With its latest release of the MoCA 1.1 specification, the Multimedia over Coax Alliance has been able to achieve increased throughput performance and incorporate a novel approach to enabling quality-of-service over the coax home network. These spec extensions enable in excess of 175 Mbps of data throughput at the MAC layer and with the new Parameterized QoS (PQoS) method, critical traffic flows, like video streams, can be assured the bandwidth they require. These two key enhancements are designed to address the growing need for more bandwidth within the home LAN and ensure that video traffic is never compromised during situations where heavy, but less time-critical, traffic arises.

MoCA background
Designed for multimedia networking within the home, MoCA crafted a standard that specifically considered the requirements necessary to support real-time streaming of entertainment media. Network technologies that are primarily concerned with providing PC connectivity, Internet access and file sharing capability do not necessarily need nor meet the stringent requirements that a multimedia network does.

Rapid consumer adoption of high-definition television is driving the need for throughput large enough to support multiple high-definition video streams. However, a multimedia network must also support low latency, low error rates, quality-of-service and reach all of the home's coax outlets in order to fulfill the expectations of system operators and consumers. The original MoCA 1.0 spec was targeted to achieve approximately 100 Mbps of aggregate MAC throughput. The newest additions to the specification enable MoCA to achieve 175 Mbps with only MAC layer changes. This increase in MAC efficiency is achieved by aggregating multiple Ethernet packets into a single MoCA frame.

With streaming video data, any extended delays in the delivery of data packets can potentially lead to video decoding artifacts that spoil the viewing experience. Thus, the network requires a QoS mechanism that protects video content from the side effects of heavy network usage. The brute force approach to ensuring QoS is to simply increase bandwidth such that data flows never experience a condition of limited bandwidth. However, increasing bandwidth for QoS concerns is not a very efficient use of network resources and is just a stop-gap measure as applications are continually increasing their needs and will eventually consume this additional bandwidth. Relying on just adding more bandwidth also doesn't take into account that simple network file transfers can consume all of the bandwidth in transient bursts. System operators need a QoS mechanism that protects their premium content, yet is flexible enough to work in a network environment that includes QoS and best-effort traffic.

MoCA 1.0 supports a prioritized traffic scheme to enable QoS data flows simultaneous with best effort traffic. The MoCA 1.1 spec takes QoS one step further by introducing parameterized QoS. Whereas prioritized QoS simply ranked data packets by priority relative to one another, parameterized QoS operates by reserving the actual bandwidth required to ensure that peak data rate transfers are always accommodated over the network.

Packet aggregation
Depending on the applications being used, a mixture of small and large packets may be found traversing the home network. Traditional applications such as file downloads, heavily buffered web video and web page refreshes can use large packets that are transferred from the wide area network (WAN) in long back-to-back bursts. Applications that require real-time data flows like interactive gaming typically use small IP packets transmitted at regular intervals. Even the TCP/IP protocol makes use of small packets in the form of ACKs, of which there may be several within a 1 millisecond period.

View full size
Figure 1 shows a simplified example of a MoCA packet aggregated frame.

In addition to the actual payload, MoCA requires transmission of its own control and header information that serve to delineate the start of a packet and describe the contents of its payload. This additional data is comprised of PHY preambles and MAC headers that do not usually vary in size for any given payload length. By improving the ratio of payload to control/header data using packet aggregation, the efficiency of the network's bandwidth can be maximized. Packet aggregation combines multiple Ethernet packets into a single MoCA MAC frame and greatly increases the ratio of useful payload data to header information. Thus, more network bandwidth is used for application data rather than header information, resulting in improved packet throughput.

Additionally, the overall inter-frame gap (IFG) time and number of reservation requests are reduced since packets that have been aggregated no longer need their own individual MoCA frames. For example, a MoCA frame that has six aggregated packets saves 5*10 = 50 microseconds of IFG time.

Next: Throughput results

Throughput
Throughput results from Entropic Communications' tests are shown in Figure 2 and illustrate the dramatic improvement in packet-per-second throughput that packet aggregation can yield. Additionally, the MoCA network coordinator node typically has better throughput performance for all packet sizes, but with packet aggregation, client node performance can come close to matching the NC performance over all packet sizes.

On a percentage basis, packet aggregation dramatically improves small packet throughput on client nodes the most, but large packets will experience throughput increases as well. Multiple full-size Ethernet packets can be combined together or in a combination of small and large packets to share the overhead of a MoCA frame destined to the same MoCA node. This combination approach is especially useful when using MoCA for a WAN connection where virtually all packets can be aggregated for reception by a MoCA broadband router.

View full size
Figure 2: Packet Aggregation Throughput Improvement (x = Packet Size, y=% Improvement)

In order to keep packet aggregation from impacting higher layers of the network stack, the MoCA 1.1 implementation relies only on MAC layer extensions. Packet aggregation is completely transparent to the higher network layers and is an autonomous function within a MoCA 1.1 node. Thus, packet aggregation is automatically performed within MoCA 1.1 devices and only occurs between nodes that are capable of aggregation.

During the admission process, a MoCA node declares its capabilities to the network coordinator, which in turn broadcasts those capabilities to the other nodes in the network. Because the network coordinator is not involved in the packet aggregation decision process, packet aggregation can occur even in networks where the network coordinator is not aggregation capable. This feature allows operators to incrementally deploy packet aggregation capable devices into existing MoCA 1.0 networks and utilize this advanced feature at a later date when the rest of the network devices are upgraded to support MoCA 1.1.

