PACKETEER TECHNICAL PAPER

Enhancing MPLS Network Performance

Packeteer, Inc.
10201 N. De Anza Blvd., Cupertino, CA 95014 Tel: (408) 873-4400 info@packeteer.com | www.packeteer.com

Packeteer, the Packeteer logo, combinations of Packeteer and the Packeteer logo, as well as AppCelera, PacketSeeker, PacketShaper, PacketShaper Xpress, PacketWise, and PolicyCenter are trademarks or registered trademarks of Packeteer, Inc. in the United States and other countries. Other product and company names used in this document are used for identification purposes only and may be trademarks of other companies and are the property of their respective owners. Copyright 2001-2004 Packeteer, Inc. All rights reserved. No part of this publication may be reproduced, photocopied, stored on a retrieval system, transmitted, or translated into another language without the express written consent of Packeteer, Inc. Packeteer software is licensed, not sold, and its use is subject to the license terms set forth in the end user license agreement. (05/2004)

Table of Contents
Enhancing MPLS Network Performance ......................................................................................... 3 Brief MPLS Review.........................................................................................................................3 Limitations and Solutions for Delivering QoS With MPLS ............................................................4 Challenge 1: Match the Right Traffic and Service Class .................................................................5 Precise Traffic Classification ...................................................................................................6 Challenge 2: Prevent a Bottleneck ...................................................................................................8 Challenge 3: Provide Performance Feedback .................................................................................. 9 Measuring Response Times....................................................................................................10 Access to MPLS Performance Data .......................................................................................11 Other Benefits of Using Packeteer in MPLS Networks ................................................................. 12 Preparing for MPLS ...............................................................................................................12 Weighing the Cost of a Premium Class of Service ................................................................13 Marking, Pushing, Popping, and Swapping ...........................................................................13 Delivering Network Backup...................................................................................................13 Assisting Voice/Data Network Convergence......................................................................... 14 Summary ........................................................................................................................................14

Packeteer, Inc.

Enhancing MPLS Network Performance

Multi-Protocol Label Switching (MPLS) is here to stay. Recent surveys indicate that the majority of large organizations have either changed to MPLS-based IP VPNs (virtual private networks) or intend to change soon. Most businesses make the move to take advantage of different classes of service to ensure appropriate application performance. Consider these perspectives: MPLS is a fundamental protocol thats going to be used for the next ten to twenty years because its the first complete and stable convergence protocol for data, voice, and video infrastructure.1 By 2007, IP VPN (frequently MPLS-based) will become the most popular WAN technology in the U.S.2

MPLS has become a leading vehicle for connecting an organizations distributed locations. It offers advantages to both service providers and enterprises. For the service provider, MPLS reduces cost, simplifies provisioning, provides wider service coverage, and enables differentiated services. In addition to the promise of multiple levels of quality of service (QoS), MPLS offers the enterprise a meshed architecture, scalability, and network convergence, eliminating the need for multiple networks. Service providers need to make their bandwidth services more valuable to enterprises. After all, pitching bandwidth as a stand-alone commodity is no longer compelling. One opportunity for service expansion is differentiated performance standards. Meanwhile, enterprises realize the importance of ensuring that key business applications operate consistently, reliably, and promptly requirements that become essential as organizations adopt converged networks supporting voice, video, and data. Based on these advantages, its no wonder that MPLS, with its ability to offer different performance levels to different applications, seems to be the ideal solution for both parties. On paper, it all makes sense. However, once MPLS is implemented, organizations frequently discover that placing key applications into premium service classes does not reap the expected benefits. What goes wrong? Bandwidth contention within service classes, overloaded premium classes, bottlenecks outside the MPLS cloud they all can result in poor and unpredictable application performance. Furthermore, the complex requirement of mapping large numbers of applications (often numbering in the hundreds) to a relatively small number of MPLS classes creates confusion, and it leaves businesses with unfulfilled performance and cost-benefit expectations. Packeteer, the leading provider of application traffic management solutions, offers an integrated platform that addresses the obstacles associated with MPLS networks. Packeteer boosts application performance, manages the complexities of juggling each service class traffic load, eases bandwidth bottlenecks, and verifies application and service-level performance, providing a very cost-effective complement to MPLS. This paper details the obstacles that emerge in MPLS environments and the value Packeteer brings to application performance as it solves these operational issues.

