN has been installed onthe respective computer in addition to theBrowser which can be downloaded from theSUN-Homepage or copied from the CDdelivered free of charge with the product. Incontrast to the user interface, with the WEBinterface it is possible to completelyconfigure all functionalities of the systems.

After call-up of the IP addresses of the switcha Java applet is transferred from the switchto the PC. If you dispose of many systems ofthe same type and with the same softwareversion it is reasonable to pre-install theapplet on a management station as well soas to save downloading time.

An argument in favour of this process is thespeed of the applets which can becharacterized as extremely fast compared tosimilar applications of many othermanufacturers. Besides the surface hasbeen designed very clear and user-friendlyand there are numerous and detailed helpfunctions. The surface and the help menucan be loaded either in German or English.Apart from a separate help function at thebottom of the selection menu (see Picture 6),most of the functions are also documenteddirectly and in a very detailed way in thewindow where the adjustments are made.From the help function it is again possible tojump to the configuration of the respectivefeatures.

Apart from pure configuration, the differentavailable stati of a system can be inquiredand indicated, such as voltage supply, inter-face stati and counter, port stati of the HiPER


Ring ports. Serious errors are already indi-cated, right after login, on the entry page inthe “Alarm“ field. In the example shown here,the alarm is marked in red and a failure ofone of the two voltage supply sources is sig-naled. What alarms are reported and in whatcolour (green, yellow or red) can be config-ured individually.

Besides, it should be noted that, apart fromsome basic adjustments, the configurationchanges at Hirschmann will become activeimmediately but, after a new start, will remainvalid only if explicitly stored.

Network Management Tool HiVision

Picture 7 shows the survey page of theprogramme. At first sight you can see whichsystems are included in the system. Besidesyou get further important information on the


Picture 6: Access via WEB Interface

stations managed. Thus you can recognizederrors immediately by the red ”Smilies“ andsearch for the reasons.

In addition to the survey, this entry mask offersthe possibility of managing VLANs and events


reported by the switches are gathered and listedin a table.

On the basis of this mask you can configure theswitches individually but a configuration of severalsystems at a time is possible, too.


The drawback of a configuration via WEBinterfaces consists in the fact that only oneunit can be configured at a time. However,often one has to proceed at adjustmentswhich are to be valid for a number ofsystems; for instance, you want to configurethe trap receiver for SNMP messagesnetwork-wide or you want to implementVLANs over many switches. Here thenetwork management Tool HiVision isparticularly useful. If you use HiVision forconfiguration all adjustments can be effected

alike in the WEB Interface. The currentstatus is presented in a clear way by meansof a photo-realistic picture. Thus defectiveports or errors committed by the agent canimmediately be recognized from the colour.

If you configure several systems at a timeyou can just use those parameters which areavailable on all systems selected.

Picture 7: Network Survey with HiVision

Management of this tool can be learned wi-thin a short time provided, however, you arefamiliar with the units and network protocolsused.

Unit Exchange with the Auto-Configuration Adapter (ACA)

To facilitate the unit exchange so that staffmembers who normally do not work with net-works will also be able to exchange switchesHirschmann has introduced the Auto-Configuration-Adapter. It is a “stick“ withflash memory which can be inserted into theV.24 console port.

The ACA functions as follows:

• There is an ACA for each switch as onlyone configuration can be stored per ACA.


• As soon as the configuration or recon-figuration of a switch has been completedthe ACA is plugged onto the switch and theconfiguration is stored locally. The switch isable to recognize by itself that an ACA wasinserted and stores his configuration, bothin the local configuration list and on the ACA.

• If a switch is exchanged the ACA is insertedinto the new switch before the unit is turnedon. Now the new switch recognizes the ACAand loads the configuration stored thereon.With industrial switches of the RAIL andMICE series, it might be required to adaptthe position of the DIP switches of the ex-changed unit as well.

It is a precondition for the application of aACA that the new and the exchanged switchare identical. Furthermore, the softwareversions must be identical, too.

An advantage of this method consists in thefact that you can exchange a switch withouthaving to access to the switch itself via thenetwork. In case of a software update or aconfiguration modification the data areloaded into the ACA anew.

Modularity: Addition/Exchange inOperation

When it comes to failure security, one alsohas to consider the question regarding theexchange of units and whether it is possibleto exchange modules during operation.

