Boosting the Efficiency of OTT Delivery With State-of-the-Art Streaming Optimizations 

An increase in streaming capacity has driven OTT service providers to look for ways to reduce the cost of video streaming without decreasing the overall quality of experience (QoE). Through the latest streaming optimizations, video service providers can achieve significant bandwidth savings, while retaining exceptional video quality.

This article will examine the current state of the art in streaming innovations.

The Evolution of Video Codecs 

Moving to a new codec can dramatically reduce the bandwidth for streaming. Using HEVC, video service providers can save up to 40% bandwidth compared with AVC. AV1 reportedly outperforms HEVC (on some content), offering up to 30% bandwidth savings. 

Launching a new codec will not address all devices.

Therefore, a “simulcast” mode needs to be implemented where some devices will use the old codec (i.e., AVC), others will support the newer codec (i.e., HEVC), and the newest devices will leverage the latest codec (i.e., AV1). The savings will only apply to a limited set of devices.

Table 1 shows what savings can be achieved using a new codec, assuming certain distribution of the codecs on all devices. The bitrates are normalized by taking 100 as the AVC bitrate reference. 

Codec	AVC	HEVC	AV1	Total
Bandwidth	100	60	42
Popularity	35%	60%	5%	100%
Traffic	35	36	2.1	73.1
Contribution	48%	49%	3%
New codec savings	27%

It is important to note that the HEVC device popularity evolves over time, location, and device targeted.

• In a recent session at the Streaming Media Connect 2023, Alex Zambelli of Warner Bros. Discovery mentioned that in the U.S. 70% of the targeted devices were HEVC capable.
• Comcast’s Alex Giladi stated that 60% are HEVC devices. 

Compared with a 100% AVC scenario, bringing HEVC and AV1 will only reduce the total traffic by 27%, with AVC still having the 48% of the traffic, HEVC 49%, and AV1 only 3%.

AVC is still creating a bottleneck, generating a need for additional streaming optimizations. 

Content-Adaptive Encoding

Given the codec situation, video service providers may need to use other methods to decrease AVC bandwidth.

A well-known technique called content-adaptive encoding (CAE) can reduce the bitrate without impacting the quality. There is a general agreement that CAE on average, reduces the traffic by 40% for AVC and 20% for HEVC, with no savings for AV1 as no AV1 CAE has been demonstrated so far.

CAE brings the bitrate to a level that is very close to what HEVC achieves.

As a result, some operators prefer to deploy AVC and CAE instead of HEVC. Using CAE with AVC and HEVC reduces the overall bitrate by 48% vs. 100% in the AVC scenario. 

AI Compression

For many years, AI has been used for optimizing video compression. Video service providers can take several paths using AI.

The first approach involves optimizing encoding parameters, similar to what CAE does. The savings is negligeable when CAE is used. 

The second technique optimizes the encoding ladder, saving some bandwidth, but much more on the storage side. Storage will be addressed in a later section.  

The third method is to dynamically change the encoded resolution in the same profile, in order to adapt to the content complexity. Results have shown savings between 20% and 50% depending on the resolution used. 

The last technique is to dynamically change the encoded frame rate, resulting in about 5% to 10% bandwidth savings depending on the motion. Overall, video service providers can safely count on another 40% savings when using AI in addition to other practices.  

Zero Rating Streaming

In some countries like the U.S., an agreement can be made between the mobile network operator (MNO) and the OTT service provider where the content consumed will not be accounted inside the quota and in return it can be throttled by the MNO.

The benefit is to reduce the entire traffic over a mobile network to SD resolution instead of delivering HD or UHD. Then  SD AVC traffic is capped to 1.5 Mbps. Streaming 720p30 video would be in the range of 3 Mbps, meaning zero rating can save up to 50% of the bandwidth. 

Note that this technique will decrease the QoE since viewers will now be watching SD instead of HD, but on mobile devices that level of quality is generally good enough. 

CDN Throttling

CDNs, when running out of capacity, tend to reduce the profiles served to the clients. This situation is difficult to quantify, as no data is publicly available. But it’s akin to moving from an HD to SD profile, similar to zero rating. CDN throttling is more of an exception, while the zero rating is mainstream.