Brief MPLS Review

As a brief reminder of how MPLS operates, recall that in the typical network without MPLS, packet paths are determined in real time as routers decide each packets appropriate next hop. Conventional IP routing requires time and eliminates opportunity to influence packets paths. With MPLS, explicit and pre-defined network paths transport specific types of traffic.

Hossein Eslambolchi, CTO and CIO of AT&T; President of AT&T-Labs; from Point of View/Networking and Business Strategy, 06/03/2003. 2 IDC, U.S. IP VPN Services Forecast, 2002-2007 (IDC # 28575).

Packeteer, Inc.

MPLS solved the problem that router manufacturers faced when incorporating QoS into very large IP-VPN networks: ensuring that each and every router can identify and process each and every traffic flow appropriately requires so much processing power as to be ineffective and non-scaleable. A better approach, and the one that MPLS adopts, is to label traffic flows at the edge of the network and let core routers identify the required class of service with a simple and quick label check. MPLS reduces the burden of differentiating types of traffic and assigning appropriate class-of-service labels by focusing the task on the edge of the MPLS network. The onus is usually on a router, called the label edge router, or LER. Optimally, the MPLS labels indicating the best and fastest service classes go into the most urgent applications packets (and only the most urgent applications packets).

Limitations and Solutions for Delivering QoS With MPLS

The ability to provide diverse applications with appropriate bandwidth and performance is part of MPLS purpose and design. So, why do organizations frequently find it difficult to realize the benefits of MPLS? Service providers do an excellent job of providing an MPLS core network with different service classes. However, an MPLS solution degrades as it faces three major challenges: Challenge 1: The right traffic does not get placed in the right MPLS service class. Premium classes deliver sub-premium performance as they drown in copious non-urgent traffic; important applications are improperly assigned to only best-effort classes. Challenge 2: Traffic gets hung up in a congested bottleneck just before each entry point to the providers MPLS network. In addition, unmanaged traffic heading into a LAN (inbound) grows unruly, using an inappropriately high flow rate. Challenge 3: Providers and enterprises need information on the performance of each application and each service class transported over their MPLS network. Concrete, quantified service-level assessments are rare.
MPLS and Packeteer improve response times.

Packeteer complements MPLS installations and overcomes each of the three QoS challenges listed above with pragmatic, real-world solutions. Packeteer offers visibility into and control of application traffic and enhances MPLS performance gains as it: Detects, identifies, and classifies diverse applications, assigning distinct QoS tags Packeteer, Inc.

Ensures that the traffic within a particular MPLS service class is the right traffic, meant for that class Eases the bottlenecks that form at the entry points to MPLS networks Extends MPLS performance benefits to the network edge and users premises Measures and graphs per-application and per-MPLS-class performance, enabling assessment of service level agreement (SLA) compliance

Unlike other elements of the MPLS network, the enterprise typically owns and manages Packeteer products. A service provider can also use Packeteer to offer customers additional services. Packeteer appliances are deployed at the network edge just before each locations entry point to the MPLS network.
Packeteer appliances sit on the LAN immediately before the service providers routers leading into and out of the MPLS network.

Challenge 1: Match the Right Traffic and Service Class

The ability to match an application to an MPLS class of service is a prerequisite for an effective QoS implementation. After all, if network equipment is unable to identify which application is passing, how can it determine the appropriate service level? With the most granular and specific application classification, comes the most accurate and beneficial MPLS service-class assignments. If it is not possible to identify applications, it is also not possible to ensure that a particular application gets its proper MPLS label and corresponding class of service. Precise application classification requires the ability to analyze traffic flows at all networking layers including the application layer, layer 7. Traffic classification must find and decipher application-specific information contained within transported data. If MPLS solutions depend only on layer-4 port numbers to identify applications, not only do non-critical (or recreational) applications pass undetected, but they also seize bandwidth from premium (and more costly) MPLS classes. These questions highlight examples: Some ERP applications use any one of 10,000 layer-4 port numbers. Which ERP traffic actually gets identified as ERP? All? Some? None? Some recreational traffic searches for open ports on firewalls, disguising itself as ERP or HTTP. Will KaZaA get a class of service that is appropriate for recreational traffic? Or will it get ERPs service class?

