Skip to content

Files

Latest commit

urlofmarurlofmar
Restructure docs, add abstract, RPC and more
Jul 5, 2023
28da69a · Jul 5, 2023

History

History
56 lines (45 loc) · 4.25 KB

thinknode_cache.md

File metadata and controls

56 lines (45 loc) · 4.25 KB

Cache

CRADLE caches objects whenever possible; there are two levels:

  • A primary memory cache (default size: 4GB)
  • A secondary cache: a local disk cache (default size: 4GB) or a remote cache

The main local disk-based cache is formed by an SQLite database, which also contains small blobs. Large blobs are compressed via LZ4 and stored in a file.

For all types of secondary storage, structured data is first serialized into a form similar to the MessagePack serialization used between client and CRADLE. (Preferred way forward: no serialization, CRADLE need not understand the data format, it can store raw data.)

What does CRADLE cache?

CRADLE caches most Thinknode responses:

Thinknode request Entry type Link Synopsis
GET /iam/contexts/:id get_context_contents Available apps
GET /apm/apps/:account/:app/versions/:version get_app_version_info App version info
POST /iss/{:type} post_iss_object post_iss_object Reference id to an immutable object
GET /iss/:id/immutable resolve_iss_object_to_immutable Data Immutable id
GET /iss/immutable/:id retrieve_immutable Data Immutable data as dynamic
GET /iss/immutable/:id retrieve_immutable_blob Data Immutable data as blob
HEAD /iss/:id get_iss_object_metadata iss_object_metadata Metadata for an immutable object
POST /calc/:id post_calculation post_calculation Calculation request id
GET /calc/:id retrieve_calculation_request calculation_request Calculation descriptor
GET /calc/:id/status (not cached) Retrieves calculation status
POST /iss/:id/buckets/:bucket N/A (see notes)

Notes:

  • The entry type identifies what kind of data the cache entry contains. The cache keys are a hash calculated over several items, the first one always being the entry type string.
  • If the link refers to a CRADLE request, the cache entry contains the data returned in the response to that request.
  • "N/A" means that the Thinknode request is issued only for CRADLE requests that are candidate for removal according to the overview (see below).

CRADLE also caches:

Entry type Link Synopsis
local_function_calc perform_local_calc Return value from local function calls
resolve_named_type_reference Part of schema returned via get_app_version_info (is it useful to store this?)
coerce_encoded_object Conversion of an object passed in a post_iss_object request

CRADLE currently also caches information on behalf of messages that are candidate for removal according to the overview:

Entry type Status
resolve_results_api_query Astroid
locally_resolve_results_api_query Astroid
type_contains_references Thinknode
deeply_copy_iss_object Thinknode
deeply_copy_calculation Thinknode