It sounds like you are sending one sample per request? If so, I agree with Federico: you are asking too much of the network to make that many requests per second. There is a lot of overhead involved in contacting the server, the server accepting the connection, sending request headers, processing the data, receiving response data, then closing the connection. Then you want to do it all over again a millisecond later? It's not the Yun that is the limitation, you're just asking too much of the network because you're trying to do a lot more work than is necessary.
Suppose you're giving a book to a friend. Do you take the first page over, knock on his door, wait form him to answer, give him the first page, say goodbye, then go home to get the next page and start it over again? Of course not, you take the whole book over at once.
You should consider doing the same thing here. Have the Yun collect a bunch of samples, and store them in an array in RAM. Then when you have a few hundred samples, make the connection to the server and send them all at once. You will be spending more time actually sending data, and less time sending all of the overhead to open and close connections so often.