As with any project, content transport network design and construction is not successful without some amount of performance criteria. Content transport networks are different than ordinary voice and data networks. Ordinary voice and data networks don’t typically do content transport well. On the other hand, design and build a network capable of transporting valuable program content, and voice and data can come along for the ride. Below is a list of considerations and criteria that can be used when preparing to undertake a content transport network project. The following list is not to be taken literally, nor is it exhaustive or all-inclusive:
§ Access, switching, and transport elements
§ Access and transport facilities can be terrestrial, satellite, or a combination
§ Switch facilities will be time (TDM), cell (ATM), or packet (IP)
§ Service availability is full-time 24/7 or shared
§ Availability, reliability, robustness, grade, and quality of service
§ Capital and operating cost
§ Geographical or physical coverage includes local (LAN), metropolitan (MAN), regional, national, and global turf (WAN)
§ LANs may have single or multiple segments covering a room, floor, building or group of buildings in a campus arrangement
§ A MAN typically involves third party telco or ISP service and uses standard telephone facilities, such as E1/T1, E3/DS3
§ WAN extends LAN and MAN to wider geographic areas not covered by local telephone companies and ISPs
Content transport networks can be built or bought, but practical realization is a combination of buying equipment and the rights to use facilities and services.
Live, streaming content requires continuous, uninterrupted connections with an equal amount of bandwidth. That’s the theory; however, in practice it’s always prudent to leave just a tad of headroom. So how much is a tad? Practicality drives such in the form of how the service provider divides up the bandwidth and sells it. For example, a 10 Mbs ATM or IP network facility likely won’t be precisely 10 Mbs. These animals usually break out in increments of octal numbers. So somewhere around 10 Mbs will be something like 10240000. If that is your choice of network transport channel, then the compression system output bitrate should be set at some number less than the channel rate. This parameter is also a victim of practical circumstances as well because these devices commonly have to deal with octal numbers. So a tad in practice happens to be the difference between the highest speed the encoder can be set at, and the channel rate. (See Appendix II for an example of calculating payloads and matching channel rates.)
Non–real-time content can be transported using continuous, uninterrupted connections, but it can also be carried on discontinuous bandwidth connections, usually at lower cost and improved utilization of the facilities. Be aware that realization of lower cost is dependent on obtaining use of facilities and services at unit prices based on time used and type of bandwidth occupied for each session or transmission just like the old fashioned long distance telephone call.
Standard network performance and characteristics must be understood before they can be applied to content transport networks. The next few paragraphs provide an introduction to time division multiplexing (TDM), ATM, and IP network technology.
TDM technology characteristics and performance are the standard cell and packet based network performance should be measured against. If a standard for TDM is required, then use wire, fiber, or another passive conductor of known performance. The characteristics of interest include available channel bandwidth, bit error rate, and jitter. However, in cell and packet networks, bit errors cause cell and packet loss or impairment, as can jitter.
ATM transport technology offers 5 classes of service. Constant bit rate (CBR), variable bit rate—real-time (VBR-rt), variable bitrate— non–real-time (VBR-nrt), unspecified bit rate, and available bit rate. While it may change in the future, ATM CBR is currently the only ATM class of service capable of transporting high-quality, high bit rate content in real time.
Packet-switched networks are inherently chaotic unless specifically configured to deal with continuous signal, or mixed-signal traffic and class-of-service. Packet networks are either Ethernet or IP. (Several packet or packet-like techniques exist; however, they only support content transport as a file transfer, not real time.)
In general there are two types of IP technology and methodology: Ethernet and Internet. The IEEE 802.1 standard defines Ethernet. Internet or more precisely, IP is defined in RFC 791. Ethernet transport of IP is defined in RFC894.
Ethernet architecture is built around shared media in the form of common set of cabling where the information is carried in packets, and the device such as a workstation or server listens or monitors the buss before attempting to establish a connection or session. The way the process works, end-to-end, has the sender and all the receivers constantly listening or monitoring the buss. A session is kicked off after a sender sends an initial transmission to all stations using a unique address. If the initial transmission has a valid destination address, that is an actual receiver connected to and listening to the buss, it responds with an acknowledgement. After the sender receives the acknowledgement, then and only then do the two computers establish a connection and carry on with the session using their unique address information.
IP networks, the Internet in particular, behave in similar fashion as Ethernet.
