Skip to main content

Minimize transmitted data size

Description​

From an energy-efficiency perspective, it's better to minimize the size of the data transmitted so that less energy is required because the network traffic is reduced.

Solution​

Minimize network traffic by applying both data reduction and compression techniques. These complementary approaches work together to achieve maximum efficiency:

Approach 1: Reduce Data at the Application Layer​

Curate the data being transmitted by only sending properties or values deemed necessary:

  • Send only the specific fields required by the recipient
  • Filter out unnecessary metadata or properties that can be correlated later
  • Choose efficient data formats that are more compact (e.g., Protobuf is much more compact than XML, JSON is more compact than XML)
  • Use pagination or streaming for large datasets rather than sending everything at once
  • Implement field selection APIs that allow clients to request only needed properties

Approach 2: Compress Data at the Transport Layer​

Apply compression mechanisms to files or payloads before transmission:

  • Use gzip, Brotli, or zstd compression for text-based data (JSON, HTML, CSS, JavaScript)
  • Enable HTTP compression at the server and CDN level
  • Compress binary files and payloads where appropriate
  • Consider the compression-decompression energy tradeoff (modern compression algorithms are efficient enough that network savings outweigh computational cost)

Best Practice: Apply both approaches together. First reduce the data set to only what's necessary, then apply compression to the payload being transmitted. This maximizes bandwidth savings and energy efficiency.

SCI Impact​

SCI = (E * I) + M per R Software Carbon Intensity Spec

Minimizing transmitted data size will impact SCI as follows:

  • E: Reduces the total electricity required by reducing network traffic. While compression/decompression adds slight computational cost, the network energy savings typically outweigh this overhead.
  • I: May have a slight increase in location-based marginal carbon emissions due to compressing and decompressing data at different locations, but this is generally offset by reduced transmission energy.
  • M: By reducing the total size of data stored and transmitted, the total embodied carbon of network infrastructure is lower.

Assumptions​

  • Rather than transferring complete data payloads because some properties might be needed later, a better approach is to curate the data set to send only necessary properties, with the ability to request additional data if needed
  • Different data formats significantly affect network traffic - choose compact formats when possible
  • Compression mechanisms are available and compatible on both client and server sides

Considerations​

  • May reduce cloud bills because minimizing the size of the data transmitted costs less
  • If data sets are minimized by removing properties, there may be overhead to correlate missing properties or request them separately
  • If compressed assets cannot be properly decompressed, there is overhead to resend in alternative format
  • Balance between data completeness and transmission efficiency should align with application requirements

References​