ROCKITPLAY API
  1. ROCKIT Edge - Backend API
ROCKITPLAY API
  • ROCKIT Edge - Admin API
    • Minimal Tests
      • ping
      • submit test task
    • Admin Edge Organizations
      • Create Organization
      • Delete Organization
  • ROCKIT Edge - Backend API
    • Backend Edge Organizations
      • Create Organization
      • Delete Organization
    • Authentication
      • user login
      • authentication
    • Account Management
      • get user details
      • list users
      • Modify user
      • delete users
      • Create User / Create Machine User
    • Deployments
      • Creating Deployments
      • Get Deployments
      • Get Deployment Details
      • Modify Deployment
      • Deleting Deployments
    • Apps
      • Creating Apps
      • Import native builds
      • App Details
      • List Apps
      • patch app
      • Delete Apps
    • Triggers
      • Process Traces
      • Retry Task
    • Tasks
      • Listing Tasks
      • Task Details
      • Acknowledge Warning / Expire Task
    • Subscriptions
      • Subscribe
      • List Subscriptions
      • Unsubscribe
    • Roles
      • List Roles
      • Create Role
      • Patch Role
      • Role Details
      • Delete Roles
    • Training
      • Training Details
      • Training Infos
      • Patch Training
  • Changelog
    • Changelog
  1. ROCKIT Edge - Backend API

Training

Game Training#

After uploading the native game build to the ROCKITPLAY, the internal service ROCKIT Engine generates ROCKIT images and ROCKIT patches which will be automatically deployed to the CDN origin. A game launcher which is linked against the ROCKIT RTE downloads the ROCKIT image / patches from the CDN. After the game play session ROCKIT Trace data (block access pattern) is uploaded to the ROCKITPLAY Cloud Service in order to enable the FastStart experience. Typically, only a few iterations are necessary to obtain a ROCKIT FastStart image. The process is illustrated below:
ROCKITPLAY-Roundtrip
This process is referred to as training and will be discussed in the following paragraphs
The ROCKITPLAY Cloud Service constantly monitors uploaded ROCKIT trace files. Once a specified number of traces have been received it automatically triggers the ROCKIT trace file processing processes. If the new trace files lead to a significant quality improvement of the computed sequence, a new ROCKIT image will be generated and deployed automatically.
The training process is applicable to various scenarios such as:
QA – Use traces obtained during the QA process prior to publishing
Live – Dynamically improve progressive download through user activity (crowd sourcing).
Hybrid – A combination of pre-release and post-release training.
When uploading subsequent game builds the current training state is being applied automatically. The system represents this training state as a single trace. However, after uploading new game builds additional traces may need to be collected and processed to achieve an optimal FastStart result.

Training Parameter#

There are multiple parameters that control the trace processing and the image generation. These parameters determine
the quality requirements for individual ROCKIT trace files to be accepted by the ROCKITPLAY Cloud Service and
how often ROCKIT images may be updated in order to avoid too frequent compute cycles and ROCKIT image updates.

Quality Requirements#

Trace length and coverage. Valuable information is mainly obtained from longer game sessions which access many files (file access coverage) during gameplay. Furthermore, unique first-time game sessions, i.e., game sessions from the beginning of the game, are of particular interest for sequencing the ROCKIT image. Only those data generated from long-lasting first-time game sessions should be considered when generating new ROCKIT images.
Unique traces. In order to obtain long first-time game file access information, ROCKIT trace files are concatenated as the gamer progresses through the game during multiple sessions. Thus, a gamer generates several ROCKIT trace files in multiple game sessions that are merged into a single ROCKIT trace. In the discussion below these concatenated ROCKIT traces are referred to as unique ROCKIT traces.

Compute Cycle Management#

Besides the above mentioned quality criteria ROCKITPLAY allows controlling the compute cycle in order to reduce compute costs and to avoid unnecessary ROCKIT image computations and update cycles. This section discusses the corresponding parameters.
Received Traces. ROCKITPLAY will not trigger the ROCKIT trace processing before the specified number of ROCKIT traces have been received.
Reliable traces. In order to avoid unnecessary ROCKIT image computation and update cycles, the ROCKIT image process depends on the minimum of reliable ROCKIT traces that have been received. Reliable traces are those unique long-lasting first-time ROCKIT traces with a minimum file access coverage. Only if a given number of reliable traces are available the ROCKIT image processing is being triggered.
Additional information about the above described parameters can be found in the documentation of the endpoint PATCH /be/v1/trainings.
Modified at 2025-10-10 13:19:15
Previous
Delete Roles
Next
Training Details
Built with