| Age | Commit message (Collapse) | Author |
|---|
|
There's not a big performance difference between 8192 and 1024
so let's use less memory at the start of the connection.
|
|
Cowboy will set the socket's buffer size dynamically to
better fit the current workload. When the incoming data
is small, a low buffer size reduces the memory footprint
and improves responsiveness and therefore performance.
When the incoming data is large, such as large HTTP
request bodies, a larger buffer size helps us avoid
doing too many binary appends and related allocations.
Setting a large buffer size for all use cases is
sub-optimal because allocating more than needed
necessarily results in a performance hit (not just
increased memory usage).
By default Cowboy starts with a buffer size of 8192 bytes.
It then doubles or halves the buffer size depending on
the size of the data it receives from the socket. It
stops decreasing at 8192 and increasing at 131072 by
default.
To keep track of the size of the incoming data Cowboy
maintains a moving average. It allows Cowboy to avoid
changing the buffer too often but still react quickly
when necessary. Cowboy will increase the buffer size
when the moving average is above 90% of the current
buffer size, and decrease when the moving average is
below 40% of the current buffer size.
The current buffer size and moving average are
propagated when switching protocols. The dynamic buffer
is implemented in HTTP/1, HTTP/2 and HTTP/1 Websocket.
HTTP/2 Websocket has it disabled because it doesn't
interact directly with the socket; in that case it
is HTTP/2 that has a dynamic buffer.
The dynamic buffer provides a very large performance improvement
in many scenarios, at minimal cost for others. Because it largely
depend on the underlying protocol the improvements are no all equal.
TLS and compression also impact the results.
The improvement when reading a large request body, with the
requests repeated in a fast loop are:
* HTTP: 6x to 20x faster
* HTTPS: 2x to 6x faster
* H2: 4x to 5x faster
* H2C: 20x to 40x faster
I am not sure why H2C's performance was so bad, especially compared
to H2, when using default buffer sizes. Dynamic buffers make H2C a
lot more viable with default settings.
The performance impact on "hello world" type requests is minimal,
it goes from -5% to +5% roughly.
Websocket improvements vary again depending on the protocol, but
also depending on whether compression is enabled:
* HTTP echo: roughly 2x faster
* HTTP send: roughly 4x faster
* H2C echo: roughly 2x faster
* H2C send: 3x to 4x faster
In the echo test we reply back, and Gun doesn't have the dynamic
buffer optimisation, so that probably explains the x2 difference.
With compression however there isn't much improvement. The results
are roughly within -10% to +10% of each other. Zlib compression
seems to be a bottleneck, or at least to modify the performance
profile to such an extent that the size of the buffer does not
matter. This happens to randomly generated binary data as well
so it is probably not caused by the test data.
|
|
This can be used to limit the maximum frame size before
some authentication or other validation is completed.
|
|
|
|
While we are identified as a supervisor in the tree,
we no longer manage children processes at that point,
so do not need to trap exit signals. Users can still
enable trap_exit if they prefer to.
|
|
|
|
This reduces the number of times we need to ask for more packets,
and as a result we get a fairly large boost in performance,
especially with HTTP/1.1.
Unfortunately this makes Cowboy require at least Erlang/OTP 21.3+
because the ssl application did not have active,N. For simplicity
the version required will be Erlang/OTP 22+.
In addition this change improves hibernate handling in
cowboy_websocket. Hibernate will now work for HTTP/2 transport
as well, and stray or unrelated messages will no longer cancel
hibernate (the process will handle the message and go back into
hibernation).
Thanks go to Stressgrid for benchmarking an early version of this
commit: https://stressgrid.com/blog/cowboy_performance_part_2/
|
|
It has been deprecated in OTP and the new way is available
on all supported OTP versions.
|
|
This allows changing the normal exit reason of Websocket
processes, providing a way to signal other processes of
why the exit occurred.
|
|
|
|
The old interface with ok|reply|stop tuples is deprecated.
|
|
This allows disabling the UTF-8 validation check
for text and close frames.
|
|
|
|
They allow the server to configure what it is willing to accept
for both the negotiated configuration (takeover and window bits)
and the other zlib options (level, mem_level and strategy).
This can be used to reduce the memory and/or CPU footprint of
the compressed data, which comes with a cost in compression ratio.
|
|
Thanks Artem.
|
|
|
|
Option allows to limit a frame by size before decoding its payload.
LH: I have added a test for when the limit is reached on a nofin
fragmented frame (the last commit addressed that case but it had
no test). I have fixed formatting and other, and changed the
default value to infinity since it might otherwise be incompatible
with existing code. I also added documentation and a bunch of other
minor changes.
|
|
|
|
I have decided not to include a manual page for
cowboy_stream_h at this point because it clashes
with the cowboy_stream manual page. This decision
will be revisited in the future.
|
|
This option allows customizing the compacting of the Req object
when using Websocket. By default it will keep most public fields
excluding headers of course, since those can be large.
|
|
Before this commit we had an issue where configuring a
Websocket connection was simply not possible without
doing magic, adding callbacks or extra return values.
The init/2 function only allowed setting hibernate
and timeout options.
After this commit, when switching to a different
type of handler you can either return
{module, Req, State}
or
{module, Req, State, Opts}
where Opts is any value (as far as the sub protocol
interface is concerned) and is ultimately checked
by the custom handlers.
A large protocol like Websocket would accept only
a map there, with many different options, while a
small interface like loop handlers would allow
passing hibernate and nothing else.
For Websocket, hibernate must be set from the
websocket_init/1 callback, because init/2 executes
in a separate process.
Sub protocols now have two callbacks: one with the
Opts value, one without.
The loop handler code was largely reworked and
simplified. It does not need to manage a timeout
or read from the socket anymore, it's the job of
the protocol code. A lot of unnecessary stuff was
therefore removed.
Websocket compression must now be enabled from
the handler options instead of per listener. This
means that a project can have two separate Websocket
handlers with different options. Compression is
still disabled by default, and the idle_timeout
value was changed from inifnity to 60000 (60 seconds),
as that's safer and is also a good value for mobile
devices.
|
|
|
|
Also fixes a small mistake in cowboy_websocket.
|
|
|
|
https://github.com/ninenines/cowboy/blob/master/examples/websocket/src/ws_handler.erl
|
|
A few small revisions were made, and Erlang.mk has been updated.
|