All these types of transport work well for moving files, including hypertext markup language—coded pages, fixed images, and other static objects. Uncongested networks may even support low volume continuous signals such as produced by voice or telephone service over IP, and even ‘‘work okay’’ with higher bandwidth continuous signals. Make no mistake about it though, unstructured networks cannot be relied on for transport of continuous signal, high bit rate, valuable content such as audio, video, closed captioning, control, or other signals associated with, or embedded in, program content.
Reliable, predictable, safe, and secure content transport requires network connections with sufficient bandwidth, grade, and quality of service (GOS, QOS). Even non–time-sensitive or non–real-time transport—so-called FTP—should be planned and implemented with care because of the size of the files and the time required to move them have significant economic implications.
Obtaining sufficient bandwidth, GOS, and QOS is a matter of specifying and configuring LAN, MAN, and WAN network resources.
Sufficient, continuous bandwidth means the network must exhibit bandwidth equal to or greater than the bandwidth of all traffic, not just program content if the network is required to accommodate email, web surfing, network management, and perhaps voice. Insufficient network bandwidth results in denial of service or, at best, delayed service. Program content payload bandwidth is roughly equivalent to the sum of compressed audio, video, and other signals multiplexed into a program stream or included in a file object stored on the system. When more than one real-time stream is present on the interface simultaneously, the aggregate of all the program streams cannot exceed the bandwidth available on the interface points of the sending and receiving systems and the network connecting the systems. In other words, the bandwidth of the sending and receiving systems must equal or, preferably, exceed the aggregate of all traffic.
GOS means the network connecting all workstations and servers must be available to all users within the design limits agreed to or promised to its users. For example, telephone network services use statistical probability based metrics to define and measure GOS level, inside and outside the network. A P.01 GOS means the network is designed and performs, or doesn’t perform, within the limits of probability that the network will enable the user to complete the call in 99 of 100 attempts. This model can be applied to workstations, servers, and a LAN, MAN, WAN or combination of all and will perform satisfactorily 99 of 100 times when someone wants to transfer a file, or set up and use a connection to deliver streaming content originating on a server platform and terminating in one or more peer platforms at other locations or interfaces served by the network.
QOS means that the quality of the connection in terms of bandwidth, bit-error rate (BER), jitter, packet loss or any other parameter the payload may be sensitive to, is of sufficient level to support program content transport between and amongst the service points. The basic model for this category of network is classic TDM facilities found in ANSI/ITU standards-based networks. The acid test of performance is measurement and comparison to TDM private line facilities such as E1/T1, E3/DS3, and OC3/STM1. A good question of network equipment, facilities, and service suppliers is: Can you emulate T1, or DS3, etc.? The right answer is not ‘‘Yes.’’ The right answer is, ‘‘You can expect jitter, packet loss and bit error rate performance of x, y, and z. This compares to TDM emulation performance of x, y, and z.’’ Then you can decide if the differences fit into your required performance and compare one supplier to another.
Ingest, play out, and file transfer of program content as promised in many product and service descriptions require network connections with sufficient bandwidth, GOS, and QOS. Even non–time-sensitive or non–real-time transport—so-called FTP—should be planned and implemented with care because the size of the files and the time required to move them have significant economic implications.
Standard, so-called out-of-the-box or plug-and-play default LAN configuration included with recent generation Microsoft Operating systems (OS; W2000 Workstation & Server; WXP) enable non– real-time or FTP program content transport. Connect Ethernet to a network interface card (NIC) with access to the Internet, install the OS, run the Internet wizard, and voila! Instant success. No further fuss or effort and file transfer across the Internet from one host to another is possible.
These operating systems also permit configuration of an NIC to enable QOS as specified in IEEE 802.1p, a method whereby packets carrying continuous content can be marked and differentiated so LAN segments can isolate and protect the content from the effects of congestion and chaos mentioned above. Ethernet packets mapped to IP enable QOS marking to be passed to the IP network. If the network has differentiated services capability, real-time content transport across the network is possible. Older operating systems (NT 4; 95/98) do not include 802.1p/QOS capability.
Two types of connections are possible, and both may be required by the application. These include Unicast, or point-to-point, and multicast, or point-to-multipoint. These types of connections enable single or multiple deliveries of files or streams, sometimes referred to as objects.
The basic elements of a content transport network include customer premises equipment (CPE), access facilities at each location, and backbone transport in between. The equipment must be selected and configured to support the level and type of traffic. For example, if the traffic is program content only, that’s one set of circumstances. If the network is to carry voice, data, and provide Internet access, that’s another. If the network is to carry multiple types of traffic, the equipment and facilities will have to be structured to accommodate it. Figure 1 shows a general reference architecture capable of supporting voice, data, and content transport.
