Rehabilitate quadratic pool #12711
Rehabilitate quadratic pool #12711
Conversation
lorban
force-pushed
the
experiment/jetty-12.1.x/rehabilitate-quadratic-pool
branch
from
c62eede to
7769a33
Compare
…stic for factor Signed-off-by: Ludovic Orban <[email protected]>
lorban
force-pushed
the
experiment/jetty-12.1.x/rehabilitate-quadratic-pool
branch
from
7769a33 to
c1c1153
Compare
…tty-12.1.x/rehabilitate-quadratic-pool
jetty-core/jetty-io/src/main/java/org/eclipse/jetty/io/ArrayByteBufferPool.java
Outdated
Show resolved
Hide resolved
jetty-core/jetty-io/src/main/java/org/eclipse/jetty/io/ArrayByteBufferPool.java
Outdated
Show resolved
Hide resolved
Signed-off-by: Ludovic Orban <[email protected]>
jetty-core/jetty-io/src/main/java/org/eclipse/jetty/io/ArrayByteBufferPool.java
Outdated
Show resolved
Hide resolved
Signed-off-by: Ludovic Orban <[email protected]>
jetty-core/jetty-util/src/main/java/org/eclipse/jetty/util/MathUtils.java
Outdated
Show resolved
Hide resolved
| @@ -723,15 +724,12 @@ protected void acquire() | |||
| * A variant of the {@link ArrayByteBufferPool} that | |||
| * uses buckets of buffers that increase in size by a power of | |||
| * 2 (e.g. 1k, 2k, 4k, 8k, etc.). | |||
| * @deprecated Usage of {@code Quadratic} is often wasteful of additional space and can increase contention on | |||
There was a problem hiding this comment.
Interestingly enough, we have seen a few small contents being put into 4K buffers with the current buffer pool, so this pool may be more space efficient for those contents, as it can return a 1K buffer.
The main issue is for TLS, which needs an approx 17KB buffer, so it will get a 32KB one.
I wonder if it would be worthwhile to create a pool that wraps another pool and provides just a single specific sized buffer. So we could have a pool specialized for TLS buffers wrapping a quadratic pool? Best of both worlds?
There was a problem hiding this comment.
In the default case, we are the only users of the buffer pool and we know upfront all the sizes we're going to use.
So I wonder if we shouldn't simply have a buffer pool that works with an arbitrary list of sizes that we would hardcode to the ones we need, i.e.: 1K, 4K, 8K, 17K and 64K?
There was a problem hiding this comment.
Should be easy to implement. Give it a go.
Or how about a dynamic one, that learns the sizes it needs to allocate?
Signed-off-by: Ludovic Orban <[email protected]>
As discussed with @sbordet,
ArrayByteBufferPool.Quadraticcould be useful in certain scenarios so we should un-deprecate it and fix its minor issues.