|
The prevailing perception in broadband wireless is that as adoption grows, so does the need for guaranteeing quality of service (QoS). The issue of QoS, therefore, has become a critical area of concern for suppliers of broadband wireless access (BWA) equipment and their customers. Enforceable QoS is an essential foundation for widespread acceptance of broadband wireless, since it allows for more efficient sharing of the operator's infrastructure as demand for capacity increases with subscriber take-up.
New 802.16 broadband wireless products are breaking the QoS "gridlock" that has beleaguered wireless deployments to date. What 802.16 brings to the QoS equation is a number of unique QoS parameter sets that guarantee levels for throughput, latency, and jitter. This enables service providers to offer flexible and enforceable QoS guarantees — a benefit that has never been available with other fixed broadband wireless standards. This article will discuss the evolution of QoS in broadband wireless and the specific design elements of the 802.16 standard with respect to QoS.
The QoS Evolution
QoS directly affects the ability of providers to meet the service level agreement (SLA) requirements demanded by customers. To date, this has been difficult to achieve with most wireless options.
By way of definition, QoS metrics measure loss, throughput, delay and delay variation (also known as jitter), sequencing, and total errors. These metrics are equally applicable at the data link layer and the network layer, and have been generally adopted by industry groups for testing QoS over networks such as ATM, Frame Relay and IP.
Defining metrics helps network operators to match the subjective quality statements from subscribers to actual, measurable network conditions. Automating the collection of these metrics allows network operators to evaluate the QoS of a network before, during and after the network goes live.
Much of ATM's success is directly attributable to its QoS guarantees. ATM's attempt to deliver QoS is defined by different classes of service based on constant and variable bit rates. These classes are:
- Constant bit rate (CBR) — In CBR, traffic is characterized by a continuous stream of bits at a steady rate. This class is for low-bandwidth traffic that is highly sensitive to delay and intolerant to cell loss.
- Variable bit rate (VBR) — VBR applies to voice or video applications that use compression. Within this class are real-time VBR (RT-VBR) where real-time end-to-end delivery is critical and non-real time VBR (VBR-NRT), where delay is less critical.
- Best effort services — These include available bit rate (ABR) and unspecified bit rate (UBR) and apply to LAN traffic that is more tolerant of delays and cell loss. Since UBR is subject to increased cell and packet loss and does not specify bit rate or traffic parameters, it has no QoS guarantees. ABR is a managed service based on minimum cell rate (MCR) and has a low cell loss. Neither ABS or URB service classes offer delay variation guarantees.
When it came to delivery of real-time services over wireless, however, the QoS challenge was considerable. In the early stages some vendors integrated ATM technology into their systems. While this addressed the QoS needs, it also made equipment prohibitively expensive. Others chose to focus on an Ethernet or IP packet transport with prioritized delivery over wireless. In this case, service level parameters were limited to two classes: high and low priority. As needs grew, we saw the evolution of more customized parameters that drove a product's ability to accommodate more than these two classes of service.
Most broadband wireless networking systems have been based on the 802.11 standard — one that is absent of QoS capabilities and limited to proprietary solutions. Since it uses the contention-based media access control (MAC) layer, 802.11 is inefficient when it comes to allocating bandwidth and scheduling traffic, and therefore cannot meet the required QoS standard that users have come to expect in the ATM world.
One of the first attempts to refine parameter definitions in the LAN world occurred with the development of the 802.1Q/p standard, which allowed for the tagging of seven different priority levels. The 802.1Q/p standard improved upon the inefficiencies of 802.11 by providing priority handling of tagged traffic. Unfortunately, cost effective components supporting the 802.1Q/p standard were not developed with the ability to prioritize to multiple levels, often providing only two queues: high and low. While this provides some level of QoS guarantees to the high-priority traffic it offers little in terms of traffic enforcement.
The inability — or unwillingness should we say — for developers to increase the number of service queues was a matter of technology and financial resources. From a design perspective, multiple service queues represent an expensive proposition. Each queue requires separate memory to interrogate packets, increased processing power and multiple state machines for maintaining timing, patterns and tagging.
Authors of 802.16 standards recognized the value of the ATM service model and incorporated four analogous classes of service, each of which can be tailored to distinct application flows. Let's take a closer look at these classes.
802.16 QoS Features
The demand for QoS in wireless has been addressed by the IEEE 802.16 standard. What specifically sets this standard apart is a polling-based MAC layer that is more deterministic than the contention-based MAC used by 802.11. 802.16's MAC layer enables classification of QoS and non-QoS dependant application flows and maps them to connections with distinct scheduling services, enabling both guaranteed handling and traffic enforcement.
Each connection is associated with a single scheduling data service and each data service is associated with a set of QoS parameters that quantify aspects of its behavior. The following are the four types of scheduling service supported by 802.16:
- Unsolicited Grant Service (UGS)
- Real-time Polling Service (rtPS)
- Non-real-time Polling Service (nrtPS)
- Best Effort (BE)
The UGS service provides the most stringent scheduling, maintaining guarantees on throughput, latency, and jitter to the levels necessary for time division multiplexed (TDM) services. This scheduler is analogous to ATM's CBR.
The rtPS service also provides guarantees on throughput and latency but with greater tolerance on latency, making it suitable for video conferencing applications. This scheduler is analogous to ATM's RT-VBR service.
The nrtPS service provides guarantees in terms of throughput only and is therefore suited to mission critical data applications that are not latency dependent (e.g. e-mail). This scheduler is equivalent to ATM's VBR-NRT technique.
|