A uniform answer cannot be given as thisdepends on the components selected.

Of course, the RAIL series cannot beexchanged without downtime of the switchconcerned since this system is composed offirmly structured units.


Picture 8: Unit Presentation of HiVision

This is different for the MICE systems: an in-dividual module can be exchanged duringoperation without having to switch off theentire switch provided that the module isidentical or that the module is operated asplug’n-play. Thus it is recommendable for thespare parts management to have at least onepiece of all modules used in store.

In case of the MACH switches 3002 and3005, it is possible to exchange the individualbasis boards during operation but not themodules on the basic boards.

Help Functions

Here it should be mentioned that the entirehelp system of Hirschmann as well as thedocumentation are of excellent quality. Bothis available in English and German and bothversions are up-to-date. With the exceptionof the text-based user interface via consoleor telnet, the online help is not only verydetailed but supplies extensive informationon the techniques applied. Handling iscomfortable. In HiVision, there is thepossibility (in the form usual for manyWindows programmes) to call the helpfunction by way of the menu bar. In the caseof the WEB interface, the help function islocated in the same mask where theadjustments can be effected, too. Further-more, one has the possibility of activating thehelp function by means of the left menu bar.

The documentation goes beyond the purehelp function: most of the chapters have ashort introduction into the respective problemso that the reader will quickly becomeacquainted with the general subject and doesnot get the feeling to be left alone with justthe pure help description. The documentshave been subdivided according to subjectsso that the reader easily finds what he islooking for.


Summary

• Due to missing functionalities andrelatively complicated implementation, theuser interface is only recommendable forthe commissioning of the Hirschmannswitches.

• For a limited number of systems, theBrowser Interface is sufficient.

• For medium and large-size networks,HiVision should be applied since itfacilitates considerably both handling andmonitoring.

• The use of ACAs is reasonable if you wantto set up a network which also functionswithout the continuous presence of skilledpersonnel. Likewise, it is possible that, incase of maintenance work, standstill timescan be reduced to a minimum which is anabsolute must, especially in the industry.However, one should take into accountthat the network personnel should showthe necessary discipline to keep theconfiguration stored on the ACAs alwaysup-to-date.

• Both help function and documentation area success: the theoretical knowledgerequired is in most cases introduced byway of a short comment and theexplanations on configuration parametersare informative and easy to understand.


RSTP

It would go beyond the scope of this productreport to describe the rapid spanning tree indetail. Instead, we recommend the lecture ofthe ComConsult Report by Petra Borowkaand Markus Schaub („Design-VariantenLokaler Netzwerke im Vergleich“, only inGerman) with examples of applications or theInsider article published by Markus Schaubin December 2002 („Der schnelle Baum”,only in German).

An essential characteristic of this process isthe fact that it does no longer work on a timebasis alike the classic spanning tree but byway of actions which are triggered on anevent-controlled basis. If a line fails theRapid Spanning Tree Process canimmediately react to this failure and thuskeep the failover time to a minimum. Afterrebuilding of a new, stable topology allswitches of the tree are instructed to cancelthe necessary entries of their bridge table.

Three peculiarities of the RSTP should beunderlined as they could retard the end-to-end communication considerably:

1. A switch must recognize a link failure or alink activation by way of a link-up or link-down. If, for instance, a media converter isswitched between two switches, whichdoes not transfer the link status, it will takesome time for the failure to be detected, i.e. until non-arrival of the BPDU has beenreported.

2. Not all manufacturers are using event-triggered BPDU packets to circulatechanges but use the ones regularly sentout. This, however, in the worst case,might take 12 seconds.


3. For the topology change messages oftenthe regular BPDUs are sent so that here,too, it might take relatively long until allswitches have received the information.

Furthermore, it should be noted that, with therapid spanning tree, frames may be doubledor changes in the order of the frames trans-mitted from the sender to the receiver maybe produced by a switch-over.

Hirschmann Process

The core of the processes developed byHirschmann is the HiPER Ring. The ideabehind is that a complete network can besubdivided into rings which are inter-connected with one another on a redundancybasis.

Picture 9 shows such a design of differentinterconnected ring networks: the backboneand the individual control levels are set up bymeans of the HiPER Ring; the rings them-selves are either connected with one anotherby means of the new HiPER Ring coupling(brown lines in the picture) or the olderprocess, i. e. Dual Homing (red lines in thepicture).

