Alloy NS-16T Manual do Utilizador descarregar pdf (Página 2)

CAIA Technical Report 031217A December 2003 Page 2 of 5

card 3 was connected to as the source of all the MAC

addresses, thus packets would then be sent to card 3. To

avoid possible CRC errors, fragmented/undersized

packets, card 1 cycled through the first 4,000 MAC

addresses in the CAM table and card 2 through the last

4,000 MAC addresses. This test was repeated with the

number of packets per card increasing until flooding

occurred after 8,320 MAC addresses. To validate the

results the test was repeated with card 4 filling the CAM

table and cards 1, 2, and 3 sending packets to it. Again,

there were no errors in these tests.

The next step of this investigation focused on

repeating the tests above but this time instead of half (or

a third) of the destination addresses being sent by each

card, all two (or three) cards simultaneously cycled

though the 8,000+ MAC addresses to the destination

card. It was once again found that the CAM table could

hold 8,320 MAC addresses before flooding occurred.

Also, there were no CRC errors and/or

fragmented/undersized packets recorded by

SmartWindow.

IV.FLOODING TCPDUMP

A. Using tcpdump to investigate packet collisions.

This set of tests was to investigate whether starting

the cards at different times would create any difference

in packet error or loss. We hypothesized that errors

could be caused by multiple packets arriving at the same

time on the switch’s backplane and colliding. By starting

the cards at different times, the packets might not collide

on the backplane.

For these tests the packet inter-arrival time was set to

200ms so as to ensure that the rate of packet arrivals is

slow and gives enough time for packets to pass through

the switch. The size of the packet payload was set to

1,024 bytes. All four cards flooded tcpdump with 10,000

packets each (see Figure 1, page 1) to send a total of

40,000 packets. This was so that tcpdump could record

all packets that were sent out of the switch and measure

the time intervals between these packets.

B. Card Group Burst.

In the first test, all four cards were started at the same

time using the Group feature of SmartWindow. Table 1

below shows packet loss and CRC errors were recorded.

Tcpdump only captured 38,714 packets from the switch.

Card 1 Card 2 Card 3 Card 4

Packets

Received

24,465 25,878 25,878 29,524

CRC Errors

0 0 4 13

Frag/Undersize

6 0 14 52

Table 1: Group burst

C. Start All Cards Burst.

In the second test instead of starting the Group

feature of SmartWindow we used the option “Start All

Cards”. This option starts all cards with a 50 to 100ms

lag between each card beginning to transmit. As we can

see from Table 2, there were only very few errors, most

likely because the packets did not converge on the

switch backplane at the same time. Tcpdump captured

39,836 packets, the rest being dropped by the switch.

Card 1 Card 2 Card 3 Card 4

Packets

Received

29,918 29,917 29,835 29,835

CRC Errors

0 1 0 0

Frag/Undersize

6 6 0 0

Table 2: Start All Cards burst

D. Manual Card Burst.

The final test involved manually starting each card.

This resulted in 1 to 2 seconds delay between each card

starting to transmit packets. As can be seen in Table 3,

there were no CRC errors, fragmented/undersized

packets and all 40,000 packets arrived and were

accounted for.

Card 1 Card 2 Card 3 Card 4

Packets

Received

30,000 30,000 30,000 30,000

CRC Errors

0 0 0 0

Frag/Undersize

0 0 0 0

Table 3: Manual burst

E. Verifying timestamps of packets in the tcpdump file.

Each tcpdump test above produced a tcpdump file

that was consulted as to the reason for the packet loss

behaviour. Since the alleged errors occur in the switch,

not all packets arrived to the machine running tcpdump.

As can be seen in Figure 2, both the test run manually

and with the “Start All Cards” option have a relatively

even gradient, where the rate of packet arrivals over time

is steady. It can be seen that the Group option graph does

not have an even gradient, suggesting that the

cumulative number of packets falls due to packets being

dropped by the switch. Figure 3 shows a close up of the

beginning of the test. Each point on the graph represents

one packet. As seen for the Group and Start All graphs,

four bursts of packets are followed by a 0.2 sec interval

of time where no packets arrive. Also, the time interval

between the Group packets within a burst is slightly

smaller than the Start All option packets. The Manual

graph clearly shows between 1 and 2 seconds several

packets were sent by only one card, corresponding to the

time it took to manually click “Start” on the second card.

1 2 3 4 5

Comentários a estes Manuais

Sem comentários

Alloy NS-16T Manual do Utilizador Página 2

Comentários a estes Manuais

Manuais e produtos relacionados com Comutadores de rede Alloy NS-16T