-- Page: 1
There will be 8 computers:
All have 10/100 NICs in the them and
are running at 100 base, full duplex.
The 5 Pentiums will be running a linux
IGMP DoS (Denial of Service) attacker, attacking each other to
simulate a network load. The linux 2.4.2 kernal seems to
ignore these packets so it doesn't have the same effect as it does
with windows (ie killing it). These machines can send out
around 3.5-4.1 megabytes/sec roughly the same as 5 computers doing
TCP/IP file transfers (ie. Windows network transfer) without having
to worry about performance hit of writing to a hard drive on an old
Pentium machine. The settings for the IGMP attacker is as follows:
65K packet sizes aimed at port 1
The Celeron and the Tbird will do a majority of the
benchmarks since they will be least affected due to low resources.
The Pentium2 350 will be monitoring the Pentiums DoS attack and
recording the incoming and outgoing speeds, which will be reported
The reason why we aren't using the P2 350 as part of the
benchmarking is that it returns a very high error rate and very slow
Each network transfer test will be done 3 times and
we will take the average of the three transfers.
- Timed windows network transfer, with and without network
traffic: Using a 696 Meg test file, we will time how long it
takes to transfer the test file over the network. We will take
the average of 3 transfers both with and without network
- Timed windows FTP transfer with and without network traffic:
Again, using a 696 Meg test file, we will time how long it takes
to send the test file via FTP. The Ftp client that we used is
LeapFTP, and the server that was used was Serv-U FTP (for those
who think it matters). We will take the average of 3 transfers
both with and without network traffic.
- Timed Linux FTP transfer, with and without network traffic:
Using our 696 Meg test file, we will time how long it takes to
send the test file via FTP. The Ftp client that we used is NcFTP,
and the server that was used was wuftpd. We will take the
average of 3 transfers both with and without network traffic.
- Sandra network throughput bench.
- IGMP attack: We will use the IGMP attacker to test the full
duplex of the switch, with the Celeron and the Tbird both DoSing
each other. We will record the average amount of data (in
megs/s) that was transferred in the attack.
- MAC address buffer refresh times: It seems that at every LAN
party there is a self-proclaimed know-it-all with a weak ass
box, well at least there are at ours. And this know-it-all
swears up and down that there is a "sweet port" on the
switch and he is determined to find it. So he goes through and
switches out his computer with another persons computer until he
is satisfied that he has found the "sweet port." So
what ends up happening on some switches? Everybody ends up not
being able to see the network, why? Because the internal MAC
address buffer in the switch hasn't refreshed, so its still
trying to switch the packets to the computer that use to be
there, but is now in a new port. So what you have to end up
doing is unplugging the switch from the wall and re-plugging
back in, or wait for the buffer to be refreshed and the
know-it-alls blood spilled all over the floor.