Due to the combination of couplingprocesses and rings a typical topology will becreated which connects a backbone ring tothe control rings by a star-shaped structure.As both in each individual ring and in theredundant ring coupling one line isdeactivated a loop-free topology is created.


Redundancy Process



MICE 1

RT2-TX/FXRS2 2RS2 4 RS2 1RS2 3

Standby ManagerMICE 2

Redundancy ManagerMICE 3 MICE 5MICE 4

Redundancy Manager

MACH 4

MACH 2Redundancy Manger

MACH 1

MACH 3

Backbone

MICE 1 MICE 2Redundancy Manager

MICE 3 MICE 5MICE 4

RT2-TX/FXRS2 2RS2 4 RS2 3 RS2 1

Picture 9: Design with Hirschmann Process

HiPER-Ring

Picture 10: Monitoring of a HiPER Ring



In each ring, there is a station calledredundancy manager. To avoid a loop theredundancy manager will interrupt the ring asfollows: data frames received on a passivering port, which belongs to the ring, will nomore be processed nor will they sent out fromthe port (red-dashed line in the picture; theport of MICE 2 in the direction of MICE 1 isthe “passive ring port“).

For the monitoring of the ring, the redundancy

manager will send out, every 100 ms, socalledwatchdog packets in both directions of the ring(see Picture 10), and, of course, also to theport which stopped processing incomingframes. If three watchdog packets in one ofthe two directions get lost the redundancymanager will recognize an interruption of thering and activate the passive port. Thus againa completely connected topology will begenerated despite failure of a line or a unit.

Picture 11: Error removal in the HiPER Ring

Following activation of the blocked port bythe redundancy manager the latter will go onsending watchdog packets so as torecognize the rebuilding of the startingtopology as well. If, as shown in the example,the defective line between RS2 2 and RS2 3is replaced the watchdog packets will againbe able to pass the line and the redundancymanager will then deactivate the port leadingto the system MICE 1. To avoid the creationof short-term loops in the ring, the systemswill discard the data frames in the memoryuntil the ring has been rebuilt.

In contrast to the spanning tree, the failovertime of the HiPER Ring does not depend onthe number of systems used. The retardingfactors additionally caused by a greatnumber of switches are the transit time of the

watch-dog packets on the cables, the delaysper switch due to the system and one packetmaximum which a switch has just begun tosent out as a watchdog packet comes in.However, these factors have just a minorinfluence on the failover time, as will becomeevident in the measurement tests as well sothat even in the case of 50 switches and aline length of 3.000 km a failover of less than500 ms can be guaranteed.

According to the indications given by themanufacturer, it is not possible that a framewill be doubled or the frame order will getmixed up as a consequence of the failure orthe rebuilding of a ring as can happen withthe RSTP process. This statement is alsoapplicable for all other redundancyprocesses of Hirschmann.


MICE 3 MICE 5MICE 4




Picture 12: Dual Homing

Dual Homing

Originally, a HiPER Ring consisting of hubsystems was connected to the backbone bymeans of Dual Homing. To do so, two units ofthe ring are connected to a unit, a modularswitch of the MACH series (see Picture 12).

The Dual Homing is configured on the MACH3 which switched one of the tow prots intothe discarding mode. If the primaryconnection fails the MACH can put thesecondary connection almost immediatelyinto operation.

An advantage of the Dual Homings is the veryshort failover time. A serious disadvantage,however, is the fact that the backbone switchis a single-point-of-failure. To compensate thisdrawback, it is possible that Dual Homing portsuse different basis boards of the same MACHunit. However, with this process there is stillthe risk of a failure of the entire modular system– a default which cannot be compensated.

Another drawback is the fact Dual Homing isnot suitable for the coupling of two control ringswith one another since the function is onlyavailable for the MACH series.

HiPER Ring coupling

The more modern HiPER Ring coupling doesno longer have these disadvantages. In thisprocess, redundancy is provided by meansof the switches in the control area. For thispurpose, a switch is configured as standbymanager. Subsequently, the latter andanother switch of the ring are connected toone or two different units of another ring.Furthermore, the two are connected witheach other via a special “standby port“. Oneof the two switches is now the master whichactivates the coupling line to the other ring;the other switch, which has become theslave, sets his port into the discarding mode.

