In the rapidly evolving landscape of cloud services, AWS has consistently set the standard with its comprehensive suite of offerings. However, even the most robust ecosystems have their limitations. Today, I'm diving into a significant gap in AWS's video processing capabilities: the lack of tiered storage options for Kinesis Video Streams (KVS).

As organizations across various sectors increasingly rely on video data for critical operations, the absence of storage class diversity in KVS has become more than just an inconvenience — it's a substantial financial burden that limits adoption and scalability for many use cases.

Understanding Kinesis Video Streams

AWS Kinesis Video Streams is a powerful service designed to stream video from connected devices to AWS for analytics, machine learning, and other processing applications. It efficiently captures, processes, and stores video data at scale, making it an attractive solution for numerous applications:

• Real-time monitoring of manufacturing equipment
• Smart city infrastructure surveillance
• Telehealth and remote patient monitoring
• Security camera networks for facilities
• Connected vehicle systems and autonomous driving

The service's primary strength lies in its seamless integration with other AWS services like SageMaker for machine learning, Lambda for serverless computing, and Rekognition for video analysis. This integration enables sophisticated workflows and real-time insights from video data.

The Single Storage Class Problem

Despite its advanced capabilities, KVS has a fundamental limitation: it offers only standard storage with no tiered options. Unlike Amazon S3, which provides multiple storage classes (Standard, Intelligent-Tiering, Standard-IA, One Zone-IA, Glacier, and Glacier Deep Archive), KVS users are locked into a one-size-fits-all approach.

This limitation becomes particularly problematic when we consider the typical lifecycle of video data:

1. Short-term (0–30 days): Highest access frequency, often requiring real-time or near-real-time processing 2. Medium-term (30–90 days): Occasional access, primarily for reference or specific analysis tasks 3. Long-term (90+ days): Rare access, mainly for compliance, archival, or potential future analysis

In the current KVS model, you pay the same premium rate whether the video was captured minutes ago or months ago, regardless of access patterns or business value.

The Cost Implications

Let's break down the numbers to understand the financial impact. At its current pricing (as of April 2025), KVS storage costs approximately $0.023 per GB per month, with additional charges for data retrieval and processing.

For perspective, consider an automotive company managing a fleet of 1,000 vehicles, each generating an average of 2GB of video data daily:

• Daily data generation: 2,000 GB
• Monthly data generation: 60,000 GB
• Annual data generation: 720,000 GB

At standard KVS rates, storing just one year of this data would cost approximately $16,560 per month or nearly $200,000 annually — solely for storage, before factoring in ingestion, retrieval, or processing costs.

If AWS offered tiered storage similar to S3, where Glacier Deep Archive costs roughly $0.00099 per GB per month, the same company could potentially store older, less frequently accessed video data at less than 5% of the current cost.

Industry-Specific Challenges

Automotive and Autonomous Vehicles

The automotive industry faces particularly acute challenges with KVS's storage limitations. Advanced Driver-Assistance Systems (ADAS) and autonomous vehicle development require massive video datasets for:

• Training machine learning models
• Regulatory compliance and safety verification
• Incident analysis and continuous improvement
• Insurance claims processing
• Fleet management and driver behavior analysis

Tesla, for instance, has reported collecting over 3 billion miles of driving data, with video comprising a significant portion. Without tiered storage options, maintaining this data in KVS becomes prohibitively expensive, forcing companies to either delete valuable historical data or implement complex custom archiving solutions.

Smart Cities and Urban Infrastructure

Cities implementing intelligent transportation systems, traffic monitoring, and public safety networks face similar challenges. These systems typically involve hundreds or thousands of cameras operating continuously, generating petabytes of data annually.

While recent footage has high operational value for immediate incident response, older footage often serves primarily archival purposes for statistical analysis or rare investigations. The inability to transition this older data to lower-cost storage tiers significantly impacts municipal budgets already stretched thin.

Healthcare and Telemedicine

The healthcare industry has seen explosive growth in video utilization, particularly since the pandemic accelerated telehealth adoption. Medical providers must retain video consultations for:

• Patient record completeness
• Medical training and education
• Quality assurance reviews
• Regulatory compliance (sometimes requiring retention for 7+ years)

With KVS's current pricing model, storing years of telehealth sessions becomes financially unsustainable for many healthcare organizations, potentially forcing them toward less integrated but more cost-effective storage solutions.

Industrial IoT and Manufacturing

Manufacturing facilities employing computer vision for quality control, process optimization, and predictive maintenance generate continuous video streams from multiple production lines. This data has a natural lifecycle:

• Current data (hours/days old): Critical for immediate process control
• Recent data (weeks old): Valuable for short-term trend analysis
• Historical data (months/years old): Essential for long-term improvement initiatives and compliance

The absence of storage tiers in KVS creates an artificial choice between comprehensive data retention and budget management.