Because the routers at the edge of the MPLS network use port numbers to identify traffic, most service providers require organizations to manually map each TCP port number and IP address to each application and user, and then in turn to a provisioned MPLS class. Even if a business doesnt mind tending to this manual-mapping task, another fundamental problem remains. Once traffic is assigned to an MPLS class, any single flow can consume the class entire bandwidth. One highcapacity user or one flow from an unsanctioned bandwidth-hungry application can undermine performance of ERP or other mission-critical applications. Packeteer, Inc. 5

Without tools to differentiate and analyze each application, the premium MPLS class traffic can appear to be in compliance with the providers service level agreement. However, behind the scenes, rogue traffic might be the only beneficiary of costly premium performance. Label edge routers application classification capabilities are limited. They lack precision and accuracy when tagging or labeling MPLS traffic. This is the core problem behind the discrepancy in organizations expected and actual application performance.

7 6 5 4 3 2 1

Application Presentation Session Transport Network Data Link Physical

Routers classify traffic by IP address or TCP port number, but many applications share the same TCP number, and IP addresses are assigned dynamically. This traffic tree organizes traffic first by class-ofservice and then by application. Users can control bandwidth and examine performance for traffic in any branch of the tree no matter how granular its classification. Packeteer classifies traffic using application-specific information contained within the data and at all layers through layer 7.

Precise Traffic Classification

The ability to distinguish one type of traffic from another as it passes from a non-MPLS edge to an MPLS core is essential to ensure that applications are assigned to the correct class of service. Only with precise classification can applications be treated according to their respective business importance. Many products can differentiate traffic based on MAC, IP address, and port numbers, and then perhaps infer an application using a static port number. A few vendors offer limited application-layer classification for specific traffic types over specific ports. Packeteer offers all these same address- and portbased techniques, but goes far beyond them to identify an incredible diversity of complex applications. Packeteer scrutinizes multiple packets in a flow, behavioral characteristics, application markers, and traffic histories to determine precisely which applications are in use. Packeteers classification feature builds a hierarchical classification tree, inserting an entry for every distinct traffic type it identifies. A traffic tree provides an organized view of a traffic mix and makes it easy to assign bandwidth-allocation or QoS-tagging policies to entire traffic categories with one step.

Packeteer, Inc.

Some of the applications and protocols Packeteer automatically detects, identifies, and classifies include:
Client/Server
CVS FIX (Finance) Folding@Home INFOC-RTMS INT-1 (Unisys Interact) MATIP (Airline) MeetingMaker NetIQ AppMngr OpenConnect JCP PEPGate (Attachmate) Unisys-TCPA

File Server
AFS CIFS-TCP CU-Dev lockd Microsoft-ds NetBIOS-IP NFS Novell NetWare5 rsync SunND

Legacy LAN and Non-IP

AFP AppleTalk DECnet FNA, FNAonTCP IPX LAT MOP-DL/RC NetBEUI PPPoE SLP
SNA

Music P2P
Aimster Apple-iTunes AudioGalaxy Bit Torrent Blubster DirectConnect EDonkey FileRogue Filetopia Furthurnet Gnutella
Acquisition Ares BearShare Furi Gnotella Gnucleus gtk-gnutella LimeWire MyNapster Mactella Morpheus Mutella Nap Share Phex Qtraxmax Qtella Shareaza toadnode XoloX Emule Overnet Rhapsody Mac Satellite

Network Management
Cisco Discovery Day-Time ICMP(by packet type) IPComp Microsoft SMS NTP RSVP SMS SNMP SYSLOG Time Server

Session
GoToMyPC pcAnywhere REXEC radmin rlogin rsh Telnet Timbuktu VNC Xwindows

Content Delivery
Ariel Backweb Chaincast EntryPoint Kontiki Marimba NewsStand PointCast WebShots

Games
Asherons Call Battle.net Diablo II Doom EverQuest Half-Life Kali LucasArts (Jedi*) MSN Zone Mythic Quake I, II, & III SonyOnline Tribes I,II Unreal Warcraft III Yahoo! Games

Messaging
AOL IM,
Talk, Image,File, ISP,

Print
IPP LPR TN3287 TN5250p

Thin Client or Server Based

Citrix
Published Apps, Nfuse, IMA

ICQ IRC Lotus IM MSN Messenger Windows-POPUP Yahoo! Messenger

RDP/Terminal Server

Voice over IP
CiscoCTI Clarent CUSeeMe Dialpad H.323 I-Phone MCK Commun. Megaco Micom VIP MGCP Net2Phone RTP RTCP SIP Skinny (SCCP) T.120 VDOPhone

