The networking ballgame has certainly changed a lot since the first decade of the 2000’s, but fortunately the basics I learned all those years ago remain the same. As a recovering IT guy, and now broadcast engineer, I have found my past life as a LAN Manager/System Administrator comes in handy during our current transition from SDI (Serial Device Interface) to new audio and video over IP technologies such as DanteAV and NDI (Network Device Interface).
I guess you could say that I never really left the IT field when transitioning to broadcast engineering. I basically made my move along with the broadcasting industry. In fact, that is really what launched my broadcasting career. The station where I worked needed someone familiar with IT to help out in the engineering department to implement all the new computer and network-based equipment. Since I was already on staff, the move over to the engineering side of the house was a smooth transition. Now a veteran of both fields, I get the chance to share my knowledge with young minds eager for new adventures in the crazy ever-evolving world of media and communication.
It doesn’t take long to see that broadcast television today is making a full transition to streamed delivery, both over the airwaves using ATSC 3 or via live streaming services and on demand streaming like Netflix and Amazon Prime. The world of broadcasting is ever-changing, but the transitions today are flying in at exponential rates when compared to the last few decades. Network advancements are the impotence for these vast changes.
In this article, we will begin a journey to understanding broadcast quality streaming over IP based workflows. To get there, we begin with editing over a network. The purpose of this guide is to provide a broad, but hopefully nuanced enough overview of the networking factors that play a crucial role in the delivery of audio and video over IP and the ability to edit and work on networked AV files. I hope this will serve to provide a stronger understanding of these principles and allow for more intelligent conversations with the IT staff that I know will prove to be integral in any network-based workflow setup within your facility.
I will start with transfer rates of common serial bus connections. Here is a quick look at some of the more recent (and some of these have been around for years, so I use the term recent loosely) common serial bus connections- i.e., USB and Thunderbolt. We will dig deeper into this as we move along.
- USB-3.2: USB-3.2 is the latest version of the USB-3 interface and supports three different transfer rates:
· USB 3.2 Gen 1x1 (formerly known as USB 3.0): up to 5 Gbps
· USB 3.2 Gen 2x1: up to 10 Gbps
· USB 3.2 Gen 2x2: up to 20 Gbps
- USB 4: USB 4 supports a maximum data transfer rate of 40 Gbps, which two times faster than USB-3.2 Gen 2x2 and similar to Thunderbolt 3 and 4.
- Thunderbolt 3: Thunderbolt 3 supports a maximum data transfer rate of 40 Gbps, which is significantly faster than USB-3.2. This makes it suitable for transferring large files quickly and running high-bandwidth applications, such as video editing.
- Thunderbolt 4: Thunderbolt 4 supports the same maximum data transfer rate of 40 Gbps as Thunderbolt 3. However, Thunderbolt 4 has stricter hardware requirements than Thunderbolt 3, such as support for two 4K displays or one 8K display, and ensures a minimum performance level for connected devices. Additionally, it improves compatibility with USB-C and includes enhanced security features.
So, what do all these numbers represent, and why should you care? I think a good approach here might be to discuss data transfer rates and speeds in terms that most of us have dealt with over the years. Since the year 2000 the broadcast industry has moved from editing analog video using tape-to-tape editors, to SD digital video on early NLE’s, to 720p, then 1080i, followed by 1080p, and now 2K and 4K. If you have been involved in the technical side of this business for very long, I am sure you have answered the call for support of someone attempting to edit video straight from a USB thumb drive or an SD card. The throughput of those older external drives just couldn’t keep up, causing stuttering and laggy video playback or system errors. Editing video over an early USB or SD technology might have worked OK for low-res video, but you were sure to experience trouble with anything over 720p (of course the codec plays a big part in all of this as well… that’s another discussion for another day). The reason for this trouble is... you guessed it, data transfer rates.
Up until a few years ago, most thumbdrives used USB 1 or 2 connectivity meaning that the transfer rate was maybe 12 Mbps in the case of USB 1 and up to 480 Mbps for USB 2. Likewise, a Class 10 SD card had a transfer rate of 80 Mbps with newer UHS Class 3 cards transferring data at around 240 Mbps. There are faster alternatives out there, but for the sake of explanation we will go with these examples.
