Nguyễn Việt Đức created JAMES-3135:
--------------------------------------
Summary: Write a CassandraDumbBlobStoreCache
Key: JAMES-3135
URL: https://issues.apache.org/jira/browse/JAMES-3135
Project: James Server
Issue Type: Improvement
Reporter: Nguyễn Việt Đức
Location: {{blob-store-cassandra}} maven project.
{{interface BlobStoreCache { Publisher<Void> cache(BlobId blobId, byte[]
data); Publisher<Void> cache(BlobId blobId, InputStream data);
Publisher<byte[]> read(BlobId); Publisher<byte[]> delete(BlobId);}}}
*DOD* Write a {{Cassandra}} implementation of it.
* The cache is responsible of caching only small input, below a configurable
threshold. Calling {{cache}} on a too big input should result in nothing being
stored. (default 8kb)
* Failure handling:
** Circuit breaking: if retrieving the content takes more than
{{blobStore.cache.timeout}} (default to 50ms) then return empty which will
result in reading the main data store
** Ignore errors: return empty.
GC grace period for this table should be set to 0
ConsistencyLevel ONE
API + contract test + implementation
----
Why a cache:
* We don't want to read the blobStore for small input as it comes with high
latency.
* eg headers
Why Cassandra: distributed on disk storage without "yet another strategy"
Limitation:
* Replication factor. Should we consider a dedicated 'cache' keyspace with a
replicationFactor of one
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
