Content transfer and process execution speed in broadcast operations impact the content assembly process and time-to-air. Moving large HD files around the facility and through the production process at greater-than-real-time rates open new possibilities in content assembly. In a sense, speed can bring new life and creative opportunities to the drudgery of indistinguishable on-air presentations.
Speed in the BOC relates to all layers of the infrastructure. On the physical, 3- and 10Gb/s SDTI standards are under development by SMPTE. On the media network layer, 10Gb/s is becoming a default uplink capability of switches, while greater than 15Krpm disk spin rates are being attained. Application layer compute platforms are now dual core and dual processor at ever increasing clock rates. And on the security front, real-time intrusion prevention and detection benefits from all these advances.
The performance of each of the subsystems on each layer influences how rapidly media traverses the production chain. A trend in equipment design is to move software-based functionality into hardware. No matter how fast your processor, software algorithm execution takes significantly longer than hardware logic processing at wire speed.
Let’s get physical
On the physical layer everything happens in real-time. The opportunity to correct a transmission glitch is nonexistent. IT network technology is not ready for 1.5Gb/s HD-SDI uncompressed real-time video transfers.
SMPTE has been actively seeking methodologies that will support increased SDTI speeds through extension of already defined protocols. A study group is investigating high-speed interfaces. SMPTE 424 specifies the 3Gb/s SDTI implementation. This will facilitate 1080 60p distribution. A 10Gb/s proposal was presented by Sony last year.
10Gb/s speeds are necessary to support the digital cinema rates of 2K and 4K. The protocol can also support five, 1.5Gb/s HD streams. Data is formatted similar to SMPTE 292 and delivered over trunked lines. These are guaranteed throughput rates.
For live production and highlights editing, SMPTE 305 SDTI transfer of content at 540Mb/s and1.5 Gb/s, as implemented in EVS systems, has satisfied the fast turnaround requirements of live sports production. Up to six HD-SDI 1.485Gb/s streams can be transferred in real time. I-frame compression rates are selectable from 100- to 360Mb/s, and 48 uncompressed 24-bit, 48Khz audio tracks are supported. With 3- and 10Gb/s SDTI, a greater number of HD sources could be handled simultaneously in a similar fashion.
These speed increases open the possibility to produce content in 1080 60P or higher resolution formats, can lead to improved efficiency in workflows, facilitate storing content in the highest resolution, and downconverting in real-time as distributed.
Audio: Using IT network
Audio has taken the first steps in moving to commodity IT network switches for distribution and routing. Driven by the expense of traditional broadcast routers, the relatively low data rates, when compared to video, have made this a viable option for AES3 and MADI audio.
AES/EBU digital audio bus (AES3) is a digital audio transmission standard that is based on a synchronous, self-clocking RS-422A compatible physical layer on top of which stereo digital audio and associated subchannel data is transmitted. Multiple sample rates (32kHz to 48kHz) and sample bit depths (up to 24-bits per sample) are supported.
48KHz, 20-bit mono samples have a data rate of 1Mb/s — a piece of cake for even a 10BaseT network. Even when allowing a 50 percent network overhead, five AES-3 audio streams can be supported, with 100Mb/s up to 50 streams!
Multichannel audio digital interface (MADI) is a time division multiplex, unidirectional multichannel digital audio transmission standard (AES10). MADI is based on Fibre Distributed Data Interface (FDDI) transmission technology, but usually uses coaxial cable instead of optical fibre. It can support up to 56 channels and 24-bits per sample distributed on a LAN constructed with commodity IT resources.
Media nets
Moving HD, even when compressed to 100Mb/s, places a great strain on contemporary IT network technology. Routing equipment used in a BOC. Besides the requirement for robust and resilient operation and a deterministic QoS, routing equipment used in a BOC must attain the highest data rates possible. This has commonly been done by trunking lines together. As switches advance to commonly have 10Gb/s ports, trunking allows 40Gb/s and greater backbone to be implemented.
Speed and distance are related to the protocol used for file transfers. TCP has performance limitations (dependant on cable length and the speed of light) because a handshake confirms that a valid packet transfer has been successful. This reduces the data transfer rate. UDP is a best effort transfer. If packets are corrupted they are not resent. Real-time protocol adds a counter field to the UDP header and provides some assurance that packets have been received and can be reassembled into their correct order.
InfiniBand is an emerging technology that solves the distance problem encountered with Fibre Channel-based SANs. Data rates of up to 30Gb/s are supported. Cable lengths of over 100m are possible.
With the proliferation of SANs and NAS, speed is now an important attribute of disk storage systems. Disk spin rates of 15Krm, seek times and emerging holographic and storage switch technologies all strive for maximum performance.
A relatively new design approach for routing and storage devices attain consistent data rates up to theoretical limits by moving the internetworking operating system (IOS) from software to ASICs. Packet header address decoding happens at wire speed and maintains consistent throughput as data demand increases.
Compute platforms
PC manufacturers like to tout their processor clock specs. Make no mistake about it, higher processor clock rates are good, but are no substitute for inferior system design, incompatible chip sets and poorly written code.
With the ubiquitous infiltration of software development kits and development environments writing code for you, it is easy to understand why so many applications have so many similar bugs. Less than optimally reliable code is repeatedly used and propagates unpredictable errors. Yet code compilers are improving execution speed and application reliability. One would expect that sufficiently tested software would be bug-free.
Increased internal chip transfer rates attainable with single-die dual-core processors can eliminate I/O bottlenecks. Multi-thread, larger on chip cache, look-ahead instruction fetching on a reduced 45nm die result in higher execution speeds and increase microprocessor instructions per second. This leads to decreased render times, composite times and eventually will allow full 1080 60p HD field acquisition on a PC. HDV’s day may be a short one.
Guarding the fort
Real-time intrusion detection and prevention systems (IDS/IPS) are indisputable defense requirements. A small hit can have a large impact on the media network and platform. Viruses can infect an on-air system at near instantaneous speeds. The security of a BOC benefits from increases in speed by moving defense technology functionality into hardware.
By moving from software- to hardware-based technologies, network traffic analysis is at wire speed. And new analysis protocols can be downloaded into FPGA’s and other programmable logic arrays for defense against newly discovered threats.
Deep packet inspection, traffic pattern histogram tracking and other forms of analysis are compute intensive and will slow down file transfers. The ability of these systems to act transparently in your infrastructure is dependent on processing speed.
The actual security appliance topology will have a significant impact on network performance. How the core, aggregate and edge switches are connected and IDS/IPS are deployed will impact the effectiveness of your BOC defense. This will also minimize the performance hit to your IT production network and decrease the likelihood of a successful attack.
Related linksInfiniBand Trade Association
InfiniBand Architecture: Transforming I/O Technology
Best of all worlds
It’s not easy to design reliable, maximum speed infrastructure with so many design and equipment variables. Overall system performance is directly dependent on the slowest link. Therefore, increases in speed on each layer of the BOC will affect overall performance. Coupled with an intelligent, technology integrated workflow, production time and assembly time can be reduced.
For 1080 60p production increases in speed throughout the facility is imperative. Keep in mind that Moore’s Law only applies to the number of transistors on a die, doubling every one and a half years. In reality there may be a physical limit to silicon-based performance. At this point, an elegant, efficient engineering design will be required to pump every speed improvement out of a BOC infrastructure.