If the connection of the masters to the otherring is interrupted it will immediately informthe slave and the latter will activate hisconnection.

Since, theoretically, the connection of theother ring can be interrupted as well – in thisexample between MACH 1 and MACH 3 – orthe standby line between RS2 1 and RS2 2can break down master and slave will ex-change test packets as in case of the HiPERRing. If these test packets do no longerarrive RS2 2 will also activate his connectionto MACH 3, independently of the fact that theorder regarding the control line in default hasnot been received; however, this kind offailover is significantly slower, as shown bythe test.

Approx. by the end of 2003, a solution for theMICE switches will be available where theadditional control line can be dropped byreplacing it with control packets.

HiPERRing

HiPERRing

RS2 1

MACH 1

RS2 2

MACH 3



Picture 13: HiPER Ring coupling

HiRRP

The Hirschmann Router RedundancyProtocol, shortly HiRRP, is a typical routerredundancy process alike VRRP. The basicfunctioning of these protocols will not beexplained in detail here; as literature, werecommend the article published by PetraBorowka in the Network Special Insider(“Protokolle zur fehlertoleranten Konfigurationdes Default Router Eintrags”, only in German)or the more detailed ComConsult Report withapplication examples (“Design-VariantenLokaler Netzwerke im Vergleich”, only inGerman) by Petra Borowka and MakusSchaub.

A special characteristic of the HiRRP processis the fact that, compared to the standardprotocol VRRP, it has a distinctly shorterfailover time of less than 800 ms instead of 3to 4 seconds with VRRP. Furthermore, con-figuration has considerably been facilitated.However, each time only two routers can useone virtual IP address, not, as with VRRP, asmany as required which, however, suffices inmost cases.

HiPERRing

HiPERRing

RS2 1

MACH 1

RS2 2

MACH 2



Test of the Redundancy Processes

Picture 14: Test Sequence of the Fast-Ethernet HiPER Ring

HiPER Ring

Test Processes Applied

As it is difficult to determine the “real” failovertime we opted for an indirect measurementprocedure which is also more relevant inpractice: the time period is measured duringwhich a client system does not receive pack-ets. For this purpose, a continuous data flowis initiated between two PCs on UDP/RTPbasis by means of a load generator, i. e.Chariot by Ganymede.

Each test period lasted 10 seconds duringwhich the ring was interrupted and rebuiltonce. During these 10 seconds, 2000packets with 1000 bytes each were sent,which corresponds to a frequency of 0.5 msas time distance between two packets.

At the end, the duration of the connectioninterruptions was calculated on the basis ofthe loss rate of bytes.

In the tests, five MICE systems and fourRAIL switches were used. The MICE-MICEcoupling was entirely put into practice withglass fiber, for the connection to the RAILsystems one time twisted-pair (MICE 1 toRS2 4) was used and one time glass fiber(MICE 5 to RS2 1).

In all tests, MICE 2 was the redundancymanager and the ring port to MICE 1 was thepassive port in the default-free ring.

The two test units were connected to MICE 1and MICE 3 so that in the thus generatedring the packets were forced to run throughall switches; this way the packet loss wasmaximised since each time the line had themaximum extension.


MICE 3 MICE 5MICE 4


PC1 PC2



Result 1: Reaction to different line failures

Failure of Different ConnectionTypes in a Fast-Ethernet HiPER Ring

At first, the reaction time to the failure of linesfor a HiPER Ring with nine industrialswitches was tested. Here it was verifiedwhat impact the various failure possibilitieshave on the recovery time. The results areshown in Result 1.

At first, it is striking that there is a con-siderable difference whether one or bothdirections of a glass fiber connection aredown. While it is possible to repair the failureof both directions within approx. 200 ms, theHiPER Ring needs more than twice the timefor the repair of a failure of only one glassfiber direction and nearly reaches thedeadline of 500 ms indicated by Hirschmannas maximum down time.

The reason of this significantly varyingreaction is the fact that the link failure isimmediately recognized as link-down by oneof the units only and then sent to the ringmanager. If, in an unfavorable case, the ringmanager receives this message on thepassive port, it will not interpret the packetand thus switch-over is not exercised byevent-control but is recognized by the ringmanager due to the missing watchdogpackets and remedied accordingly. Thisreaction of the ring manager is identical bothfor the MICE and the RAIL switches. A MACHsystem as ring manager would, according toHirschmann, also process the signal packeton a passive port.