Routing
AURP BGP CBT DRP EGP EIGRP IGMP IGP MPLS (+tag, +app) OSPF PIM RARP RIP Spanning Tree VLAN (802.1p/q)

Database and ERP

Baan FileMaker Pro JDENet (JD Edwards) MS SQL Oracle (and by database) Oracle JVM Oracle EM PostgreSQL Progress SAP

MiddleWare
CORBA Java RMI SmartSockets SunRPC (dyn port) JavaClient

Healthcare
DICOM HL7

MultiMedia
MPEG (Audio, Video) Multi-cast NetShow NetMeeting QuickTime Radio@Netscape Real (Audio, Video) RTP RTSP SHOUTcast Streamworks VideoFrame WebEx WinampStream WinMedia

Host Access
ATSTCP Attachmate Persoft Persona SHARESUDP SMTBF TN3270 TN5250

Directory Services
CRS DHCP DNS DPA Finger Ident Kerberos LDAP RADIUS RRP SSDP TACACS whois WINS

Groove Hotline iMesh KaZaA KaZaA Lite Napster

Internet
ActiveX BITS FTP, Passive FTP Gopher HTTP HTTP Tunnel IP, IPv6, IPIP, UDP, TCP IRC Mime type NNTP Socks2http SSHTCP SSL TFTP UUCP URL Web browser type

E-mail and Collaboration

Biff ccMAIL DCOM (MsExchange) Groupwise (Novell) IMAP LotusNotes MSSQ OSI POP3 SMTP

Napster2 PeerEnabler Scour Tripnosis Winny

Amster audioGnome File Navigator Gnapster Grokster gtk napster jnapster MacStar Maxter My Napster Napigator NapMX Napster Fast Search Napster, MacOSX OpenNap Rapster Snap Spotlight WebNap WinMX

Security Protocol
DLS DPA GRE IPMobility IPSEC ISAKMP/IKE key exch L2TP PPTP RC5DES SOCKS Proxy SSH SSL (+shell) swIPe

Packeteer can classify traffic by a wide range of variables, including the following: Advanced Layer 7 application signatures Application sub-classification including Oracle and PostgreSQL (by database), Citrix (by published application or priority tag), FTP (by file name or extension), NNTP (by name or type of newsgroup), VoIP (by protocol or CODECs) HTTP sub-classification by URL, URL wildcard, content type, MIME type, browser type, HTTP tunnel, HTTP 1.1 virtual server SSL sub-classification by certificate common name Packeteer, Inc. 7

Layer 4 UDP and TCP ports, port ranges, and port lists Layer 3 IP addresses, address ranges, subnets, subnet ranges, MAC addresses, host lists, and LANs or physical device QoS markings, including DiffServ, IP-ToS, IP-CoS, IP precedence, MPLS label, MPLS experimental bits Frame relay interface, PVC/DLCI, ATM PVC, ATM interface, ISL-VLAN, 802.1q-VLAN, 802.1p-LAN

Once Packeteer identifies each passing packet, it can tag each packet with a DiffServ or IP TOS (type of service) identifier to give LERs the criteria they need to assign proper MPLS labels. As Packeteer identifies, differentiates, and tags applications, it ensures that only appropriate applications are assigned to any MPLS service class. It eliminates the contention issue that happens when too many applications go into the same service class because of a lack of visibility into various types of traffic. The frequency of missed SLAs declines. For example, with Packeteer, SAP and Oracle traffic can share one service class or have two distinct classes. Either way, they can sit in a separate MPLS service class than the best-effort class used for email and downloads. In another example, Citrix Published Applications and NFuse (using the ever-popular port 80) can have a higher service class than that assigned to web surfing and file sharing, also on port 80. Typically, the enterprise does not get involved with MPLS labels, leaving all MPLS involvement (including labels) to the service provider. However, Packeteer does have the ability to tag packets directly with MPLS labels if the enterprise or service provider wishes to do so.

