WideBand managed Ethernet switch: Affordable and fast

Network managers driven by frugality, patriotism or both, might want to consider the WideBand WB28GMPRO, a low-cost managed Gigabit Ethernet switch made in the American heartland.

WideBand says it gains a price advantage by manufacturing in Missouri, where labor costs are relatively low. Nearly all other network equipment is made in Asia, often through outsourcing to component assembly firms.


How we tested WideBand switch

Chart tracks switch latency

Archive of Network World tests

Subscribe to the Network Product Test Results newsletter


The WB28GMPRO performed well in some areas of our tests, while lacking polish in others. This is a fast switch, delivering line-rate throughput for all frame sizes in tests lasting 60 seconds, and near-line-rate throughput for tests lasting 300 seconds. Latency was in line with other low-cost gigabit switches we've tested.

On the downside, the switch's user interface is quite limited in terms of features supported, and we were unable to complete a test of link aggregation because of performance issues.

Switch setup is fast but not entirely straightforward. Most switches offer a command-line or Web interface, and usually both. In contrast, WideBand's Windows-based management software has a proprietary interface, accessible via serial or Ethernet ports.

That's where we hit our first snag: Software supplied with the switch would communicate only over a serial link attached to COM1 of a PC running Windows. That was a problem for us, because the machine we used for configuration allocated COM1 to an infrared port. Within a day, WideBand released an updated version of the management software that let us select serial ports.

Even so, we'd be happier with a simple command-line interface (CLI) to the switch. A CLI also has the advantage of not requiring a given operating system or serial port. We'd be even happier if the switch management software supported Secure Shell for remote access.

The management interface is serviceable but limited compared with competing offerings. The interface displays information about port counters, virtual LAN (VLAN) assignment, SNMP and link aggregation. One notable feature is the switch's support for 4,094 VLAN IDs; many access switches support only a few hundred VLANs.

A port can be assigned to as many as four VLANs based on frame type, and that port will accept untagged traffic from each VLAN. The switch also will accept tagged frames, but managing trunk ports requires WideBand's nMU network management software, which we didn't test. We found a few other functions available only through nMU, such as jumbo frame configuration (though jumbo handling is enabled by default) and controlling address aging timers.

Performance measurements

In performance tests, we measured the WideBand switch in four areas: throughput, latency, address learning capacity and link aggregation (see How we tested WideBand switch). We determined throughput and latency by attaching a Spirent TestCenter traffic generator/analyzer to all 28 ports of the switch and running TestCenter's RFC 2889 suite of switch tests.

It's common practice to run this type of test for 60 seconds, and here the WideBand switch was perfect. For every frame length, from the Ethernet minimum of 64 bytes all the way to 9,216-byte jumbo frames, the switch forwarded traffic at line rate without dropping a frame (see WideBand throughput chart, below).

WideBand throuughput

When we increased the test duration to 300 seconds - a practice increasingly used by service providers to model long-lived flows such as video feeds - the switch forwarded traffic without loss at a rate equivalent to 99% of line rate for all frame sizes.

The distinction between 99% and 100% of line rate is academic for most enterprise networks, where use is usually far lower. However, for applications that require zero frame loss, this isn't the right switch (or the right price range, for that matter).

We also measured switch latency, which was much higher in our 60-second tests than in our 300-second tests (see WideBand latency chart, below). That's because throughput was lower in the 300-second tests, and standard testing practice is to measure latency at the throughput level. We should note that latency with this switch is comparable with that for other low-cost gigabit switches we've tested.

WideBand latency

Latency at line rate may be much higher than at the 99% level, but that isn't necessarily a cause for concern. Delay introduced by the WideBand switch in any of our tests is unlikely to hamper application performance.

Congestion concerns

The one place where switch latency could be a concern is in congestion handling. At line rate, the lowest latency we recorded was 72 microsec with 64-byte frames. At gigabit line rate, that means there are more than 100 frames outstanding between transmitter and receiver at any one time. The loss of any one of those frames (perhaps because of congestion somewhere else in the network) could cause slowdowns for connection-oriented protocols such as TCP.

In a test of learning media access-control addresses - the maximum number of hosts a switch can see without flooding traffic - the switch learned 4,857 addresses. That's plenty for an access switch like this.

Link aggregation has long proven a sore point in switch testing, and unfortunately the WideBand device is no exception. We were unable to test the IEEE 802.3ad standard for aggregating multiple ports to appear as one virtual pipe.

During throughput tests, the switches kept dropping members of the link-aggregation group, and all tests failed from that point forward until we power-cycled the switches. WideBand reproduced the issue and corrected a flow-control issue in its firmware, but we were not able to verify the fix by press time.

We've noted performance problems with link aggregation before, often because of poor hashing algorithms in switches. That's one issue the WideBand switch doesn't have: Commendably, it offers users numerous hashing methods. While we weren't able to complete link-aggregation testing because of the flow-control issue, we still appreciated the choice of hashing methods in the switch.

Deciding between a WideBand switch and the better-known competition is like the choice car buyers face when looking at U.S. and Japanese models. The imports offer plenty of creature comforts, usually at a price premium. Think Ford or Chevy when it comes to the WideBand switch: It's a decent performer with limited features, and it's less expensive than the imports.

Newman is president of Network Test, an independent engineering services firm in Westlake Village, Calif. He can be reached at dnewman@networktest.com.

NW Lab Alliance

Newman is also a member of the Network World Lab Alliance, a cooperative of the premier reviewers in the network industry, each bringing to bear years of practical experience on every review. For more Lab Alliance information, including what it takes to become a member, go to www.networkworld.com/alliance.

Learn more about this topic

Check how Wideband compares in our Access Switch Buyer's Guide

SMC, Asante pump out low-cost Gigabit Ethernet gear

08/01/06

Dell'Oro: Lower end switches drive Ethernet growth

02/16/06

Foundry launches new Gigabit Ethernet switches

01/13/06

Performance problems with link aggregation

4/16/01

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Now read: Getting grounded in IoT