Since in case of a failure of the twisted paircable both directions are down this failure isalso remedied by event-control; the samegoes for a simultaneous failure in bothdirections of an optical fiber cable.

After rebuilding of the ring, the reaction willalso occur event-controlled. The problem of“one-sided“ error messages does not arisethis time. The down-time of the failover isapprox. 200 ms for glass fiber, for twisted-pair it is approx. 300 ms as the auto-negotiation process takes longer andadditionally retards the switch-over process.

Failure under Load

The structure of the previous test remainedunchanged. This time, however, in addition tothe data flow of the test process a basic loadof 40 Mbit/s was generated between the twotest stations so that the ring was exposed toa throughput of approx. 56 Mbit/s while onlyone direction of an optical fiber connectionwas interrupted.

0,000

0,100

0,200

0,300

0,400

0,500

0,600

Fiber oneDirection

Fiber bothDirections

TwistedPair

Failure Rebuild



Result 2: Switch-over with ring load Result 3: Impact of the measurementdirection

In Result 2, the results of an unloaded ringare compared to those of a ring with load.One can see very clearly that, in the threemeasurements taken, there are no significantdifferences between the loaded and the un-loaded ring.

As no difference had been measured in thetests run with or without load all further testswere effected without load.

Test Flow in Both Directions

The next test with unchanged test sequencewas designed to check whether the directionof transmission is of importance.

For this test, a data flow was simultaneouslytransmitted from PC 1 to PC 2 and from PC2 to PC 1. In Result 3, the results of threetests for each direction are compared to ameas-urement in only one direction for therebuilding of the ring. As expected, the resultdemonstrates that there is no difference as tothe direction the data flow or whether theanalysis is made one or both directions.

For reasons of simplification, only one dataflow in one direction was used for themeasurements in the following tests.

Influence of the Number of Switchesin the HiPER Ring

In their leaflets Hirschmann indicate that theextension of the ring and the number ofswitches hardly have any repercussion on thefailover time. For this reason, the number ofswitches was increased from three to nine.inthe following test. At first, the MICE switcheswere used in an increasing number, andsubsequently the RAIL switches wereimplemented as well until the ring sequenceof the previous tests was achieved again. Foreach measurement, the one of the two glassfiber connections most distant from the ringmanager was disconnected so as to measurethe delay of the greatest number of switchespossible during transmission of the eventmessages.

0

0,1

0,2

0,3

0,4

0,5

0,6

1. F

ailur

e

2. F

ailur

e

3. F

ailur

e

1. R

ebuil

d

2. R

ebuil

d

3. R

ebuil

d

with Load without Load

0

0,05

0,1

0,15

0,2

0,25

0,3

1 2 3

Only one DirectionDirection 1Direction 2



Result 4: Number of the switches in the HiPER Ring

From the result diagram, you can see fouraspects very clearly:

1. The number of switches has no influenceson the failover time of a HiPER Ring. Thevariations during the recovery of a lineinterruption are within the scope ofdeviation to be expected.

2. The switch series does not seem to haveany impact on the failover time either.

3. The mixture of optical fiber and copperconnections necessary from the seventhswitch onward due to the test units useddoes not make any difference either.

4. The failover time was always below 500ms.

3 4 5 6 7 8 9

0

0,05

0,1

0,15

0,2

0,25

0,3

0,35

0,4

0,45

0,5

Rebuild Failure



Gigabit HiPER Ring

Picture 15: Test structure of Gigabit Ethernet HiPER Ring

Result 5: Comparison of Fast-Ethernet vs.Gigabit-Ethernet HiPER Ring

In the tests, two MACH 3002 and one MACH3001 were used. The MACH switches wereinterconnected to the glass fiber basis bymeans of a Gigabit HiPER Ring. During thetests, the cable between MACH 2 and MACH3, which was not connected to the ringmanager, MACH 1, was interrupted.

In the diagram, the results of the interruptionof one and both glass fiber connections arecompared. The results reveal the followingaspects:

• In case of a failure of the ring, the GigabitHiPER Ring is significantly faster thanFast-Ethernet ring. The two reasons arethe faster reaction of the MACH systemswith processors of higher performancecapacities and the ten times highertransmission speed of error messageswith Gigabit.

