Performance and Caching v6.0.2

Where file-based configuration is used, performance and caching parameters are configurable through the cache section of the file.

Each Quantum Origin entropy generation cycle generates a fixed block-size of entropy depending on use-case and circumstance.

In some situations, many small entropy requests - far smaller than a single block size - are required. To avoid wasted cycles generating unused entropy, Quantum Origin can cache entropy between each request. This is called synchronous caching. Any excess entropy from a previous generation call will be used in the next call.

In other situations, immediate responsiveness is critical. For this, background/multithreaded caching can be used, this starts a new thread or threads that can independently generate entropy outside of direct requests. This generates a cache pool that can be immediately accessed on demand without further generation.

Tip

Where high performance, multi-core generation is required, use multithread-caching.

The standard cache mechanisms can be significantly extended for almost any use case.

Cache modes

Option	Description
NONE	Do not cache between each request.
SYNCHRONOUS-CACHING	Cache excess entropy between requests, no new thread is used.
CACHING	Start a new thread to generate cached entropy in the background. Cache excess entropy. A new thread is used.
MULTITHREAD-CACHING	Start multiple new threads to generate cached entropy in the background. Cache excess entropy.

If caching is enabled, the size of the cache must be provided.

A particularly useful cache size for synchronous caching is 3973 bytes, the size of a large block - this ensures no entropy is discarded.

Cache Option

Option	Description
size	Total capacity of the cache in bytes.
prefill	The amount of bytes to generate at startup prior to any randomness call being accepted. The cache will be filled to this level synchronously on initialisation even when using a background/multithreaded cache.
refill_at	The level below which refilling of the cache will be automatically triggered. Only used for synchronous caching. It does not apply to background/multithreaded cache which will always attempt to keep the cache filled to capacity.
threads	The number of threads to use in multithreaded extraction. It is recommended this is kept fairly low.

The following are examples of cache configs, using no caching and multithreaded (background) caching respectively.

cache:
   type: none

cache:
  type: multithread-caching
  size: 4096
  prefill: 128
  refill_at: 1024