Owners of enterprise networks expect their hardware and software to put in long, hard hours, and – in a perfect world – never go down.
But users and vendors alike know equipment will fail. The goal is to create boxes that will be up under all but the rarest circumstances; thus, the promises of five-nines availability and comfortingly long MTBF numbers. Networking equipment is also subject to regulations set by standards authorities (CSA and Underwriters Laboratories, for example) and communications bodies like the FCC and DOC. Bottom line is, your switches and routers have to go through a battery of tests to make sure they’ll live up to your expectations.
“Battery” is an apt word for what sample equipment goes through at Hewlett-Packard’s Hardware Test Centre in Roseville, Calif., where HP’s ProCurve networking equipment is dropped, shaken, sweated and pounded with packets to see if the designs will live up to spec. Though it’s owned by HP, the lab is described as “an independent, fully accredited facility” by the company.
Recently, HP invited Network World Canada for a behind-the-scenes look at the Roseville facility – the Shake ‘n’ Bake lab, as it’s known within the company. The Dynamics Lab is where the shaking takes place. “It’s one half of the shake-and-bake equation,” says Ken Klingler, senior regulatory engineer with ProCurve. The lab simulates what a piece of equipment might go through in transport and end-use environment.
Equipment is strapped to a shock test machine, fitted with accelerometers, and dropped with about 30 Gs of force. It’s used on equipment early in the development process. “We’re trying to break the product,” says Klingler. “This is a good R&D tool before we get to the finished product.” Nitrogen-charged cylinders at the base shape the shock into half-sine or trapezoidal pulses.
Vibration tables simulate transport and end-use environment, shaking products at speeds between five and 500 cycles per second, with up to five centimetres of displacement. A 10-minute operational test is followed by an hour-long swept-sine test to identify the four top resonant frequencies. Then, the product is powered up again.
Testing might shake a heat sink loose, requiring a re-engineering of the mount, or cause a machine to sporadically reset for no apparent reason, calling for engineers to root-source the problem. One single-unit rackmount piece of equipment developed what an engineer called “a smiley face” under vibration testing – it bowed. The solution was to mount the unit on a rack shelf rather than the standard L-brackets.
Telco-class equipment undergoes testing on a vibration table that simulates an earthquake – very low frequencies (less than five cycles per second) with displacement of up to 30 centimetres.
Equipment is also drop-tested in its packaging, on all of its faces, to ensure it survives cosmetically and functionally intact. Much of the information gathered in the shake portion of the program is used to design packaging. You don’t want to over-engineer the packaging, says Klingler: “It takes up more space on the pallet, and it’s more stuff to go into the landfill.” Products also undergo live testing in a Highly Accelerated Stress Testing chamber, which combines vibration with wild fluctuations in heat and humidity.
Products are tested for RF emissions in a US$3-million 10-metre semi-anechoic chamber. HP’s goal is to beat the regulated limit of RF emissions by six decibels, since manufacturing variances will inevitably move the numbers around. The chamber is a vertigo-inducing cube lined with round-cornered, square Styrofoam panels on the four walls and ceiling. Behind the squares and foam damping material, the chamber is lined with 90,000 ferrite tiles.
These combine to isolate the chamber from ambient RF interference from radio stations and other transmissions. Two antennae, one oriented vertically and one horizontally, are aimed at the product. Each is routed through a series of amplifiers so powerful that cell phones and other electronic devices aren’t allowed in the chamber in case they fry an amplifier.
An ESD lab simulates electrostatic discharge events. The test product is passing packets while being tested against static discharges. It’s an immunity standard as opposed to an emission standard, Klingler says, and it’s not regulated in the U.S. or Canada, though it is in the U.K. , where 8 KV air discharge and 4 KV contact discharge are the standards. There can’t be any “user-perceptible events”; dropped packets is one type of failure, “catastrophic” is the other. HP tests for ESD up to 25 KV. “If one of our products won’t make 25 KV, we might back off a little,” Klingler admits, especially with consumer products with a lot of exposed ports.
The hardware shakeout puts product into a simulated enterprise environment connected in various configurations to some of the roughly 600 Linux-powered PCs, each capable of simulating as many as 20 client streams. 078943