• It is without importance for the failovertime whether one or both fibers of anop-tical waveguide connection are down.Thus the statement is confirmed sayingthat a MACH switch can also processevent-controlled packets on a passiveport.

• In contrast to the Fast Ethernet, on re-building the Gigabit ring switches moreslowly than in case of a failure. Accordingto Hirschmann, the reason for this is thedelay caused by the auto-negotiationprocess.

• The asymmetric delay caused by the auto-negotiation process can be measured onlyif an optical fiber cable is disconnected.

MACH 3MACH 2MACH 1Redundancy Manager

PC1 PC2

0,000

0,100

0,200

0,300

0,400

0,500

Gigabit

1 F

iber

Gigabit

2 F

iber

FastE

ther

net 1

Fibe

r

FastE

ther

net 2

Fibe

r

Failure Rebuild



Picture 16: Test structure of the HiPER Ring coupling

Ring coupling

For the switch-over test of the HiPER Ringcoupling, the Fast-Ethernet HiPER Ring andthe Gigabit-Ethernet HiPER Ring wereconnected with each other. The structure isshown in Picture 16; the orange line is thecontrol line between the standby managerand the standby partner.

For the test, the switches were configured ina way to ensure that they return to theirstatus after rebuilding of the original

connection. Subsequently, the failover timesof failure and rebuilding of the line betweenMICE 4 and MACH 3 were measured. In asecond test, the down time was measuredresulting when the control line is alsodefective or not connected (Result 6).

If the control line is connected the failovertimes are approx. 200 ms; this is valid for thefailure as well as for the rebuilding of thering.

MACH 3MACH 2MACH 1Redundancy Manager

PC1 PC2


MICE 3 MICE 5MICE 4


PC2

Standby Manager

PC1



Dual Homing

Result 6: Failover times of the HiPER Ringcoupling

Picture 17: Test structure Dual Homing

Result 7: Dual Homing vs. Ring Coupling

If the control line is missing down time will beincreased by a multiple to over 1.4 secondswhich is by far in excess of the tolerancevalue of 500 ms defined by Hirschmann.However, this case would only occur if thecontrol line was defective. Thus, followingcommisioning of a HiPER Ring coupling thecontrol line should be examined carefully bythe management as to correct functioning andshould also be monitored during operation.

0,000

0,200

0,400

0,600

0,800

1,000

1,200

1,400

1,600

Failure Rebuild

with control line without control line

HiPERRing

HiPERRing

MACH 1

MACH 3

MICE 2 MICE 3

0,000

0,050

0,100

0,150

0,200

0,250

0,300

Failure Rebuild

Ring Coupling Dual Homing



HiRRP

Picture 18: Test structure of HiRRP

For the Dual Homing test the test sequenceremained nearly unchanged, only the ringcoupling was turned off and Dual Homingwas configured at the MACH 3 which wasconnected with the two RAIL switches MICE2 and MICE 3.

Also in the Dual Homing process, failovertimes will remain below the 500-ms limitvalue. In case of a failure of the primary lineswitch-over will be effected within approx. 50

ms. Rebuilding will be exercised within a timerange of 250 ms.

If you compare Dual Homing with the modernring coupling, it is striking that switch-over ismuch faster in case of a failure at the olderprocess than for the more modern onemethod. Background for this better switch-over reaction is the fact that the failure isdetermined and corrected by means of aunit.

To be able to measure the HiRRP processwithout influences by other factors the

Gigabit HiPER Ring of the MACH switcheswas replaced by a linear sequence.

MACH 3MACH 2MACH 1

PC1 PC2

MICE 1 MICE 2Redundanz Manager

MICE 3 MICE 5MICE 4


PC2

PC1

Primary Router Secondary Router



Result 8: HiRRP

The two router modules of MACH 1 andMACH 3 were configured for HiRRP. MACH 1was an active router for both test stations aslong as both could be reached. In this test,the failure of the connection between MACH1 and MACH 2 was tested which caused thatMACH 1 could no longer be reached.