Challenge 2: Prevent a Bottleneck

The transition from a non-MPLS edge to an MPLS core or WAN typically turns into a speed-conversion bottleneck as traffic from a speedy Ethernet LAN funnels into a slower line. Critical traffic at this bottleneck is not yet the beneficiary of MPLS performance advantages, as the traffic is not yet within the MPLS network cloud. MPLS implementations are usually done in the network core, leaving local LANs at the network edge as they are. Simply deploying an MPLS core network to deliver QoS wont yield end-to-end QoS for applications. The link from the local LAN to the MPLS core is typically the lowest capacity portion of the network. It backs up with deep queues and introduces the most latency. Even if SAP, for example, has been assigned to the premium class within the MPLS network, it might wait behind FTP packets and web traffic before entering. Packeteer eases the bottleneck, ensures that key business applications can always access the network, and eliminates or reduces queuing within LERs, a main contributor to unpredictable traffic delay. As Packeteer detects and identifies each application or traffic type needing assignment to a particular service class, it can also determine how that traffic should pass through the bottleneck point. Packeteer also regulates the rate at which traffic enters the network on the other side of connections, making sure that inbound traffic doesnt grow unruly and exceed its expected flow rates. Policy-based bandwidth allocation boosts or curbs application performance over the WAN or Internet. Packeteers flexible policies protect critical applications, pace greedy traffic, limit recreational usage, and block malicious traffic. Bandwidth minimums and/or maximums apply to each application, MPLS class of service, session, user, and/or location. Each type of traffic maps to a specific bandwidth-allocation policy, ensuring that each receives an appropriate slice of bandwidth.
Its as if each application or type of traffic gets its own appropriately sized link. If an application doesnt need its bandwidth at the moment, it goes to another that does. Bandwidth is never wasted.

An MPLS class of service is based upon a given traffic load. If the load grows too large, the MPLS implementation can drop excess traffic into a lower class of service. The portion of traffic that gets demoted is not necessarily the applications that someone would choose for slower service. Packeteer can ensure that the amount of traffic Packeteer, Inc. 8

entering an MPLS class does not exceed the expected and predefined volume for that class, preventing unexpected drops in service. With Packeteer, organizations can: Protect the performance of important applications, such as SAP and Oracle, or all applications in a premium MPLS class of service Enforce limits on traffic volume in each MPLS class of service Provision steady streams for voice or video traffic to ensure smooth performance Stop applications or users from monopolizing the link Contain unsanctioned and recreational traffic such as KaZaA and AudioGalaxy Reserve or cap bandwidth using an explicit rate, percentage of capacity, or priority Strike a balance between consistent access and a bandwidth limit for applications such as Microsoft Exchange that are both bandwidth-hungry and critically important Allow immediate passage for small, delay-sensitive traffic such as Telnet

With Packeteers control features, performance matches application characteristics, business requirements, and user needs.

These graphs compare usage and efficiency, before and after using Packeteers control features. Bandwidth usage smoothes out and avoids spiky peaks and valleys that create inconsistent delays. Retransmissions and their wasted bandwidth are all but eliminated.

Challenge 3: Provide Performance Feedback

When a service provider and enterprise enter into a contract for an MPLS network, they must agree on their service classes. For example, there might be four classes of service: Voice (for VoIP calls), Gold (for urgent business-critical applications), Silver (for important but less urgent applications), and Bronze (for traffic requiring only best-effort performance with leftover resources). Each class of service maps to performance commitments, an SLA. How do providers and enterprises validate that results match expectations? Can sub-par performance be attributed to the provider or enterprises network? Clear answers can help prevent conflicts and wasted diagnostic time.

Packeteer, Inc.

Both providers and enterprises need precise performance measurements for flexible intervals of time. Both parties also need to be able to divide delay measurements between the provider and enterprises area of responsibility. Traditional methods for measuring MPLS SLAs provide information only on the performance of a service class aggregate traffic. Enterprises enter a contract for delay, availability, throughput, and/or other metrics, based on averages in each service class. If one particular application in the MPLS class suffers poor performance, it is quite possible that the overall SLA for the class would remain compliant. It is important for any organization that uses MPLS for application QoS to be able to measure all metrics for a single specific application. In addition, when auditing SLA compliance, averages based on long sample periods are of less value than exact information based on instantaneous peaks and short-term averages. Packeteers comprehensive reports provide a clear view of historical performance, load, and efficiency. Extensive measurement data lives on board for up to two months and forms the foundation for Packeteers reports, all accessible with a web browser. More than 100 types of metrics are available for export and incorporation into third-party reporting tools. In addition, Packeteer ReportCenter can centralize reporting functions for multiple Packeteer appliances at edge locations (all the MPLS network entry points at branch offices, for example). ReportCenter can render many additional predefined reports and store measurement data for long periods of time.