Let’s look at some general numbers of transfer rate requirements for editing different video formats:
- 480i (Standard Definition): A transfer rate of at least 5 MB/s (40 Mbps) is recommended for smooth editing.
- 720p (High Definition): A transfer rate of at least 10 MB/s (80 Mbps) is recommended for smooth editing.
- 1080i (High Definition Interlaced): A transfer rate of at least 20 MB/s (160 Mbps) is recommended for smooth editing.
- 1080p (Full High Definition): A transfer rate of at least 30 MB/s (240 Mbps) is recommended for smooth editing.
- 2K (2048 x 1080): A transfer rate of at least 50 MB/s (400 Mbps) is recommended for smooth editing.
- 4K (4096 x 2160 or 3840 x 2160): A transfer rate of at least 100 MB/s (800 Mbps) or higher is recommended for smooth editing, depending on the codec and bit rate of the footage.
**A quick disclaimer, the above numbers are rough estimates, the actual data transfer rate required can vary depending on factors such as codec, bit depth, frame rate, etc. These numbers represent video using the codec Apple ProRes 422 and are much lower than what is needed for uncompressed RAW video files and other less compressed codecs.
So, on the surface it seems that we should be able to edit video up to 2K that is stored on a USB 2.0 connected hard drive. Not exactly… you see each 2K video stream requires up to 400 Mbps depending on the codec. When editing video, you may have three or four tracks of video plus audio in a project. Each of those tracks use bandwidth. It wouldn’t take long to saturate that 480 Mbps pipe to your external hard drive. Also, keep in mind that your editing software is talking to your computer's CPU, RAM, GPU, and possibly other services and applications, all of which may be cutting back on your available bandwidth to your external drive or card.
Next, let’s move a few years forward to more current technology. USB 3.0 was first introduced to the market in 2010 and has, in most cases, replaced the older USB 2 technology. The initial version of USB 3.0 supported transfer rates of 5 Gbps followed by USB 3.1 and now 3.2 which support up to 10 Gbps transfer rates, with USB 3.2 Gen 2x2 supporting up to 20 Gbps transfer rates.
Before I move on let me layout the current USB nomenclature and rebranding, this may clear up some confusion from the last few paragraphs:
- USB 3.0: 5 Gbps (gigabits per second)
- USB 3.1 Gen 1: 5 Gbps (gigabits per second) - this is essentially a rebranded version of USB 3.0, with no increase in speed
- USB 3.1 Gen 2: 10 Gbps (gigabits per second)
- USB 3.2 Gen 1x1: 5 Gbps (gigabits per second) - this is also a rebranded version of USB 3.0/3.1 Gen 1, with no increase in speed
- USB 3.2 Gen 2x1: 10 Gbps (gigabits per second) - this is the same speed as USB 3.1 Gen 2
- USB 3.2 Gen 2x2: 20 Gbps (gigabits per second) - this is the fastest USB transfer rate currently available.
So, with USB 3.2 we are looking pretty good at being able to edit most digital video files and formats even up to 4K on a supported external hard drive. Cool? Again, not quite. There aren’t many USB 3.2 Gen 2x2 (20 Gbps) devices currently on the market, most are still using the older 3.2 or 3.1, @ 10Gbps, and many USB 3 ports on computers today are not ready for 3.2 Gen 2x2. We are back to only handling compressed files up to 2K smoothly over an externally connected USB 3 hard drive.
The current answer to this dilemma is to edit using proxies. Most editing software manufactures make this an easy-to-use solution built directly into their latest versions. With proxy editing the necessary bandwidth to edit 4K RAW video is reduced to around 500 Mbps to 1 Gbps per video layer depending on the codec and other factors. Any USB 3 iteration should handle proxy editing fairly well as long as we are talking about basic video editing only… none of that fancy 3D stuff.
As great as proxy editing is, it does have its drawbacks. Your video will look grainy until you render or pause playback which can cause some issues when fine tuning a project and will certainly slow down production. Likewise, many editors hate having to bounce back and forth between full res and proxy, they much prefer to edit cleanly and smoothly without jumping through any hoops. Editors dealing with 4K RAW footage must make a choice, do I live with proxy editing or not?
