Network switches?

Hi all,

Can anyone recommend a fast and reliable network switch to use with
FLEET? Some of our customers report switch problems when large packets
are fragnented by Net.

Any “success stories”?

Thanks in advance.

  • Nick

“Nikolai Gorbunov” <n.gorbunov@swd.ru> wrote in message
news:8pi8lf$91v$1@inn.qnx.com

Hi all,

Can anyone recommend a fast and reliable network switch to use with
FLEET? Some of our customers report switch problems when large packets
are fragnented by Net.

Before change for a new switch, I’d investigate what the “real” problem is.

What does “large packets are fragmented by Net” mean?

Any “success stories”?


Thanks in advance.

  • Nick

We use Cisco Catalyst. I was on the phone with there tech support and
they said fleet should be ok.

Nikolai Gorbunov wrote:

Hi all,

Can anyone recommend a fast and reliable network switch to use with
FLEET? Some of our customers report switch problems when large packets
are fragnented by Net.

Any “success stories”?

Thanks in advance.

  • Nick

We’re using the “3Com Office Connect Dual Speed Switch 8” with no problem,
both DiskShadowed nodes, and between those nodes and uplinked to our WinNT
Network.
We haven’t had or looked for timing issues.
-Paul

J. Scott Franko <jsfranko@switch.com> wrote in message
news:39BD0646.A9C6DE27@switch.com

We use Cisco Catalyst. I was on the phone with there tech support and
they said fleet should be ok.

Nikolai Gorbunov wrote:

Hi all,

Can anyone recommend a fast and reliable network switch to use with
FLEET? Some of our customers report switch problems when large packets
are fragnented by Net.

Any “success stories”?

Thanks in advance.

  • Nick

We have had very good success with Cisco Catalyst 2900 switches in several
redundant QNX networks running pure FLEET. The isolation provided by the
switch also makes it easier to troubleshoot large networks.

“Nikolai Gorbunov” <n.gorbunov@swd.ru> wrote in message
news:8pi8lf$91v$1@inn.qnx.com

Hi all,

Can anyone recommend a fast and reliable network switch to use with
FLEET? Some of our customers report switch problems when large packets
are fragnented by Net.

Any “success stories”?

Thanks in advance.

  • Nick

Before change for a new switch, I’d investigate what the “real”
problem is.

Yes, but it’d be better to use a proven hardware…

What does “large packets are fragmented by Net” mean?

Sorry for sacrificing accuracy to brevity… :slight_smile:

The “real” problem concerns transmitting lagre messages (which don’t
fit into a single Ethernet frame) over a switch-segmented network -
under a heavy load the packets just get lost. Acccording to the
customer’s report, the switches are set up right. When hubs are used
instead of switches, everything goes OK. Important: when message
fragmentation is being done at the application layer, so that every
fragment fits into ~1K, the problem disappears.

I didn’t ask them for netinfo logs yet (though maybe I should) for my
suspicions make me think of the switch misbehaviour, maybe packet
buffer overflow due to poor packet switch timing.

  • Nick

When you send a larger-than-one-Ethernet-packet message in QNX, the driver
will send two or more packets in succession to transmit the message without
waiting for acknowledgments between packets. This is, of course, important
for high network throughput. However, if the switch has limited buffer
space and no flow control mechanism, then sending large QNX messages may
cause buffer overflow in the switch, resulting in lost packets.

I would expect almost all currently sold switches to have fairly large
internal buffers, and to implement flow control to prevent buffer overflow.
We offer some low-cost unmanaged 5-port and 8-port 10/100TX Ethernet
switches that provide these features, and that we have tested (fairly
heavily) with QNX traffic. If you’re interested, check them out at
www.cormantech.com.

Regards,

Bert Menkveld
Engineer (NOT sales :slight_smile:
Corman Technologies Inc.
bert@cormantech.com

Nikolai Gorbunov <n.gorbunov@swd.ru> wrote in message
news:8pl4jr$bd6$1@inn.qnx.com

Before change for a new switch, I’d investigate what the “real”
problem is.

Yes, but it’d be better to use a proven hardware…

What does “large packets are fragmented by Net” mean?

Sorry for sacrificing accuracy to brevity… > :slight_smile:

The “real” problem concerns transmitting lagre messages (which don’t
fit into a single Ethernet frame) over a switch-segmented network -
under a heavy load the packets just get lost. Acccording to the
customer’s report, the switches are set up right. When hubs are used
instead of switches, everything goes OK. Important: when message
fragmentation is being done at the application layer, so that every
fragment fits into ~1K, the problem disappears.

I didn’t ask them for netinfo logs yet (though maybe I should) for my
suspicions make me think of the switch misbehaviour, maybe packet
buffer overflow due to poor packet switch timing.

  • Nick