[
https://issues.apache.org/jira/browse/CASSANDRA-13981?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16463929#comment-16463929
]
Michael Burman commented on CASSANDRA-13981:
--------------------------------------------
Just my 2 cents, but the approach seems slightly odd if one wants to take
advantage of the persistent memory. This stuff should override the whole
CommitLog path also. So instead of writing first to WAL and then to the
memtable, this should skip both the memtable as well as commitlog in my opinion
and basically rewrite the whole storage path. And same applies to reads,
compactions, etc as it should never create SSTables. Same way, when starting
Cassandra the recovery process should be changed also.
I would actually redesign also the query parsing, query execution and other
parts as well as these do not fit the model of persistent memory too well
either. I have no idea how this would affect the streaming processing as
suddenly the tables are not necessarily anymore immutable (as LSM isn't
designed for persistent memory). I'd say this is quite difficult approach to
integrate to Cassandra if it's anything more than "half-baked exercise". Just
replacing the data model isn't enough as the whole approach of having a single
type of memory is so fundamentally different to what Cassandra was designed for.
> Enable Cassandra for Persistent Memory
> ---------------------------------------
>
> Key: CASSANDRA-13981
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13981
> Project: Cassandra
> Issue Type: New Feature
> Components: Core
> Reporter: Preetika Tyagi
> Assignee: Preetika Tyagi
> Priority: Major
> Fix For: 4.0
>
> Attachments: in-mem-cassandra-1.0.patch, readme.txt
>
>
> Currently, Cassandra relies on disks for data storage and hence it needs data
> serialization, compaction, bloom filters and partition summary/index for
> speedy access of the data. However, with persistent memory, data can be
> stored directly in the form of Java objects and collections, which can
> greatly simplify the retrieval mechanism of the data. What we are proposing
> is to make use of faster and scalable B+ tree-based data collections built
> for persistent memory in Java (PCJ: https://github.com/pmem/pcj) and enable a
> complete in-memory version of Cassandra, while still keeping the data
> persistent.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
