QoS Controls

QoS controls can be segmented into several categories: traffic authorization, traffic modification, and traffic adaptation. Traffic authorization controls a station user’s access to resources within a domain of control. Traffic authorization methods include admission control, eligibility control, and application control. These are forms of restriction that allow traffic only if a station user provides a password, the station user is on an access list, or the station user is permitted to do so by a policy management server. Traffic modification controls the type of traffic on the network through classification (segregating traffic into different classes), shaping (smoothing out traffic peaks to avoid overload situations), or policing (dropping traffic that doesn’t respect policies). Traffic adaptation methods include protocol control, path control, user behavior, congestion avoidance, and congestion management.

There are several commonly used QoS mechanisms that are supported by most of current IP-PBX systems. The two most common class of service (CoS) mechanisms are IEEE 802.1p/Q tagging (Layer 2) and type of service (ToS) prioritization (Layer 3). Both provide prioritization but have their limitations. A better mechanism, developed by the IEEE’s IEFT, is differentiated services (DiffServ), an advanced architecture of ToS.

802.1p/Q

The IEEE 802.1p standard for QoS prioritization is a specification defining 3 bits within the IEEE 802.1Q field in the MAC header (OSI Layer 2). The 802.1Q was designed originally to support VLAN operability and then extended to support traffic priorities. IEEE 802.1p adds 16 bits to the Layer 2 header, including 3 bits that can be used to classify priority (the tag). Frames with 802.1p implementation are called tagged frames. The standard specifies six different priorities, which do not offer extensive policy-based service levels. Typically, a NIC card in a LAN system sets the bits according to its needs, and Layer 2 switches use this information to direct the forwarding process.

If multiple LANs are interconnected by routers (Layer 3 switches), then the Layer 2 bits must be used to drive Layer 3 QoS mechanisms. The 802.1p/Q mechanism does not operate on an end-to-end basis in an internetwork but does provide a simple method of defining and signaling an end system’s requirements within the entire network environment. The Layer 2 header is read only at the switch level—the boundary routers, where traffic congestion occurs—and cannot take advantage of prioritization based on 802.1p unless it is mapped to a Layer 3 prioritization scheme. Even though prioritization is achieved within the switched network, it is lost at the LAN/WAN boundary routers.

Another potential problem is installing a LAN switch supporting 802.1p in a network with non-802.1p switches, which could lead to instability: older switches would misinterpret the unexpected 16 bits specified by the standard. Implementing 802.1p in older networks could require a costly upgrade of all switches.

IEE standard 802.1D is also supported by some IP-PBX systems for traffic prioritization. IEEE 802.1D extends the concept of MAC bridging to define additional capabilities of bridged LANS: to expedite traffic capabilities in support of the transmission of time-critical information in a LAN environment and provide filtering services that support the dynamic use of Group MAC addresses in a LAN environment. IEEE 802.1D Spanning Tree Bridge Protocol is a widely used bridge standard for interconnecting the family of IEEE 802 standard LANs. In this standard, a shortest path spanning tree with respect to a predetermined bridge, known as a root bridge, is used to interconnect LANs to form an extended LAN. The spanning tree defines a unique path between a pair of LANs, but this path may not be a shortest path. Moreover, because only one spanning tree is used, some bridges and some ports may not be used at all.

ToS

ToS was first defined in the early 1980s but largely unused until recent IP traffic bottlenecks at the boundary routers required prioritization for better service levels. The IPv4 protocol always contained an 8-bit field, called the ToS field, originally intended for use in packet prioritization. The most recent version, called IP Precedence, is a control mechanism that provides end-to-end control of QoS settings. The ToS octet in the Ipv6 header includes three precedence bits defining eight different priority levels ranging from highest priority for network control packets to lowest priority for routine traffic. Three of the ToS bits are used to flag sensitivity to delay, throughput, and packet loss. Many boundary routers and ToS-enabled Layer 3 switches read the precedence bits and map them to forwarding and drop behaviors. Devices use IP Precedence bits, if set, to help with queuing management.

Differentiated Services (DiffServ)

An evolving IETF QoS control mechanisms is known as DiffServ. DiffServ will not be based on priority, application, or flow, but on the possible forwarding behaviors of packets, called per-hop behaviors (PHBs). DiffServ is rule based and offers a control mechanism for policy-based network management. The DiffServ framework is based on network policies because different kinds of traffic can be marked for different kinds of forwarding. Resources can then be allocated according to the marking and the policies. The IETF Working Group is completing a series of standards that redefine Ipv6 ToS bytes, renamed the Differentiated Services Code Point (DSCP). The new byte indicates the level of service desired and maps the packet to a particular forwarding behavior (PHB) for processing by a DiffServe-compliant router. The PHB provides a particular service level (bandwidth, queuing, and dropping decisions) in accordance with network policy.

Under DiffServ, mission-critical packets could be encoded with a DSCP that indicates a high bandwidth, 0-frame–loss routing path. The DiffServ-compliant boundary router would then make route selections and forward the packets accordingly, as defined by network policy and the PHBs the network supports. The highest-class traffic would get preferential treatment in queuing and bandwidth, and the lower class packets would be relegated to slower service.

The DSCP is 6 bits wide, allowing coding for up to 64 different forwarding behaviors. The DSCP replaces the older ToS bits, and it retains backward compatibility with the 3 precedence bits so that non–DS-compliant, ToS-enabled devices will not conflict with the DSCP mapping.

There are currently two standard PHBs, expedited forwarding (EF) and assured forwarding (AF). EF has one codepoint (DiffServ value), minimizes delay and jitter, and provides the highest level of aggregate QoS. Traffic that exceeds the traffic profile is discarded. AF has four service classes and three drop-precedences for each service class (12 total codepoints). Excess traffic is not delivered with the same level of probability as traffic within the defined profile, and it may or may not be dropped. DiffServ assumes the existence of service level agreement (SLA) between networks sharing a border. The SLA establishes policy criteria and defines the traffic profile.

Other QoS control mechanisms include RSVP, subnet bandwidth management (SBM), and multiprotocol label switching (MPLS). RSVP was used by the first generation of client/server IP-PBXs but is deemed too complex, with too much overhead for many parts of the network. SBM is concerned with layer protocols above Layer 2 for internetworking between multiple LANs. MPLS is used primarily for private network routing applications, with limited appeal for premises-only communications applications.

Another approach to IP telephony is the use of VLANs. VLANs can provide more efficient use of LAN bandwidth, are used to distribute traffic loads, and are scalable to support high-performance requirements at a microsegment level. Traffic types, such as real-time voice and delay-insensitive data, can be segmented. IEEE 802.1Q is used as a VLAN packet tagging standard.