Hi Michael, On 6/5/06, Michael Bauroth <[EMAIL PROTECTED]> wrote:
Hi, I'm interested in both directions. First of all the incoming direction: Assume that I have 10.000 concurrent client connections sending data to the server. I've to make some plausibility checks for the data and afterwards I must store it in some way. MySQL and other DB's can't handle such amount of incoming data every second without a cluster behind the server I think. That's why I want to store (append) the data in separate CSV files and import it over the CSV engine from MySQL only when I need them. To reduce the overhead of separate file / data handling and transformation I thought that it would be a good idea to take the already existing ByteBuffers and a separate connector (Mina), which forwards the data directly to some FileChannels using NIO. In this way I have a consistent data flow chain without separate overhead.
If you have to do some check before writing the data into a file, then you have to load the data into ByteBuffer, check the integrity, and write it into the file. So there's no much benefit of using MINA to write a file IMHO. In case you want to directly link SocketChannel into FileChannel, we need some work related with this here. The outgoing direction: In the same model sometimes it can happen, that
a client needs some data stored in a file. Here I want to specify the file over a filechannel and redirect the ByteBuffer data over the chain directly to the clients (eventually using some of the filters to modify parts of the content).
This can be handled with Dave's solution. Let's wait for Dave's patch! HTH, Trustin -- what we call human nature is actually human habit -- http://gleamynode.net/ -- PGP key fingerprints: * E167 E6AF E73A CBCE EE41 4A29 544D DE48 FE95 4E7E * B693 628E 6047 4F8F CFA4 455E 1C62 A7DC 0255 ECA6
