Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

New price/performance leader for Ethernet switches?

0 views
Skip to first unread message

Greg Lindahl

unread,
Sep 20, 2000, 3:00:00 AM9/20/00
to
For a long time, the HP Procurve 4000M was the price/performance leader for
fast ethernet switches. One of my co-workers (John Eshleman) points out a
new contender:

http://intel.com/network/products/exp420t.htm

The 420T has 48 ports and 2 slots which take gigabit expansion cards. It is
apparently always store-and-forward, but claims 20 usec switching latencies
for 100 mbit. Street price (www.hardwarestreet.com) is $1540, the gbit
uplinks are $377.

This compares to $1943 for a 40-port HP -- half populated, but that is about
all the backplane can handle at full speed.

The Intel switch claims 21 gbits of bandwidth, and that it can run all its
ports at full speed at any packet size.

Anyone know of any other new contenders for the crown?

-- g


_______________________________________________
Beowulf mailing list
Beo...@beowulf.org
http://www.beowulf.org/mailman/listinfo/beowulf

Joel Jaeggli

unread,
Sep 20, 2000, 3:00:00 AM9/20/00
to
been looking at the dlink des-5600 which is $1300 for the chassis and
$1100 for the 4 port gig sx blades (it holds 4) which makes it fully
configured a bit under $5,000 less than a similar configuration from 3com
(9300 which we have) or extreme (3804) (though in it's defense the extreme
is a significantly more feature rich switch). you could also have it with
4 16 port ethernet blades for ~$2800 which might buy you better
performance (beyond 40 ports on the 4000M).

layer 2 fast ethernet switches have definently arrived at commidity
pricing, gig I think is a little further off, what with hp still charging
about $900 ea for single port gig blades in the 4000's.

--
--------------------------------------------------------------------------
Joel Jaeggli joe...@darkwing.uoregon.edu
Academic User Services con...@gladstone.uoregon.edu
PGP Key Fingerprint: 1DE9 8FCA 51FB 4195 B42A 9C32 A30D 121E
--------------------------------------------------------------------------
It is clear that the arm of criticism cannot replace the criticism of
arms. Karl Marx -- Introduction to the critique of Hegel's Philosophy of
the right, 1843.

chris mccraw

unread,
Sep 20, 2000, 3:00:00 AM9/20/00
to
On Tue, Sep 19, 2000 at 10:27:13PM -0400, Greg Lindahl wrote:
> For a long time, the HP Procurve 4000M was the price/performance leader for
> fast ethernet switches. One of my co-workers (John Eshleman) points out a
> new contender:
>
> http://intel.com/network/products/exp420t.htm
>
> The 420T has 48 ports and 2 slots which take gigabit expansion cards. It is
> apparently always store-and-forward, but claims 20 usec switching latencies
> for 100 mbit. Street price (www.hardwarestreet.com) is $1540, the gbit
> uplinks are $377.

i think someone should actually test the 420T before anyone gets their
hopes up...the previous generation, 410T 24port switches performed
horribly when taxed at all. streaming at full speed over >8 ports
(that is, 4 senders and 4 receivers) even with largest-possible-legal packets
brought the performance on all ports down to half, and degraded further
as more ports were taxed. we were donated some (thank you, intel!) but
had to replace them with ciscos as our apps were actually hitting a network
bottleneck far before their time.
they are decent switches for people who don't need all the bandwidth at
once...but that's what i buy hubs for.

fortunately the intel etherexpress pro/100 cards we were also donated
are still champs in the 100mbit range.

Doug Shubert

unread,
Sep 21, 2000, 3:00:00 AM9/21/00
to
>

Hi Greg and John,
Thanks for the heads up on this new Intel switch.
Our network runs primarily on HP 2424M's.
The HP spec. sheet indicates a backplane speed of 3.2 Gbps
on the 2424M and 3.8 Gbps on the 4000 & 8000.

We are looking to setup 144 nodes in a FFN design.
The 420T would work perfectly in a 12 switch, 4 NIC's per node arrangement.
At $30 per 10/100 port price, we will have to seriously consider this candidate.

Doug


>
> http://intel.com/network/products/exp420t.htm
>
> The 420T has 48 ports and 2 slots which take gigabit expansion cards. It is
> apparently always store-and-forward, but claims 20 usec switching latencies
> for 100 mbit. Street price (www.hardwarestreet.com) is $1540, the gbit
> uplinks are $377.
>

> This compares to $1943 for a 40-port HP -- half populated, but that is about
> all the backplane can handle at full speed.
>
> The Intel switch claims 21 gbits of bandwidth, and that it can run all its
> ports at full speed at any packet size.
>
> Anyone know of any other new contenders for the crown?
>

John Eshleman

unread,
Sep 21, 2000, 3:00:00 AM9/21/00
to
Agreed!!! Anybody out there have one of these 420T's.

The 420T does have a different switching fabric than the previous intel
stuff...The 410T had 9.6 (Gbps 24-ports) (supposedly) while the 420T has
21.3 Gbps (48-ports) (supposedly).

but the marketing material for the 2 differs substantially.

the 410T says "Users receive switched performance even during high traffic
bursts."
the 420T says "100% non-blocking, full duplex on all ports Handle demanding
applications at wire speed"

sounds like you got "some sort of switched performanced during high traffic
bursts" from the 410T. I'm very curious about the 420T.

john

> On Tue, Sep 19, 2000 at 10:27:13PM -0400, Greg Lindahl wrote:
> > For a long time, the HP Procurve 4000M was the
> price/performance leader for
> > fast ethernet switches. One of my co-workers (John Eshleman)
> points out a
> > new contender:
> >

> > http://intel.com/network/products/exp420t.htm
> >
> > The 420T has 48 ports and 2 slots which take gigabit expansion
> cards. It is
> > apparently always store-and-forward, but claims 20 usec
> switching latencies
> > for 100 mbit. Street price (www.hardwarestreet.com) is $1540, the gbit
> > uplinks are $377.
>

> i think someone should actually test the 420T before anyone gets their
> hopes up...the previous generation, 410T 24port switches performed
> horribly when taxed at all. streaming at full speed over >8 ports
> (that is, 4 senders and 4 receivers) even with
> largest-possible-legal packets
> brought the performance on all ports down to half, and degraded further
> as more ports were taxed. we were donated some (thank you, intel!) but
> had to replace them with ciscos as our apps were actually hitting
> a network

____________NetZero Free Internet Access and Email_________
Download Now http://www.netzero.net/download/index.html
Request a CDROM 1-800-333-3633
___________________________________________________________

0 new messages