For the failure of the active router, a failovertime of approx. 700 ms was measured. Thistime is within the range expected of below800 ms which was calculated on the basis oftheoretical facts. The switch-over issignificantly higher than that of the HiPERRing so as to ensure that first a Layer-2recovery can be effected prior to changingsomething in Layer 3. If, for instance, theoriginal Gigabit HiPER Ring were still activein the test sequence the line failure couldhave been compensated by a switch-over tothe redundant connection and thus a switch-over of the Default Gateway function wouldhave been unnecessary.

A comparison: if you operate VRRP with thedefault values a failover time of 3 to 4seconds is measured, with HSRP evenaround 10 seconds.

HiPER Ring vs. RSTP

In the next tests, the RSTP process wastested and the results were compared withthe HiPER Ring.

Test Sequence of RSTP

For the tests of the Rapid Spanning Trees,each time four 2950 workgroup switches andfour 2955 industrial switches of Cisco wereused. Alike in the test of the Fast-EthernetHiPER Ring, the catalysts were set up asring. For the test, just the root wasdetermined, the other default values of theRSTP remained unchanged. Each time, thetime range was measured during which theexchange of packets between sender andreceiver was stopped. In analogy to theHiPER Ring tests, the times for interruptionas well as for rebuilding of the ring weremeasured. During all tests, sender andreceiver were connected to the root and to aneighbour switch. To trigger the switch-over,this connection was interrupted.

Failover time Reaction of RSTP overEight Switches

First the down-time resulting from a ring witheight switches was measured. In all, the testwas repeated ten times. The results areshown in the below Result 9.

In comparison to the HiPER Ring, there arethree striking differences:

1. The RSTP process requires considerablyless time for the rebuilding of the originaltopology than for recovery in case of afailure. The reason is that for the rebuildingof the end-to-end communication “only“the link between the root and itsneighbour had to be rebuilt and that therewere no packets in the memory of theswitch which had to be cancelled.

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

1 2 3Failure Rebuild



Result 9: Failure and Rebuilding of a ring with RSTP components

However, this may cause a mix-up of thepacket order at the receiver’s end.

2. The time required for recovery is subjectto great variations and ranges betweensharply under 200 ms and over onesecond for eight switches. In contrast tothe HiPER Ring, the duration of failurecannot characterized as deterministiccriterion. In some cases, during rebuildingeven more than two seconds weremeasured.

3. The recovery time of an average of almost800 ms is considerably higher than for theHiPER Ring.

Worst-Case Failure with RSTP

The measurements taken in the previous testdo not reflect the absolute worst case of therapid spanning tree since the link-down oftwo switches is immediately detected andreported. However, it might happen that

media transducers are inserted in the ring orthat systems are connected which cannottransmit a link-down to the switches andwhich themselves are no RSTP units. In thiscase, the failure must be detected by theabsence of the BPDU packets. To simulatesuch a default situation, two units wereinserted into the ring which do not speakRSTP but which are able to transmit BPDUs.Subsequently, the cable between these twocomponents was disconnected.

Result 10 compares the failure which can bedetected directly by the switches with the link-down which is only detected by the missingBPDUs. It is striking that the time increasesby approx. the factor five and that theinterruption is above 5 seconds.

Thus the time range is considerably betterthan the classical spanning tree but still tentimes higher than the highest possible downtime of the HiPER Ring.

0

0,2

0,4

0,6

0,8

1

1,2

1 2 3 4 5 6 7 8 9 10

Failure Rebuild



Result 10: Link-down and missing BPDUs with RSTP

Increase of the Failover Timein Relation to the number ofSwitches

In the last test with the RSTP components,the number of switches was step by step re-duced from eight to three. Each time, threemeasurements were effected.

In Result 11 the results are compared tothose of the HiPER Ring. The following canbe seen from the diagram:

• The failover time of the RSTP process de-pends on the number of switches.

• For the HiPER Ring process, however, thenumber of systems is of no importance.

• Even if at first the failover time of theRSTP implementation is faster for up tofour switches as compared to that of theHiPER Ring, it is subsequently risingstrongly and is finally significantly higher.

In the drawing, it is astounding to see thatfour switches seem to switch faster thanthree switches and six faster than five. Thisis, however, due to the significantly varyingfailover times of RSTP which statisticallyhave a major influence in the threemeasurements.

