The discussion continues… My articles from Source Material editions: 3, 4 and 6 reviewed transport mediums assuming an unmanaged network flow, but what about the nodes?!
With transceiver distances dependant on the products we wish to use, how do we route this information efficiently? Developing an understanding of IT network workflows is becoming ever more required within the AV and broadcast market. Offering extended performance and multifunction integration over one unified infrastructure provides an appealing update in system design with comparison to more traditional analogue solutions.
In order to elaborate on efficiency it is important to first look at the workflow. IT networks typically comprise of multiple processors moving information from one client to another. This can be internal (LAN), such as a Dante stream from Input to output within a given physical space. Or an external (WAN) connection, accessing and interfacing with internet or multisite communication systems. Providing a computable link between two or more networked devices, the most basic network consists of a network Hub which broadcasts data to every Ethernet based device connected. Trading this hub for a Switch, data transmission is more specified, establishing isolated links between clients. Where a switch utilises Media access control (MAC) addresses a Router introduces Internet Protocol (IP) Addresses, becoming the intermediator for connectivity of multiple local area networks (LAN) creating the start of a wider area network (WAN) otherwise known as the internet.
From there it is just a game of interoperability and functionality; Connecting two dissimilar networks? You need a Gateway to convert data from one protocol or format to another. Want to provide centralised resources? Servers provided storage, processing and data exchanges where human intervention is not required. Want to add connectivity for mobile devices? An Access Point provides a bridge between wired and wireless connections. Want to protect against unauthorised access? A Firewall monitors and restricts network traffic into and out of the network. Don’t have a broadband connection? A Modem interfaces a WAN via a traditional telephone line. Although the network topology in a domestic setting is typically compiled into one device, it is clearer to treat its processes as individual components such as can be seen in many large commercial IT networks.
Digging below the physical, in the digital domain data presents as a 1 or a 0, a transistor on or off. Multiple logic states compile to convey information and packaged into the form of a datagram. Each datagram is then placed into a packet then each packet is further placed into a frame, with most signals from source to destination consisting of multiple transmissions. For this encapsulation and transmission there are a range of transport protocols for wrapping and addressing varying in their size and quality of service (QOS). To highlight a few in simple terms; UDP (faster but susceptible to loss) and TCP (slower but can retransmit loss) are both communications protocols used to exchange data between networked devices. TCP based network protocols include: FTP/S (establishes data as well as control connection “/secure”), HTTP/S (establishes data connection only “/secure”) and SMTP (server side mail exchange and relay). UDP based network protocols include: NTP (synchronisation of clock times), DNS (network address lookup resource) and DHCP (automatic IP assignment).
The bandwidth of any given network is reliant on limitations of throughput, error rate and QOS from nodes as well as propagation, interference and stability of the interconnecting transport medium. It would be agreeable to suggest bottlenecks are a regular, unavoidable occurrence for most networks. A sudden rush in traffic between two nodes exceeding the potential maximum processing limit, often results in delayed, interrupted, or complete loss in data transfer, dependant on the type of packetisation applied. For instance It would also be reasonable to assume that a gigabit switch would have provision for gigabit links on all ports. This is not true, yet rarely a cause of alarm. The bottle neck of these switches are the central processing units, each separation (typically every 4 or 8 ports) within a product will have its own CPU limited by its maximum bandwidth indirectly related to its throughput. Typically networks comprise of low and high data usage ports equalling out the performance requirements of a CPU. The take away from this is to review your physical configuration as well as your virtual one. Subtle separations between sets of RJ45 ports may be a good indication of segmented processing, there may be throughput efficiency available by spreading your data hungry connections across the switch rather than bunching them to one side/processor.
“QOS” covers a lot of potential of network service prioritisation and throughput resource management, more so present and required in larger network infrastructures, even domestic “dumb” unmanaged switches have the capabilities of incorporating some of these more sophisticated elements although typically not to the same extent. Different network services may present variable packet sizes potential inconsistently over time. Data from A to B may be for the most part things alike: websites, email, file transfers etc. the bandwidth of which is typically low and not for the most part time restricted, nodes can buffer and or resend loss with minimal impact to the end result. Adding a Dante stream to this connection will use some of the available space but is more time dependant than the previous services so much so that we do not want to disrupt the signal with something alike a large file transfer. For this traffic type we could prioritise a dedicated, fixed data rate virtual path for our data stream, leaving all other traffic with a shared best effort and queued partition of the total connection bandwidth. This understanding scales, not only to virtual path ways but to the physical, with VLANs (allocates services to partitioned and isolated node ports) and Trunking (increases the data rate of a transmission by coupling two or more ports together at each end of the connection) both tagging Ethernet frames under the IEEE 802.1Q outline.
Network topology also plays a key role in QOS and throughput, offering alternate physical pathways for vital and or specialist network resources. Point to point (PTP) is the simplest, connecting one node to another. Two or more nodes onto a single switch may be considered a STAR, all traffic goes through a single point/processor. Extending from a star the TREE interconnects via bus networks in a hierarchical like fashion. Building in redundancy the RING offers multiple path ways, dependant on traffic or connection stability packets can be redirected to reach the same destination. With MESH interconnecting all nodes, theoretically providing multiple path ways to all traffic. Though some redundancy can be beneficial mesh networks can use substantial resource routing, running protocols such as spanning tree (STP) in an effort to reduce the occurrence of duplication and loops, which could result in a flooded and potentially redundant network. Most IT infrastructures are a Hybrid of the above with application dependant service requirements.
Innovation of Ethernet within the broadcast AV industry presents in the two main areas: “point to point (PTP)” and “Packetized”. The success of Dante by Audinate, is a prime example of a packetized solution. Conforming to a typical IT network infrastructure, audio is sourced, digitised, and sent down the network then decoded and sampled to relay at its destination alongside the likes of UDP. PTP a simplest topology, in my mind this cheats this system, utilising the medium but not the standardisation. QTP audio is a good example with a more common example presenting as a range extender, Baluns make use of the complexity and affordability of Ethernet cabling, processing signals separate from a network, piggybacking on the success and availability of Ethernet cables with the added product protection of bespoke de/encoding signals and dedicated physical links. There are a wide variety of compatibility products on the AV market for bridging the gap between analogue and digital communication systems. Deploying widely available network solutions offers a cleaner and easier platform for a variety of production applications.