Parameterized QoS
Increasingly, entertainment media is entering into homes as digitized packets from multiple avenues and ends up being stored on various devices scattered throughout the home.

• Television programs are streamed through set-top boxes and stored on DVRs typically located in the family room or bedroom
• Music and video is downloadable from the Internet, residing on PCs, game consoles and media adaptors, located in home offices, kid's bedrooms and family rooms
• Lastly, mobile carriers are ramping up their entertainment services where music and video are stored on the consumer's handset

However, all these devices are typically isolated from one another and the consumer is unable to access their content from different devices in another room - especially if that content has to cross the traditional system operator to PC/CE boundary.

The convergence of data-centric traffic with multimedia flows is necessary in order to support a more enriching usage model where consumers can watch their DVR shows, purchase VOD movies, access the Web from their PCs and play console games over the Internet from virtually any room in the home. System operators face the challenge of supporting these new multimedia usage scenarios simultaneous with asynchronous data traffic like Internet access and coexisting with premium services such as digital TV. Critical to effectively serving all these data flows is a method to ensure that each application is guaranteed the bandwidth and latency requirements necessary to provide a satisfactory user experience.

Next: Parameterized flows and asynchronous flows

Parameterized flows and asynchronous flows
MoCA 1.1 adds parameterized quality-of-service (PQoS) to ensure that asynchronous, best-effort data does not interfere with the assured delivery of time sensitive multimedia data streams.

The overall MoCA bandwidth can be thought of as being divided into two parts -- parameterized flows and asynchronous flows. The maximum bandwidth available for parameterized flows can be set to a certain percentage of the overall network bandwidth. The remaining bandwidth is used for all asynchronous traffic, including prioritized traffic, best effort traffic and link maintenance.

The MoCA PQoS architecture provides a means for OEMs and operators to either leverage the early work of the Universal Plug and Play 3.0 (UPnP) parameterized QoS model or create a custom PQoS solution using proprietary higher layer QoS services.

MoCA 1.0 Priorities The MoCA 1.0 spec supports 802.1p VLAN priorities as a method to prioritize traffic on the network. However, this approach does not make distinctions between types of data nor does it guarantee bandwidth. It only provides the network with a method to weigh the importance of one packet's priority against that of another packet's priority. Asynchronous data tagged with high priority would be treated the same as video data tagged with high priority. The result could still lead to a situation where an operator's high-value video service is adversely impacted by regular data. Even if multiple video streams are tagged as high priority, a router/switch will simply queue up the packets in its high priority buffer to be sent as soon as possible. If network bandwidth is insufficient to handle this amount of high priority traffic, packets will inevitably be dropped.

The MoCA 1.1 specification includes a parameterized quality-of-service (PQoS) feature. PQoS allows a video flow to precisely define its bandwidth needs, reserving just enough resources to guarantee robust delivery. Just as importantly, PQoS provides the intelligence to know if there is not enough network bandwidth available to support an additional video flow, thus avoiding situations similar to prioritized QoS where prioritized packets are queued up appropriately, but still cannot be delivered in a timely manner due to lack of bandwidth.

Additional extensions
Other enhancements to the MoCA spec include:

• Doubling the number of MoCA devices supported on the network
• The ability to predetermine the Network Coordinator
• The ability for any MoCA node to gather network statistics from any other node

To "future-proof" and enhance the robustness of a MoCA network, the number of devices that can be supported on a single network has been increased to 16. When you imagine all the possible devices where multimedia content could be stored, transferred or played back -- like set-top boxes, gateways, routers, bridges, PCs, game consoles, DVRs and televisions -- there exists ample opportunity for more than 8 devices in MoCA home entertainment networks.

Next: One-way traffic

In instances where traffic is known to flow predominately in one direction, like a downstream WAN link, there is a desire to be able to select what MoCA node will act as the network coordinator (NC). Any node that is not the NC must first submit a request in order to be granted time on the wire. This step is eliminated for the NC since it essentially schedules its own packets, resulting in a lower latency. Lower latency means the NC transmitter can put data on the network faster resulting in higher throughput.

Lastly, system operators are beginning to take on the challenge of providing more than just a single CPE within their subscriber's home. Increasingly, they will be providing a home network with multiple MoCA enabled devices that operators are responsible for maintaining. MoCA 1.1's Full Mesh Rate Transaction feature allows the operator to use a single node to gather PHY rate statistics from other nodes and, thereby, facilitate monitoring the health of the entire home network and enabling remote diagnostic capabilities.

Summary
The specification extensions of MoCA 1.1 are designed to address the growing needs of system operators for increased throughput, a more intelligent QoS method, larger network capacity and easier network monitoring. All these features attempt to wring out the best performance and increase the robustness of MoCA networks with minimal cost impact.

MoCA has been designed from the start with multimedia room-to-room networking in mind as the target application. With the continuing evolution of MoCA standards, system operators and consumers alike can look forward to a network technology that will keep pace and scale with services and usage scenarios yet to be imagined.

About the author
Jon Iwanaga has over 15 years of experience in the development of semiconductors for graphics, video and communications systems. Early on at Brooktree Corp., Jon worked on graphics chips that first brought color to personal computers and workstations. At Conexant Systems, he worked on cable modem ICs along with video and tuner chips that enabled the capture of video signals on personal computers. Currently, as technical marketing manager at Entropic Communications, he handles the technical aspects of the company's c.LINK products for home networking. Jon Iwanaga holds a B.S. Degree in Electrical Engineering from UC San Diego and Masters Degree in Business Administration from UC Irvine. He can be reached at jon.iwanaga@entropic.com.