Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 8 Apr 2000 20:28:07 -0500 (CDT)
From:      Joe Greco <jgreco@ns.sol.net>
To:        javier@nyi.net (Javier Frias)
Cc:        dev@inetu.net, isp@freebsd.org
Subject:   Re: flat network
Message-ID:  <200004090128.UAA92724@aurora.sol.net>
In-Reply-To: <38ECE636.CE86D01C@nyi.net> from Javier Frias at "Apr 6, 2000  3:32: 6 pm"

next in thread | previous in thread | raw e-mail | index | archive | help
> I've  had quite a bid different experience with vlans.
> At least on catalyst switches.

Don't talk to me about Catalyst switches.  Not today.  Cisco sucks.

> with vlans, arp is reduced. ARP works by broadcasting a packet to all
> hosts attached 
> to an Ethernet segment. Since a vlans virtually reduces the number of
> hosts
> on each "Ethernet Segment" or "vlan", the number of host an arp request
> reaches
> is smaller.

You're not talking to an idiot.  With routing protocols, ARP is reduced
further, to the point where the only ARP traffic on the network is for the
physical interfaces present on the network.

If you have one router and one machine on a network, with ARP you still have
the potential to have as many ARP entries as you do virtual servers.  If you
would like a practical demonstration of why this is bad, go generate about
65,000 virtual servers on such a machine, and then ask for stuff from all of
them.  Note the behaviour of the ARP cache on your routers and switches.
The behaviour is O(N), and you are screwed when N exceeds the capacity of
the ARP table on the device.  God forbid you've more than one server on the
net!

If you have one router and one machine on a network, with OSPF you have
exactly two ARP entries - and no need for the router to ARP for each virtual
server.  If you would like a practical demonstration of why this is good,
do the same test as above.  The behaviour is O(1).

> To move servers between facilities you need a flat network? you can move
> vlans accross
> switches. Plus, there are quite a large nmber of ways you can do this,
> without
> the need of a huge flat network.

I wasn't proposing the creation of a huge flat network.  My largest
production network has a netmask of 0xfffffff0.  I move servers between
facilities with no problems, thanks to OSPF.  Do a traceroute to both
dns1.sol.net and dns2.sol.net, numbered right next to each other, for a
trivial example.

> vlans also offer quite a big more security than a flat network.
> crosstalk is almost eliminated.
> 
> In the isp market, how some companies provide colocation
> without giving a customer a separate vlan is beyond my comprehesion.

Use a separate routed network.  Broaden your horizons.
-- 
... Joe

-------------------------------------------------------------------------------
Joe Greco - Systems Administrator			      jgreco@ns.sol.net
Solaria Public Access UNIX - Milwaukee, WI			   414/342-4847


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-isp" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200004090128.UAA92724>