Question

I have a server reading inbound newline-delimited strings. The pipeline looks like this:

    ...
    pipeline.addLast("framer",
        new DelimiterBasedFrameDecoder(65535, Delimiters.lineDelimiter()));

    // plain old strings
    pipeline.addLast("decoder", new StringDecoder());

    // callback to handler
    pipeline.addLast("handler", new ConnectorHandler(collector));
    ...

When I change the decoder to the following:

    pipeline.addLast("framer", 
        new DelimiterBasedFrameDecoder(65535, 
            new ChannelBuffer[] { ChannelBuffers.wrappedBuffer(
                new byte[] {'\n'})})); 

I get a 3-4X increase in server speed. I don’t understand why this would be the case. The inbound strings vary betweeen about 1K-4K in length, around 10K strings/sec. Looking at the source code, it appears that it cycles through all the delimiters to find the one yielding the minimum frame length — by inspection, I don’t see how looking for only one delimiter would result in such a performance gain over looking for two delimiters. I am tempted to try to write my own decoder but I would like to understand what’s going on here. Any insight would be appreciated.

Answer 1

thats exactly the case.. looking only for one is much more cheap.