gnunet-developers
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

The state of TNG


From: Schanzenbach, Martin
Subject: The state of TNG
Date: Wed, 25 Dec 2019 20:21:14 +0900

Hi,

over the past few days I have started working on 
https://bugs.gnunet.org/view.php?id=5547.

tl;dr:
Working on the communicators and tests, it is proceeding well. Time is the only 
constraint.
I hope to have unix and tcp communicators working (in tests) soon.
TBH I do not know if I will look at other communicators before starting to work 
on the actual
service.
If you want to help, try run the tests (see below) and maybe investigate the 
code.
The communicators are quite tame wrt complexity and size.

Long version
------------
The current state of TNG is as follows:

1. The main transport service (TNG service)
Untested :). It is also a monster (10k LoC) with non-trivial logic such as flow 
and congestion control.

2. Unix communicator
Works after some fixes. Can be tried

$ make test_communicator_basic_unix
$ ./test_communicator_basic_unix

Will create a logfile named test_communicator_basic.log.
The most interesting portion (if successful) will be:

* Short size packet test done.
* 5000/5000 packets in 1367377 us (468 kb/s) -- avg latency: 641916 us
* Long size packet test done.
* 5000/5000 packets in 3109296 us (51463 kb/s) -- avg latency: 618665 us
* Size packet test done.
* 12798/12798 packets -- avg latency: 1498559 us

Or something similar. The above results represent teh test cases from #5547.
A trained eye might notice the absolutely unacceptable, abysmal performance of 
this
communicator considering it is using UDS (we should expect ~XGB/s!).
As it is not meant to be used productively (and can't really) this is not a 
blocker, but
still something worth investigating. I do not know if this was also the case 
for the old
plugin.
For now, I am done with this. Will open a bug wrt performance.

3. TCP communicator
The test is test_communicator_basic_tcp.
However, it fails on initial key exchange and I have not idea why, yet.
Also, I had to disable IPv6 for now because in my VM it always binds to [::] but
advertises IPv4 addresses from NAT.

BR
Martin


Attachment: signature.asc
Description: Message signed with OpenPGP


reply via email to

[Prev in Thread] Current Thread [Next in Thread]