Let’s go with “or not”. There are options. Thunderbolt 3 and 4, which I mentioned off the top, both support transfer rates up to 40 Gbps. With a Thunderbolt 3 or 4 external drive and a compatible Thunderbolt 3 or 4 connection on your computer, you should be in business to edit 4K RAW uncompressed video directly on the drive. Problem solved, right? Yes, but that can be expensive. For the sole freelancer out there with plenty of money to purchase a hard drive with plenty of capacity, a suitable workflow has been found. Remember this is editing over an external drive. If your computer is equipped with a PCIe-based internal SSD you may get up to 56 Gbps transfer rate and have even better success editing on your internal drive. The elephant in the room for all this 4K RAW editing talk is hard drive size. One minute of uncompressed 4K RAW video takes up around 4 GB of storage space.
Hopefully, all the talk about transfer rates as it relates to external and internal hardware has given a good basis to make our next leap into the world of “video editing over a network” using a NAS (Network Attached Storage). Everything we have discussed so far is all well and good for the creative freelancer out their working gig to gig, but what about the education sector, corporate video, and production houses around the world? How do we deal with the folks that need to share content, collaborate with each other, and work on projects together at the same time? This just so happens to be the world that I live in each day, and I am sure many reading this are also dealing with or have dealt with this burning question. Thankfully, NAS server salesmen are very well versed on this topic and have a way of explaining how it all works, without actually explaining how any of it works. This is where I hope what I am writing might be helpful to my fellow engineers and technicians.
If you thought a high-end editing PC or Mac with a 4 TB external Thunderbolt 3 hard drive was expensive, wait until you see the price tag of something like an Enterprise level Avid Nexis E4 system. We are talking about starting prices in the $100K range. Why is it so expensive! Let me explain.
I’m certainly not a salesman for Avid Nexis, I have had my share of complaints, but when you consider what this product will do and how many editors it will serve at once, you begin to gain a better appreciation for the initial price tag and the yearly support contract expense.
NAS video servers are all very similar in how they work. In general, you have a system director device to handle media and user management. The system director is connected to media drives to house all your AV files. The drives are typically PCIe-based SSDs to handle the needed transfer rates for editing video over a network. Typically, you will have a couple of Fiber ports that can be combined to provide ample uplink bandwidth between your server and your switch. The scalability and expansion level of all of this will typically coincide with the price you pay for the server- be ready to write a big check!
Let’s start with the basics. A discussion of networking should begin with a brief look at network cabling and the bandwidth each standard offers.
Here is a list of the infrastructure cabling options we will be discussing in this section:
- Cat 5e - This type of cable can support up to 1 Gigabit per second (Gbps) of data transfer rate.
- Cat 6 - This type of cable can support up to 10 Gbps of data transfer rate over a distance of up to 55 meters.
- Cat 6a - This type of cable can support up to 10 Gbps of data transfer rate over a distance of up to 100 meters.
- Fiber optic - Fiber optic cables have the potential to transfer data at incredibly high speeds, ranging from 10 Gbps to 100 Gbps or more, depending on the type of fiber, the transceiver, and the distance of the cable run.
Most modern switches provide ethernet connectivity of 1 Gbps up to 10 Gbps. Most computers ship with either a 1 Gbps or 10 Gbps ethernet adapter. From our previous examples of USB and Thunderbolt standards, we know that compressed 1080p video requires a 240 Mbps transfer rate to edit smoothly over a medium. If you are attempting to edit multi-layers of video this number will need to be multiplied by the number of video layers.
240 Mbps x 2 video layers = 480 Mbps (necessary bandwidth 2 layers of 1080p video)
240 Mbps x 4 video layers= 960 Mbps (necessary bandwidth for 4 layers of 1080p video)
First, we will think about the above network bandwidth requirements using a 1 Gbps managed switch. In this case the medium is Cat 5e (or better) connected to a 1 Gbps switch port, connected to a video NAS with at least a 10 Gbps uplink to the switch. Someone editing a four 1080p video layered project will experience lag, stuttering, and likely will get read/write errors. The bottleneck is with the 1 Gbps network connection to the switch. The world isn’t perfect, and neither are network switches. Generally, you should allow at least a 10% overhead on your available network bandwidth. 960 is greater than 900, so, Houston, we have a problem.
The 10 Gbps uplink is fully capable of handling the above scenario, the NAS server has the drive speed to handle the necessary data transfer rates, but the 1 Gbps NIC is just not going to cut it. If we drop back to a lower resolution like 1080i or 720p, or we drop a video layer, we should be able to edit smoothly as our max bandwidth usage will decrease. Usually, three video layers and a couple of audio layers is all we will need. Therefore, most of the time, we can expect smooth editing up to 1080p over a 1 Gbps connection to a 10 Gbps connected NAS server.