1 2 3 4 5 6 7 8 9 10

0

1

2

3

4

5

6

Link down missing BPDU



Result 11: Comparison HiPER Ring vs. RSTP with differing numbers of switches

Important Remark on the DifferentRSTP Implementations

The underlying implementation of RSTP in theCisco switches does not represent that of allother manufacturers. As shown at theSommerschule 2003 of the ComConsultAkademie, there are versions of the rapidspanning tree which cannot send event-controlled BPDUs but which make use of the

regularly passing packets (i. e. every 2seconds) to dissimate modification messages.Here, with eight switches, propagation of aninformation will take up to 14 seconds.

Recommendation: If you want to implementRSTP in a failure-critical surrounding youshould in any case run your own test tomeasure the failover time!

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

3 4 5 6 7 8

HiPER Ring RSTP



Summary and Discussion of the Results

• Even in the worst case, all Hirschmannproprietary processes remain below thelimit value of 500 ms regarding the failovertimes both in failures and in rebuildingprocesses following removal of the error.

• The only exception is the Router Redun-dancy Protocol HiRRP with 700 to 800 ms.However, as this is a Layer-3 process, forlogical reasons, the time required heremust exceed that of the Layer-2mechanisms being located underneath.

• In the lab test of the ComConsult Lab, thestatement made by Hirschmann wasconfirmed saying that the failover timesare almost independent of the number ofunits forming a HiPER Ring.

• The rapid spanning tree achieves lowfailover times, too. In contrast to theHirschmann process, however, they aresubject to great variations.

• The failover times with RSTP are in-creasing significantly with the number ofsystems used and already exceed thoseof the HiPER-Ring with approx. four to fiveswitches.

• If failover times are needed which, on theone side, are determininistic and, on theother side, are far below one second thiscannot be achieved at present withouttaking recourse to proprietary processeslike the HiPER Ring.

• Furthermore, the HiPER Ring is wellsuitable for projects with a largeextension, i. e. where long lines and manyswitches connected in a row are required.

• In contrast to the HiPER Ring, with therapid spanning tree packet repetitions anda mix-up of the packet order may occurduring a link-down and a link-up.

• Advantages of the rapid spanning tree asagainst the Hirschmann process are thatthe rapid spanning tree is standardisedand that only one technique is required toset up a flat Layer-2 network.

• When comparing the design, the followingcan be said: the guidelines laid down bythe Hirschmann process are strict and noteverything is possible; the rapid spanningtree allows more freedom, however, thusenhances the potential risk that an activetopology will ge generated which canhardly be understood, even less in case ofan error.



Pictures

Picture 1: Increase of the failover time in relation with number of switches 2

Picture 2: RAIL Switch 3

Picture 3: MICE Switches 3

Picture 4: MACH 3002 4

Picture 5: Configuration via hyper-terminal 9

Picture 6: Access via WEB Interface 10

Picture 7: Network Survey with HiVision 11

Picture 8: Unit Presentation of HiVision 12

Picture 9: Design with Hirschmann Process 15

Picture 10: Monitoring of a HiPER Ring 15

Picture 11: Error removal in the HiPER Ring 16

Picture 12: Dual Homing 17

Picture 13: HiPER Ring coupling 18

Picture 14: Test Sequence of the Fast-Ethernet HiPER Ring 19

Picture 15: Test structure of Gigabit Ethernet HiPER Ring 23

Picture 16: Test structure of the HiPER Ring coupling 24

Picture 17: Test structure Dual Homing 25

Picture 18: Test structure of HiRRP 26

Results

Result 1: Reaction to different line failures 20

Result 2: Switch-over with ring load 21

Result 3: Impact of the measurement direction 21

Result 4: Number of the switches in the HiPER Ring 22

Result 5: Comparison of Fast-Ethernet vs. Gigabit-Ethernet HiPER Ring 23

Result 6: Failover times of the HiPER Ring coupling 25

Result 7: Dual Homing vs. Ring Coupling 25

Result 8: HiRRP 27

Result 9: Failure and Rebuilding of a ring with RSTP components 28

Result 10: Link-down and missing BPDUs with RSTP 29

Result 11: Comparison HiPER Ring vs. RSTP with differing numbers of switches 30

Pascalstraße 25 • 52076 Aachenphone 02408/955-400 • fax [email protected]

product hiper ring vs. rstp - interlog-it.com · product analysis: hiper ring vs. rstp the...

Documents