This technique, similar to zero rating, will reduce the QoE, although viewers agree that on a TV, HD content can be watched using a UHD screen with a good enough experience. SD is not watchable on an UHD screen. 

Content Orchestration

Content orchestration is a natural evolution of zero rating or CDN throttling, as it provides a similar outcome, but it is performed automatically.

This type of streaming optimization first looks at the content popularity and then allocates more encoding resources to reduce the bitrate of high popularity content. The goal is to reduce traffic for popular content while preserving the quality and to relax the encoding resources of low popularity content, increasing traffic while preserving the quality. 

Table 2 shows a content orchestration model for a streaming service where 20% of the content generates 80% of the traffic, and the derating factor that optimizes high-popularity content is 20%. For low popularity content it is at +20%.

Content	High Tier	Low Tier	Total
Library	20%	80%
Bitrate	100	100
Popularity	80%	20%
Default traffic	80	20	100
De-rating	80%	120%
Optimized traffic	64	24	88
Savings	12%

The model shows that 12% of the total traffic can be saved using such a content orchestration technique. 

Bandwidth Reduction Without QoE Impact 

The total savings that can be achieved using all of the above-mentioned techniques simultaneously, without impacting the QoE, is described in Table 3. The savings are calculated for a 100% AVC scenario. 

Scenarios	Codecs	CAE	Orchestration	AI
Savings	27%	48%	12%	30%
Codec	27%
CAE + Codec		48%
CAE + Codec + Orch			54%
CAE + Codec+ Orch+AI				68%

The savings that a major streaming service can realize from deploying a new codec plus CAE vs. a full-fledged approach with a new codec, CAE, orchestration, and AI) is 38%. Those gains could be achieved with the VVC codec; however it would require replacing the complete installed base of devices, an effort that could take 10 or more years.

Bandwidth Reduction with QoE Impact 

There are additional ways to optimize traffic. Table 4 shows the device distribution, according to a 2019 Conviva report.

Device	Share of Viewing Hours
Connected TV	35.3%
Connected devices	34.6
Mobile	11.3%
PC	7.1%
Gaming	6.8
Tablet	4.8%
Others	2%

The table is derived based on connected devices including TVs, smartphones, PCs and tablets.

Device	Share of Viewing Hours
Connected TV	77%
Smartphones	11%
PCs	7%
Tablets	5%

Looking at the table, one can calculate the traffic created for each type of device with an associated prediction of the traffic pattern. (See Table 5.)

			Original
Device	Share of Viewing Hours	Resolution	UHD	HD	SD	Traffic/user
		Bitrate (Mbps)	15	4	1.5
		Hour of streaming (GB)	6.6	1.8	0.7
Connected TVs	77%		10%	70%	20%	1.6
Smartphones	11%		0%	50%	50%	0.1
PCs	7%		0%	70%	30%	0.1
Tablets	5%		0%	70%	30%	0.1
Original total traffic/user (GB)						1.9

Next, video service providers can apply an optimization rule: 

• Send only HD to TVs and PCs 
• Send only SD to mobile devices (i.e., smartphones and tablets) 

The traffic results are shown in Table 6, with the red cells highlighting the one now zeroed. The savings for each user is 25%. 

	Modified
	UHD	HD	SD	Traffic/user (GB)
	0%	80%	20%	1.2
	0%	0%	100%	0.1
	0%	70%	30%	0.1
	0%	0%	100%	0.0
Modified total traffic/user (GB)				1.4

Conclusion

According to Sandvine report global internet traffic volume increased 23% in 2022, due in large part to surging streaming video usage. As streaming continues to dominate viewing consumption methods, video service providers need tools to improve bandwidth efficiency and maintain outstanding video quality.  Table 7 summarizes the different savings that can be applied depending on the starting point and the type of optimization used.

Starting Point	No QoE Change	QoE Change	Total Savings
100% AVC	68%	No	68%
	68%	25%	76%
New codecs and CAE	38%	No	38%
	38%	25%	54%

At best, starting from a 100% AVC situation and applying QoE reduction, video service providers can save 76% of the traffic. At worst, using a new codec and  CAE video service providers can save 38% of the traffic without changing the QoE.