I need to write a few thousand 24 byte values to a TCP socket which 
represents a WebSocket connection.

I must retrieve them individually from disk, but I want to group a thousand 
of them into each write, since the write represents a WebSocket message and 
each message has additional message header overhead.

At first I was instantiating a 24 byte buffer for each value, reading into 
this from disk, iterating over the array of values, calculating the size, 
creating a large buffer, copying these into it, and then writing this 
aggregate buffer out to the socket. There were a few abstractions involved 
and this was the simplest code flow.

I was concerned about the needless copying and did a rough benchmark, and 
creating a large buffer up front and writing directly into it is 3 times 
faster than creating many small buffers, then creating a large buffer and 
copying the small buffers.

I would like to change tactics and create a buffer upfront, read the small 
values directly into this from disk, and then write this to the socket. 
After receiving the callback from the write (flushed to kernel buffers) I 
would like to start writing into the same buffer at index 0.

Is it safe to modify a buffer after it has been flushed to the kernel 
buffer?

-- 
Job Board: http://jobs.nodejs.org/
Posting guidelines: 
https://github.com/joyent/node/wiki/Mailing-List-Posting-Guidelines
You received this message because you are subscribed to the Google
Groups "nodejs" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/nodejs?hl=en?hl=en

Reply via email to