Handy references:
Lab Audio
For this lab, you'll be continuing your development of a reliable data
transfer protocol over an unreliable (simulated) link, only this time, your
transfer rate should improve significantly! Your submission will still be in
the form of a library that mimics the type of functionality that you would
expect to get from an OS's transport-layer implementation.
You may implement either a go-back-N or selective repeat style of
protocol.
Requirements
- You must use C to implement your library code, and the underlying
transport that it uses to transfer messages must be UDP. TCP would give you
reliability, which would obviously defeat the purpose...
- Your protocol should provide as much performance as network conditions will
reasonably allow. If links have free space in their queues, your protocol
should attempt to take advantage of that. If it overflows the queues, it
should back off. You should mimic TCP's AIMD congestion control behavior.
You may assume that the window size will never need to go above
WINDOW_SIZE_CAP. A good default value of WINDOW_SIZE_CAP is 2000, but your
code should still work if it's lowered. I might change this value when I'm
testing, so make sure that you always refer to it via the #defined macro rather
than hard-coding a number! Please record the maximum window size you use
over the course of a run and make that value available via a function named
my_max_window that has the following prototype:
int my_max_window(int sock);
- Your protocol should provide in-order delivery. That is, the receiver
should receive packets in the same order that the sender sends them, even if
something occurs (e.g., losses) that causes a later-sent packet to arrive
before an earlier-sent one.
- Because you're pipelining multiple packets, you need to keep track of all
the outstanding data that needs to be sent and send up to an entire window at
once. This means there is no longer a 1:1 correspondence between my_send() and
send(). There are two general architectures you might use to account for this
difference. You may use either approach:
- You may code my_send() to store packets in your window in addition to
sending them. The number of packets its willing to store at any given
time will depend on your current window size. If the user ever calls
my_send when the window is full, it should wait for free window space to
open up (receiving an ACK just, just like your implementation of
my_send() in lab 5).
- You may use additional threads (at most two: one for sending, one for
receiving), which are responsible for doing the sending, receiving,
acking, timeouts, and window sizing. In this model, when the user calls
my_send(), it adds the new data to a shared location that the sending
thread will deliver later, and my_recv checks to see if there is any
buffered data to return up to the user. This is the more "realistic"
option, but is also more challenging because it requires thread
synchronization due to the shared buffers.
In either case, when the user calls my_close(), you need to ensure that all of
the outstanding data is delivered before you initiate the shutdown
sequence.
- The example code includes a test application that sits above your library
and uses it to transfer a file. You may edit the test application for your own
testing, but when grading, I will use the default version, so do not
rely on changes to these files for correct operation. Your library (lab6.h and
lab6.c) should allow any application that's built on top of it to achieve
reliable communication. When transferring files with your library, you should
get byte-for-byte identical copies. Use diff, md5sum, etc. to
veryify that everything is transferred properly.
- As a part of your implementation of reliability, you'll need to perform
RTT estimation to determine how you should set timeout values. You should use
an EWMA-based mechanism like TCP does. The library function my_rtt()
should always return your library's current estimate of the RTT. Normally you
wouldn't export such information up to the application/user, but I'll use this
to check your RTT calculation.
- Your implementation should cleanly shut down connections on both ends, even
if packets get lost. You do NOT need to implement the TCP behavior that allows
each side to shutdown independently, but in my_close(), you may want to wait
for some time, to make sure the last ACK didn't get lost (leaving one end
hanging).
Environment
For this lab, we'll be using the same mininet emulation software that
we used in lab 5, with slightly different parameters. Refer to the lab 5 environment section for details about creating a
VM. To test pipelining, you'll need buffer sizes that are larger than two
packets. When you run mininet, test with a variety of max_queue_size
parameter values. You should see the performance increase as the queue size
increases, up to a certain point of diminishing returns. A larger queue means
you're able to support more packets in flight at once.
Grading Rubric
This assignment is worth five points.
- 1½ - Your protocol delivers packets reliably and in-order, regardless of
link buffer size, with a window size that is larger than 1.
- 1 - Your protocol's performance increases significantly when network
conditions are more favorable.
- 1 - Your protocol correctly detects (within ~10%) the maximum window size
it can use for the path. (See requirement 2.)
- ½ - Your protocol cleanly terminates connections such that both ends agree
that the connection is closed.
- 1 - Your protocol has no memory leaks or invalid memory accesses (i.e.,
valgrind should not complain about anything). Your pseudo-kernel library will
need to keep track of the window and outstanding packets, and it would be
very bad for the kernel to have memory problems.
Tips
If you have any questions about the lab requirements or specification,
please post on Piazza.
Submitting
Please remove any debugging output prior to submitting.
Please do not submit output file(s) that you used in testing.
To submit your code, simply commit your changes locally using git
add and git commit. Then run git push while in your lab
directory.