Unlike the old network hub days where the port bandwidth was spread out amongst all of the clients connected to it, we now have switch technology. The beauty of a switch is that every 1 Gbps port has most of its full bandwidth available to transfer data in and out of the switch. The bigger pipe for uplinks between switches or fiber connected devices provides the switch with its ability to move many gigabits of data per second smoothly around your network. Higher end switches will support up to 40 Gbps fiber uplinks and the best of the best will support a bunch of those 40 Gbps uplinks that can be bonded to make an even bigger pipe for all that switch data to travel through. So, long story short, with modern switches, usually, uplink will not be the source of a bottleneck if you are editing 1080p video or below.
Moving on, let’s swap out that 1 Gbps switch with a 10 Gbps (or 10,000 Mbps) switch better known as 10GE (10 Gig over Ethernet). We also need to swap the cabling from Cat5e to at least Cat 6. Since we are moving to an enterprise level 10GE switch, we will likely be capable of establishing multiple 40 Gbps fiber uplinks to other network devices. Again, with the uplink to our NAS we shouldn’t have a problem. That 10 Gbps bandwidth on each port will allow us to edit compressed 4K video, proxy edit 4K RAW, and edit many layers of lower resolution video smoothly.
Here is a quick look at the numbers:
800 Mbps x 10 = 8,000 Mbps = 800 Gbps (bandwidth needed to edit 10 layers of compressed 4K video)
Even the most creative creatives out there will probably only rarely need 10 layers of compressed 4K video on a timeline at once, but it could happen. One more layer and we have saturated the bandwidth to the switch. Let’s add another 10 GE connected client editing multiple layers of compressed 4K video. Uh oh, Houston, we have another problem. Our 10Gbps uplink is now overloaded. Even though the switch can handle the data transfer needs, the fiber uplink has now met its match. With the 1080p workflow, we could have many clients working at once and not max out the NAS server’s uplink to the switch, but once we take the next leap into 4K editing (remember, we haven’t even considered RAW yet), now we must worry about our uplink speeds. Again, most 10 GE switches are going to be capable of multiple 40 Gbps uplinks. If you future proofed your NAS, you may be able to buy a couple of 40 Gbps transceivers and give your workflow some breathing room. Keep in mind, none of this is cheap!
We have finally made our way on this journey to the biggie, 4K RAW. Popular 4K RAW codecs like REDCODE or CinemaDNG can require transfer rates of up to several GB/s to edit smoothly without proxies.
Here is the math to convert Gigabytes per second to Gigabits per second…
1 GB = 8 Gb
so,
3 GB/s x 8 = 24 Gbps
This means that each video layer of 4K RAW footage could require upwards of 24 Gbps. I hate to tell you, but even your most robust 10 GE switch isn’t going to handle 4K RAW editing without proxies. Man, what a buzz kill! I thought we were really getting somewhere.
Broadcast engineers and technicians, proxy editing is currently our only line of defense to the 4K RAW thirsty creative connected to a 10GE switch via copper. Over time the technology will improve, and workflows will change... but for now, proxy editing is all we have unless we jump to a fiber switch.
This brings us back to where I began. What are the network requirements and switch settings necessary for audio and video streaming using Full NDI and Dante’?
A couple of years ago, I assisted on an NDI/Dante’ project at East Tennessee State University. Through that process I garnered much useful information that I hope to share in Part II of this series. I am currently in the process of load testing a Cisco Nexus 9372TX switch. I hope to identify any bottlenecks or limitations in the network infrastructure between a NAS and edit station to edit 4K and 6K video using proxies and various codecs. Part of this process is to find ways to optimize performance and ensure smooth operation of the network and also select an appropriate replacement switch for the 10Gbps Nexus which is nearing its EOS (end of support). Once I have completed this testing I will update this article and hopefully have a chance to write Part II which will specifically deal with the implementation of NDI and Dante’ on an existing Gigabit network. For those of you who made it with me to the end of this article, I hope this has provided a strong basis to begin the all-important conversation of transitioning from SDI to AV over IP in a broadcast facility.
0 Comments