So the question then is, what is the best way to implement QoS in conjunction with application acceleration solutions? And what role should the WAN router play versus the application acceleration appliance? Finally, can enterprises effectively enforce QoS across all applications, including both accelerated and non-accelerated traffic?
To successfully deliver applications across a distributed enterprise, IT staff require a clear QoS strategy that is tightly integrated with application delivery requirements. An advanced application acceleration solution will provide a variety of QoS options to enterprises, giving them the flexibility to support numerous applications and traffic types in an optimized manner.
Implementing QoS
QoS involves two functions: Classification of packets into traffic classes based on source, destination and/or application; and enforcement of policies using queuing and service mechanisms.
QoS classification is typically performed either in the WAN router, or by an "upstream" device (e.g., host or Ethernet switch) and
Requires Free Membership to View
Application acceleration appliances introduce a challenge to the way that QoS is implemented. These devices sit on the LAN side of the WAN router, in both the data center and branch offices. As traffic flows from the internal network to the WAN, application acceleration appliance remove repetitive information, compress headers and payload content, modify IP addresses and port numbers, and sometimes even encrypt traffic. By completely obscuring the original data (and its headers), application acceleration appliances prevent downstream devices, such as WAN routers, from applying QoS classification logic based on normal packet inspection.
To account for this, QoS classification can be performed upstream of the application acceleration appliance (by a host, Ethernet switch or LAN router) and then honored by the appliance. Or, QoS classification can be performed by the application acceleration appliance itself. The latter approach is often more desirable, as enterprises sometimes do not have intelligent upstream devices in a branch office that are capable of performing QoS classification.
With respect to QoS queuing and service disciplines, most application acceleration appliances can pass existing tags to downstream WAN routers, enabling these devices to participate in the QoS process as they normally would. In addition, queuing and service disciplines can be enforced within the application acceleration appliance itself. This is often desired, because application acceleration appliances are well equipped to collect real-time metrics, like packet loss and delay, and adapt QoS techniques accordingly. (This is usually not part of a typical WAN router's feature set).
Conclusion
As enterprises move towards a centralized model for branch officer servers and storage, new solutions are required to improve application delivery. While these solutions increase bandwidth efficiency and improve perceived application response time, they do not eliminate the need for robust QoS capabilities. Certain types of traffic, such as voice and data, will almost always benefit from QoS when being delivered across shared resources.
When application acceleration solutions are deployed on both ends of a WAN link, special consideration needs to be given to how QoS is delivered. Traditional methods of performing QoS classification and enforcement in the WAN router are often no longer viable as application accelerations solutions modify headers and obscure payload information. To account for this, enterprises should ensure that their application acceleration solution provides different QoS options to accommodate unique traffic requirements. Fundamental QoS capabilities include packet marking, application classification, queuing and traffic shaping. In addition, QoS must be applied across all traffic – accelerated or otherwise. By satisfying all of these QoS requirements, enterprises can ensure consistent and reliable application delivery across all business locations.
Dr. Hughes founded Silver Peak Systems in 2004 and previously held senior architect positions with Cisco Systems, Stratacom, Blueleaf and Nortel. Dr. Hughes has a PhD in packet network optimization.
This was first published in January 2006
Network Management Strategies for the CIO

Join the conversationComment
Share
Comments
Results
Contribute to the conversation