Figure 1: Premises Equipment Architecture
There are several characteristics of the architecture that should be pointed out and commented on. First, note the presence of a network clock reference and a separate station synchronizing reference. Neither has anything to do with the other and that’s the point. The network clock reference is to make sure the network is stable and jitter-free because it must carry the embedded program clock reference along with the content. After all, if the network isn’t capable of carrying the program clock reference to a satisfactory degree of accuracy, then the content will suffer impairment.
Although there appears to be a single-thread router and network interface, this is purely symbolic, and emblematic of the same level of redundancy as implied in the private branch exchange, LAN router, and Moving Picture Experts Group (MPEG) Codec. Resolving reliability, robustness and network performance concerns may require redundant equipment and facilities, with emphasis on content value and specific traffic levels. The terms and symbols are generic and intentionally chosen to cover several alternatives without stating them implicitly. For example, any new facility design should take a serious look at voice-over IP telephone service. New installations or even replacement/upgrade installations, may find economic advantage in fully integrated voice and data on LAN wiring. And although, not likely, it may be more appropriate to use ATM switching and transport for real-time program content than IP or TDM transport.
On the network side, there are similar issues and concerns; however, they must be addressed with carriers or service providers instead of manufacturers of equipment. As a design exercise, network access, transport, and switching should logically follow the food chain whereby the network facilities support movement of content within and between the creation, distribution, and delivery sections of the model. For example, moving raw, unedited content from a location to an editing facility, or moving finished program material from the post-production facility to a network operations center. And of course there’s the end link, which requires the content to be moved from anywhere else to cable head end, DBS uplink, Internet access facility, or digital television transmitter input.
Figure 2 is a network topology diagram showing the details of how the basic elements fit into an overall architecture serving users located at separate sites, or operating centers.
Figure 2: Reference Architecture
All the various elements must be specified and priced out in detail. CPE is a capital investment. Access and backbone transport is an operating expense and can be provided by third parties, such as Internet service providers (ISPs), ILECs, CLECs, or inter-exchange carriers. Obviously, it is advantageous to deal with a single source for these services. Decisions on the end-to-end solution should only be made after following a due diligence process. Building the simplest of networks is not easy. Scaling start-up or small networks to larger networks becomes geometrically more complex. Churn and change after a network is built, debugged, and operational can be risky and should not be attempted without careful planning and deliberate, task oriented, sequential steps.
Similar to the end-to-end service model, the reference architecture simply lays out the functional components and shows how they relate. The NID or premises equipment interfaces and interoperates with the network to set up and tear down connections, monitor performance, and process alarms. The desired content transport network leverages one or more routing, switching and transport capabilities, depending on requirements and configuration of the access facilities. In situations where multiple types of traffic are converged onto a common access facility, the access facility must be channelized and mapped to the particular transport. For example, voice grade dialup or switched service would have channel capacity sufficient to accommodate peak voice demand on the public switched telephone network (PSTN) or integrated services digital (ISDN) network. However, if the design called for voice-over IP, some amount of bandwidth would be required to accommodate a similar level of voice traffic.
CPE is a router configured to connect to peer routers at the other locations. The router must be sized and have features selected to perform the functions required by the servers. These functions vary and depend on the encoded bit rate or payload of the content and level of traffic.
Another factor for careful consideration is distance between peer devices at other locations. If the distance is short, such as a nearby building, Ethernet could be an option. But outside adjacent buildings within a campus environment it’s likely construction and capital cost will quickly add up to make third-party service providers with IP transport capability attractive.
Access facilities provide basic connectivity between the CPE and the MAN or WAN. Likely alternatives include E3/DS3 or OC3/STM1. Choosing an appropriately sized access facility is a matter of making a conservative estimate of initial traffic level, then monitoring the traffic and adjusting capacity to levels consistent with acceptable utilization and growth plans.
Backbone transport varies based on requirements and usually comes with significant and critical services attached. For example, the access facility is a dumb, point-to-point, TDM unchannelized facility. But routed networks include services such as routing and configuration protocols, IP address provision, service configuration and management that depend on processing functions resident in edge and core routers. It naturally follows that the owner of the core backbone equipment should provide these services.
No comments:
Post a Comment