Quality of Service

In traditional time division multiplexing (TDM) networks, voice and video calls were dedicated resources from source to destination. Packet based networks use much more efficient statistical multiplexing mechanism that lacks the ability to provide an explicit guarantee to real-time communications (voice and video). Quality of service (QoS) is a network technology that allows administrators to pre-allocate resources to application classes when needed. Cisco’s class-based weighted fair queuing (CBWFQ) and low-latency queuing (LLQ) technologies allow idle application class bandwidth to be used by non-idle bandwidth. The sharing of resources during periods of non-congestion allows individual application class bandwidth usage to spike up to the full bandwidth available on the link. During periods of congestion, application classes are guaranteed minimum bandwidth resources. All QoS tools can be summarized into the following categories outlined in RFC2475 “An Architecture for Differentiated Services” and expanded in the Enterprise QoS SRND: • Classification and Marking • Policing and Re-Writing (markdown) • Dropper • Shaper • Queuing • Link Specific Mechanisms Classification is the most fundamental building block of quality of service. Classification will inspect traffic as it enters the network. Depending on the application class in which the traffic belongs, the traffic will be marked with a technology specific marking so devices can treat the traffic differently than other traffic. Access control lists (ACL) is normally used on access layer switches to inspect data traffic as it arrives into the network. Applications can normally be identified by the transport layer (UDP or TCP) port(s) the application uses. Network Based Application Recognition (NBAR) is a classification mechanism with ability to look up to 400 bytes into the application layer header to identify traffic. Packets belonging to a particular application class are marked in alignment with that class configuration. Marked packets can then be distinguished by QoS mechanisms so the packet belonging to different application classes are treated differently. Marking mechanisms vary by the technology being used in the network. Marking mechanisms can occur at both layer 2 (data-link) and layer 3 (network) of the OSI reference model (OSI-Rm). The marking mechanisms include the following: • Ethernet 802.1Q and Cisco ISL Trunking • IP Precedence • Differentiated Services Code Point (DSCP) • MPLS Labels • Resilient Packet Ring (RPR) CoS • Frame relay • ATM In the next blog we will begin to explore markings in more detail. REFERENCES: Enterprise QoS SRND: http://www.cisco.com/en/US/docs/solutions/Enterprise/WAN_and_MAN/QoS_SRND/QoS-SRND-Book.html QoS Design Recommendations for Media Ready Networks: http://www.cisco.com/en/US/docs/solutions/Enterprise/Video/qosmrn.html Network Infrastructure Chapter of CUCM SRND: http://www.cisco.com/en/US/docs/voice_ip_comm/cucm/srnd/7x/netstruc.html QoS Section of Telepresence SRND: http://www.cisco.com/en/US/docs/solutions/Enterprise/Video/tpqos.html An Architecture for Differentiated Services http://www.ietf.org/rfc/rfc2475.txt Network Based Application Recognition http://www.cisco.com/en/US/docs/ios/12_1/12_1e11/feature/guide/dtnbarad.html

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.

Copyright © 2008 IDG Communications, Inc.