Expose support for asynchronous interactions

Servlet 3.1 and frameworks like Spring 4 support non-blocking request execution. This partially supported by the raw Storage API which exposes its HttpRequest objects with their executeAsync() operation. However, this does not support non-blocking transfers.

Full async support would allow an application to transfer data to/from Cloud Storage without having to consume a Thread while waiting for IO to complete.

I doubt that will change while using google-api-java-client (did you create a feature request there we can link too?), so the current plan is to support it for every service that is available via gRPC.

@jgeewax

@jgeewax I don't this should not be a Needed for public announcement blocker but rather more of a future enhancement (when service supports it).

I can buy that. We could add these later pretty easily. Moving out the of the milestone.

+1

Turns out that AppEngine Standard has async support, Jetty 3.1 in Flex supports Async (at Top of Trunk), for generic Java, there is Apache's HttpAsyncClient

I think we should just naturally surface the async functionality as we move each API to gRPC instead of having an old tracking issue. Closing this out.

Has this been addressed elsewhere? I'm not finding another issue with this topic.

What is currently the best way to stream a file from storage with Netty? As in https://github.com/netty/netty/blob/60250f3795d4082e59df6806dade2966ec583859/example/src/main/java/io/netty/example/http/file/HttpStaticFileServerHandler.java#L188-L202.

The way I have this working now is:

ReadChannel reader = storage.reader("bucket", "blob");
InputStream in = Channels.newInputStream(reader);
ChunkedStream stream = new ChunkedStream(in);
HttpChunkedInput input = new HttpChunkedInput(stream);

I send the HttpChunkedInput stream through the Netty channel and ChuckedWriteHandler writes the stream out in chunks to the client.

I wonder, though: the ReadChannel reader, is it non-blocking? I looked at the code and there are some functions that sleep the thread.