Why Hasn't AWS Addressed This Gap?

Given AWS's typical responsiveness to customer needs, the persistent absence of tiered storage for KVS raises questions. Several factors may explain this limitation:

Technical Challenges

Unlike object storage (S3), video streams have unique characteristics that complicate tiering:

• Sequential access patterns
• Complex indexing requirements for random access
• Metadata and timestamp dependencies
• Variable bitrate considerations

Implementing effective cold storage for time-series video data while maintaining reasonable retrieval performance presents genuine engineering challenges.

Product Positioning

AWS may intentionally position KVS as a service optimized for real-time and recent video processing rather than long-term archival. This would align with common video workflow patterns where only a small percentage of footage requires permanent retention.

The expectation may be that customers will implement their own lifecycle policies, moving aged data to S3 or other storage solutions — a reasonable approach for sophisticated users but a significant barrier for smaller organizations.

Market Segmentation

By maintaining distinct capabilities between services, AWS creates natural integration points between products. The current limitation may encourage users to develop architectures that combine KVS's streaming strengths with S3's storage efficiency, potentially increasing overall AWS service adoption.

Current Workarounds and Best Practices

While awaiting potential improvements to KVS storage options, organizations can implement several strategies to manage costs:

Automated Lifecycle Management

Develop custom solutions to automatically migrate aging video data from KVS to tiered S3 storage. This typically involves:

1. Setting up Lambda functions triggered on a schedule 2. Using the KVS API to retrieve specific fragments or streams 3. Storing the retrieved data in appropriate S3 storage classes 4. Implementing a metadata database to track video locations across storage systems

This approach requires significant development effort but can dramatically reduce long-term storage costs.

Selective Retention Policies

Rather than storing all video indefinitely, implement intelligent retention policies based on:

• Video content analysis (keeping only frames/segments with relevant activity)
• Business value scoring (retaining high-value footage longer)
• Sampling techniques (storing representative footage rather than exhaustive records)
• Compression or resolution reduction for archived footage

Edge Processing and Filtering

Deploy edge computing solutions to process video at the source, sending only relevant segments or extracted insights to the cloud. This approach can reduce storage requirements by 60–90% for many use cases.

Hybrid Cloud Architectures

For organizations with substantial on-premises infrastructure, consider hybrid architectures where KVS handles real-time processing while on-premises solutions manage long-term storage of historical video data.

The Path Forward: What AWS Should Consider

To address this significant gap in their video processing ecosystem, AWS could implement several improvements:

Integrated Lifecycle Policies

Similar to S3, KVS could offer built-in lifecycle policies allowing automatic transitions between storage tiers based on age or access patterns. This would eliminate the need for custom migration solutions.

Storage Classes for KVS

Introducing even basic tiering options would significantly improve the service:

• KVS Standard: Current offering, optimized for frequent access
• KVS Infrequent Access: Lower storage cost with slightly higher retrieval fees
• KVS Archive: Substantially reduced storage costs with longer retrieval times

S3 Integration for Archival

Native integration between KVS and S3 would streamline the process of archiving older video data while maintaining metadata relationships and simplified retrieval.

Intelligent Data Reduction

Automated compression, resolution reduction, or frame rate adjustments for aging video could provide cost savings while preserving essential information.

Conclusion

The Business Case for Change

The absence of tiered storage in Kinesis Video Streams represents more than just a technical limitation — it's a strategic opportunity for AWS to address a growing market need.

As industries from automotive to healthcare increasingly leverage video data for competitive advantage, the ability to cost-effectively store and manage this data throughout its lifecycle becomes critical. Organizations currently face a difficult choice between the seamless integration of KVS and the cost-efficiency of alternative solutions.

By addressing this gap, AWS could:

1. Increase adoption of KVS across cost-sensitive industries
2. Enable longer retention of valuable historical video data
3. Provide a more complete end-to-end solution for video-intensive workloads
4. Differentiate from competing offerings from other cloud providers

Until AWS implements such improvements, organizations must carefully weigh the convenience of KVS against its long-term storage costs, potentially implementing complex workarounds or accepting the premium pricing as the cost of integration.

For data-intensive applications like autonomous vehicles, smart cities, and healthcare, this limitation may ultimately determine whether KVS becomes the foundation of their video processing architecture or merely a component in a more diverse solution landscape.

What are your experiences with managing video data at scale? Have you found effective strategies for balancing the benefits of Kinesis Video Streams with its storage limitations? Share your thoughts and approaches in the comments below.