Measuring Response Times

Packeteers position in the network monitoring all traffic that passes gives it a unique opportunity to provide accurate response-time measurements at a very low cost. Because it already sees every packet, Packeteer can calculate the time traffic spends traveling between a client and a server, the time used by the server, and the time spent on either side of a Packeteer appliance. Rather than collecting response data, Packeteer notes response times as traffic passes. This simple approach provides rich data without network impact or overhead. Packeteers response-time management facility offers techniques to measure performance, troubleshoot sluggish response, set service-level goals, and assess compliance. Organizations can: Track delay statistics for flexible traffic categories. Measure response times for an MPLS service class, individual application, host, subnet, and any transaction-oriented TCP traffic. Split each response-time measurement into network delay (time spent in transit) and server delay (time the server used to process the request). Identify users and servers with the slowest performance. Set acceptability standards and track whether performance adheres to them. Set the speed that divides good responses from bad (500 ms for the Gold MPLS class of service, for example), and set the percentage of transactions that should meet designated performance goals (95 percent, for example). Split network delays into time spent on either side of a Packeteer appliance. When Packeteer sits at the boundary between the enterprise and providers networks, separate metrics provide a clear, unambiguous dividing line of responsibilities. When performance is slow, these metrics indicate on which side of the dividing line the slowdown occurs.

Packeteer, Inc.

10

View current and historical performance data in intuitive tables and graphs, in a MIB (management information base) via an XML API or as raw data. SNMP management tools and third-party reporting tools integrate smoothly.

With a mechanism to compare actual and anticipated performance on an MPLS network, service-level agreements become more than empty promises.
See total transaction times divided into their network and server components. This graph shows that response time has been sporadically slow with frequent spikes. In addition, it shows that it was not the server that was causing the problems it was the network. If this is a graph of a critical application, its performance definitely needs some help. Some of Packeteers control features and perhaps a different MPLS service class are in order.

Access to MPLS Performance Data

If an enterprise owns and manages its own Packeteer appliances, then it can access any of the relevant reports to verify or diagnose performance. But if the service provider owns and manages Packeteer appliances, then the provider must find a convenient way to convey performance information to customers. The most convenient way to convey information is through a web page. But the repetitive, manual chores required to continually publish upto-date reports are painful. Packeteers customer portal is a web page that serves as a vehicle for delivering information about MPLS network usage, performance, and status to a customer. Per-customer tailored web pages offer information for any application or MPLS class of service, including average and peak bandwidth rates, response times, and efficiency. Customers can check their own status, thereby reducing the volume of support calls. Trust increases as all parties accurately monitor factors such as availability, usage, and performance. Reports are presented in HTML pages that the provider designs, using templates provided at the Packeteer website. The provider chooses either a single HTML page with customized data or a separate HTML page for each customer. Packeteer maintains the portals HTML pages, automatically customizes them for each viewer, serves the HTML pages when requested, and insulates customers from each others data. The customer portal offers the ability to: Generate metrics, graphs, tables, and reports that describe and validate usage and performance Brand the look and feel of the web page for each customer or group Notify customers of service changes or other timely messages Ensure customers are informed with no need for help-desk support

Packeteer, Inc.

11

Example of a customer portal

Other Benefits of Using Packeteer in MPLS Networks

Earlier sections of this paper covered the three most significant benefits of deploying Packeteer in an MPLS network: (1) getting the right traffic on the right MPLS path; (2) avoiding a congested bottleneck at each MPLS entry point; (3) assessing MPLS service-level performance. Additional benefits are also available.

Preparing for MPLS

Before implementing an MPLS network, organizations typically confront a few challenges during planning stages. Issues that Packeteer can determine or address include: Application mix Before determining MPLS service classes, its important to know the applications that are running over the network. As soon as Packeteer initiates automatic application discovery, organizations are often surprised by some of the types of traffic that appear in their traffic trees. Packeteer, Inc. 12

Load Organizations are often confounded when trying to select the sizes for each of their MPLS service classes. How big should the most premium class be? How about the best-effort class that will support all leftover traffic? Packeteer automatically measures load for each traffic category as it passes. If the organization classifies its auto-discovered traffic types into categories for each intended MPLS service class, then it can see each MPLS class load even before the MPLS network exists.

