• Advertisement
Sign in to follow this  

Sending A Hash List

This topic is 2055 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi everyone!

I need to send a Hash List from a client to a server and was wondering on the best way to do this. I have a connection to the server and can send data fine but I was wondering how I should handle the procedure. I'm using TCP and was wondering if I could just get the client to just fire the data one Hash at a time and allow it to queue up on the server end. I'll be sending over a thousand of them and was wondering if I was just to send a load of seperate messages if it would be considered hammering the server?

Should I get the server to send a confirm message after it's received eash hash and then force the client to wait for the confirm before sending the next data. I want the data to be sent as quick as possible so I'd rather not have to wait for a confirm before each message.

Thanks in advance!
David

Share this post


Link to post
Share on other sites
Advertisement
When you already use TCP, you have a reliable stream. If you need some kind of messaging, just send a start message, then stream the whole map (key/value) and finally a end-message. There's no need of a confimation message, if the communication breaks, you will get an according error message from the TCP implementation.

Share this post


Link to post
Share on other sites
Thanks for your quick reply!

OK, maybe I'm not understanding how this works. I'm sending data using the send() function. Does it matter how often I make calls to this function? Once I've made a connection to the server can I send data as fast as I like without worrying about hammering the server or overloading any buffers? Or should I put the hash list into a single buffer and send just once? or would multiple calls to the send() function just join them up anyway?

Thanks
David

Share this post


Link to post
Share on other sites
send() will send at least one byte, but may not send all the bytes in the buffer. As long as you handle this case (re-sending the parts that weren't sent the first time,) you're good.

There is both buffering and flow control built-into TCP. The server will not actually receive data faster than it can handle it. Once the client tries to send more than the server can handle, the client's outgoing buffer will fill up, and send() will either start sending less than you asked it to, or will block and pause the client until space frees up. In either case, data won't flow across the wire faster than the server can deal with it.

Share this post


Link to post
Share on other sites
OK, so lets say I max out the buffer and the send() function only queues half the data that I sent to it. What's the best way to handle this? Do I have to keep retrying the send with the remaining data?
Is there a way to make it only queue the data if the full buffer is accepted because I'd rather not have to hang on a partial send and resend the whole message when the buffer is free.

Thanks
David

Share this post


Link to post
Share on other sites
Do I have to keep retrying the send with the remaining data?
[/quote]

Yes.

send() is guaranteed to make at least some progress each time you call it, so a tight loop that just calls send() and updates the send pointer will eventually terminate (assuming you break out of the loop if send() returns error.)

If you are on Linux, you can use write() instead of send(); write() will block until all the data is sent, or the socket has an error.

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement