[
https://issues.apache.org/jira/browse/CASSANDRA-15393?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17173037#comment-17173037
]
Benedict Elliott Smith commented on CASSANDRA-15393:
----------------------------------------------------
{quote}But introducing an API that's error prone is dangerous. But introducing
an API that's error prone is dangerous.
{quote}
This particular issue can be somewhat resolved by using generic parameters for
methods that accept the object and accessor, including the constructor of any
object holding them, e.g.
{code:java}
public interface Accessor<Data> {}
public class SomethingWithData
{
final Object data;
final Accessor accessor;
public <D> SomethingWithData(D data, Accessor<? super D> accessor) {
this.data = data; this.accessor = accessor; }
} {code}
This avoids polluting the codebase with type parameters, but ensures the first
supplier of the Object+Accessor is safe, and later supplies are safe if they
only propagate pairs produced in this way.
We could also probably generify the objects themselves, I don't think it would
be too painful - but this at least is an easy and less invasive improvement.
{quote}The underlying issue IMO is not {{ByteBuffer}} itself but the vast
amount of those intermediate and often tiny BB instances.
{quote}
I don't have a strong opinion on the overall structural impact of this change;
I haven't thought through the various options. However I don't think we
produce all that many intermediate {{ByteBuffer}} specifically, and though we
do produce a lot of intermediate objects in many cases, there are workloads
where this isn't the issue, and object creation is still a problem.
When it comes to compaction, my preferred approach would be to avoid
materialising on heap at all - but the change here helps both compaction and
reads/writes, so it doesn't have to be either/or, and we should definitely be
trying to reduce our heap footprint in general.
> Add byte array backed cells
> ---------------------------
>
> Key: CASSANDRA-15393
> URL: https://issues.apache.org/jira/browse/CASSANDRA-15393
> Project: Cassandra
> Issue Type: Sub-task
> Components: Local/Compaction
> Reporter: Blake Eggleston
> Assignee: Blake Eggleston
> Priority: Normal
> Fix For: 4.0-beta
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> We currently materialize all values as on heap byte buffers. Byte buffers
> have a fairly high overhead given how frequently they’re used, and on the
> compaction and local read path we don’t do anything that needs them. Use of
> byte buffer methods only happens on the coordinator. Using cells that are
> backed by byte arrays instead in these situations reduces compaction and read
> garbage up to 22% in many cases.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