Pre-MPLS baseline picture MPLS networks should improve the consistency and speed of critical applications performance. But improvements are difficult to verify without a basis of comparison a performance before picture. Packeteer can measure application performance before and after MPLS deployment. Packeteer recommends three performance snapshots one before MPLS and traffic managements control features are applied; one before MPLS and after control features are applied; and one after both. These three snapshots provide a complete picture.

Weighing the Cost of a Premium Class of Service

Once an organization deploys an MPLS network, it pays a different rate for each MPLS service class. Presumably, a Platinum class costs more than Gold, which costs more than Silver, which costs more than Bronze, and so on. With Packeteer, the enterprise can compare performance in two MPLS service classes to determine if a cost premium is worth it. For example, it could compare average response times for Gold and Silver service classes traffic, revealing whether the difference in performance merits the difference in cost. A service provider could also do the same comparison using Packeteer. If the comparison reveals a less-thananticipated performance difference, the provider might consider some performance tuning on the premium class. If its greater than anticipated, the provider might decide to raise its premium rates in the future.

Marking, Pushing, Popping, and Swapping

Most Packeteer-MPLS implementations entail marking traffic with DiffServ or IP TOS tags to relay class-ofservice intentions to the LER, which does the actual MPLS labeling. However, Packeteer can also tend to MPLS labels itself, if desired. One packet can have more than one MPLS label in its header. All labels reside in the packets MPLS stack, an ordered list of labels that is processed on a last-assigned/first-removed basis. Routers frequently assign (called push), remove (called pop), or exchange (called swap) labels in each packets MPLS stack as they route. Packeteer can also perform these three operations push, pop, and swap as packets pass, relieving routers of extra overhead.

Delivering Network Backup

Organizations commonly consider MPLS networks when deploying new, business-critical applications such as VoIP or ERP. However, should the MPLS network fail, an alternate connection is needed. Typically, a back-up connection has less capacity than the primary connection. If and when the secondary connection becomes active, application contention becomes a more significant problem as key applications fail or are so slow as to be unusable. Packeteer can control traffic over slower back-up networks to ensure that key applications remain usable, even in the case of a major network failure. With multiple physical interfaces, Packeteer can implement different traffic management strategies for the primary and back-up connections. Each interface can have its own Packeteer, Inc. 13

settings for bandwidth-allocation policies, application priorities, and traffic marking. Although intervals of normal usage on the primary network might feature a mix of applications with network access, back-up mode might permit network access only to critical applications and save non-critical applications until spare bandwidth is available. Many other strategies are possible.

Assisting Voice/Data Network Convergence

For many, planning a VoIP installation is the catalyst for considering an MPLS network. A network that supports both voice and data needs multiple QoS levels, and thats precisely what brings MPLS to mind. Packeteer facilitates network convergence, again complementing the advantages that MPLS offers. In preparing for a VoIP installation, Packeteer helps organizations: Determine how much bandwidth voice traffic will need. Simulate peak call volumes impact on existing applications and their performance. Decide if a capacity upgrade is needed. Identify many types of voice traffic, including RTCP, SIP, Megaco, MGCP, Skinny, MCK-Signaling, RTP, MiCOM VIP, MCK Voice, and others. Protect bandwidth for VoIP as a whole. Clear easy passage for VoIPs setup and control traffic. Allocate the steady rate required for good performance for each voice stream. Handle over-subscription (every employee suddenly decides to use the phone at the same time) gracefully. Assign appropriate QoS tags or MPLS labels. Control bandwidth allocation appropriately for competing data applications.

Once VoIP and data are both active on one network, Packeteer can:

Summary
As a growing number of organizations turn to MPLS networks for network convergence and a range of service levels, the need for assistance in MPLS preparation and delivery of end-to-end quality of service becomes more significant. Packeteer provides that assistance. Incorporating Packeteer in MPLS networks from planning through management offers compelling advantages. It helps determine what to purchase; identifies and marks the application traffic needing special handling; assesses performance; and keeps traffic sailing smoothly at the entry and exit points of an MPLS network, extending the performance advantages of an MPLS core network all the way to the edge. With Packeteer, network and application performance align with business needs. If youd like more information about Packeteer solutions, consult Packeteers web site or call 408-873-4400 or 800-697-2253.

Packeteer, Inc.

14