Check Cowboy / Cleanup master build (push) Has been cancelled
Check Cowboy / Check examples (push) Has been cancelled
Check Cowboy / Cowboy (push) Has been cancelled
WEBTRANSPORT -> WT in most places. Only ENABLE_WEBTRANSPORT
remains (since it is a draft-02 setting that was later removed)
as well as internal functions and atoms about stream headers,
for greater clarity.
The draft also clarified that WT_SESSION_GONE can be sent on
the session stream and since it's what is already done a
related todo was removed.
This implements the upcoming draft-13
but has bits to make it work with draft-02
that most (all?) Chromium versions use.
Data and events are not going through
cowboy_stream beyond init. Since this
approach appears to work well it may
be a good idea to do the same for
Websocket over HTTP/2+ and improve
its performance.
This is meant to be used with clients such as Gun to simplify
proxying and similar operations. The set-cookie header must
not be set this way so there is still some extra processing
to be done to fully translate a Gun response into a Cowboy
response.
Over cleartext TCP the `protocols` option lists the enabled
protocols. The default is to allow both HTTP/1.1 and HTTP/2.
Over TLS the default protocol to use when ALPN is not used
can now be configured via the `alpn_default_protocol` option.
Performing an HTTP/1.1 upgrade to HTTP/2 over TLS is now
rejected with an error as connecting to HTTP/2 over TLS
requires the use of ALPN (or that HTTP/2 be the default
when connecting over TLS).
Before this change invalid return values would be detected
via unhelpful error messages such as [1] and the closing
of the connection.
[1] Bad value on output port 'tcp_inet'
Because the exit reason doesn't include the stacktrace they
were ignored. Now they are properly handled. The error message
was changed slightly to accomodate.
When idle_timeout is configured we do not need a separate
timer to detect inactivity. Disabling idle_timeout is not
recommended and should not be necessary.
When enabled the connection process will automatically hibernate.
Because hibernation triggers GC, this can be used as a way to
keep memory usage lower, at the cost of performance.
The problem was that when a request immediately following another
request with a large enough body, the data for the new request
(including headers) would be buffered waiting for more data,
instead of being processed immediately. If no more data came
in on the socket the request_timeout would eventually trigger.
Now the buffer is processed immediately.
In rare cases it was possible for active mode to be disabled
when there were no streams pipelined. This resulted in the
dropping of the connection due to timeouts as no data could
be received.
We now enable active mode when necessary even if there are
no streams pipelined.
This was found while benchmarking and I have not been able
to extract a test case.
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.
Before this commit frames could "cheat" by compressing data
below the limit which would get expanded above the limit.
Now Cowboy will stop decompressing data when the limit is
reached.
Following the same strategy as Websocket described in
commit cbed21c383e4cebb7df5a0a8b81f18c1738bef3e
Gains are comparable as far as Websocket over HTTP/2
is concerned.
`perf` has shown that Cowboy spends a lot of time
cancelling and starting this timer. Instead of resetting
for every data received, we now only reset a field in the
state.
Before it was working like this:
- start idle timeout timer
- on trigger, close the connection
- on data, cancel and start again
Now it's working like this:
- start idle timeout timer for a tenth of its duration, with tick number = 0
- on trigger, if tick number != 10
- start the timer again, again for a tenth of its duration
- increment tick number
- on trigger, if tick number = 10
- close the connection
- on data, set tick number to 0
It benchmarks binary, ascii, mixed and japanese data
using Websocket and Websocket over HTTP/2.
HTTP/2 options get set to ensure that performance is
better than the default HTTP/2 options.
It switches to Gun and Ranch branches that include
fixes that are required for tests to complete successfully.
