Enabling gigabit to the desktop - or to lots of desktops - is the name of the game for Extreme Networks' new Summit 400-48t, a 10G Ethernet workgroup switch.

In Network World's exclusive test, we pushed more than 70 million packets per second through the S400. Results were generally good. The switch delivered low, consistent delay under heavy loads, and it offers a long list of features, including 802.1X authentication for security.

On the downside, the S400's optional 10G Ethernet module is a bit pricey for a workgroup switch, although at about $4,000 per 10G Ethernet port (plus optics) it's still competitive. Also, the S400 posted subpar numbers in some tests of jumbo frames. On balance, though, the S400 is a more-than-capable performer: In fact, it's the fastest wiring closet switch we've tested.

The S400 fits in 1U and offers 48 10/100/1000M bit/sec copper ports, with an optional module for two 10G Ethernet interfaces. Features include support for major routing protocols, IPv4 multicast and up to 4,096 virtual LANs. External redundant power is also an option.

Performance tests

We conducted separate tests of the S400's 10G Ethernet and Gigabit Ethernet handling. Because the 10G Ethernet interfaces fit in an optional module, users can buy the switch with 10/100/1000 copper interfaces and add 10G Ethernet later.

Our 10G Ethernet tests measured how quickly the S400 could move traffic between 10G Ethernet uplink ports and Gigabit Ethernet downlink (edge) ports. We configured a Spirent Communications SmartBits traffic generator/analyzer to send test frames to the S400's uplink and downlink ports

In a nutshell, the S400 delivered line-rate throughput in all the 10G Ethernet test cases we tried (see graphic, below). These included baseline tests at Layer 2 and Layer 3 and tests with one and two 10G Ethernet interfaces exchanging traffic with 10 and 20 edge gigabit interfaces. The S400 didn't drop a single frame in any of these tests.

Click to see: Measuring 10G throughput

Access control lists (ACL), which can severely degrade performance on some switches, posed no problem for the S400, either. We reran the 10G Ethernet baseline tests with the maximum number of ACLs applied to every switch interface; for the S400, that is 124 ACLs per port on all 50 ports. Again, the switch delivered line-rate throughput.

Delay was also low and constant across all our 10G Ethernet tests (see graphic, below). The average delay and jitter numbers we observed are comparable to other high-performing 10G Ethernet backbone switches we've tested. Even the highest maximum delay number we observed - 104.4 microsec when moving jumbo frames - is nowhere near enough to have an appreciable effect on application performance.

Click to see: Measuring delay

Trying to get a Gig

The S400's performance in gigabit-only tests, while generally good, didn't match the flawless levels in the 10G Ethernet events.

Throughput was equivalent to 92% of line rate or better in all the tests we ran using standard Ethernet frame lengths. We observed a slight difference depending on whether we used random or non-random media access control (MAC) addresses; on production networks, random addresses are far more common. This won't necessarily harm application performance, because few (if any) production networks have sustained utilization over 90%.