Embedded diagnostics bubbling to the top

Q & A with Alan Sguigna
VP Sales and Marketing
ASSET InterTech

Q: Alan, you and the global ASSET sales forces are out there talking to test engineers all the time. What’s top-of-mind in the T&M industry these days?

Alan: Something that we’ve known about for quite some time, but which now seems to be coming to the forefront is what I call embedded diagnostics. Of course, many of the capabilities and benefits of embedded diagnostics have been known for years in telecom and high-end server applications where high-availability, remote test and diagnostics, and those kinds of issues are very critical. That’s why a company like Cray is embedding ScanWorks® IP into its next-generation supercomputers. (For more on this, click here.)

What’s changing now is we’re seeing these issues trickle down and take on greater importance for the enterprise computing and mid-range server markets as well. Much of this market space is based on Intel® x86 processors. Quite frankly, lots of people in this space are realizing it’s time to move beyond the PC mentality of: “when in doubt, reboot.” Or, the corollary to that: “When it breaks, throw it away.” So, they’re trying to incorporate some of the aspects of ScanWorks’ remote diagnostics into their systems and that’s what I’m calling embedded diagnostics.

Q: Are there specific functions that are particularly critical for embedded diagnostics?

Alan: The sky’s the limit when they embed ScanWorks, but we’re hearing that error detection and isolation, as well as recovery are imperatives and after that there is lots of other functionality that system manufacturers want to get out of embedded diagnostics. For example, the ‘no trouble found’ or NTF problem keeps coming up. Are you familiar with that?

Q: Not really. What does that involve?

Alan: In a nutshell, although the NTF problem can be quite involved, it refers to a situation where a circuit board coming off the manufacturing line or already installed in the field exhibits performance deficiencies or faults and it’s sent to a lab or repair depot for diagnostic analysis. Unfortunately, the trouble can’t be replicated in the lab and the unit under test is labeled ‘no trouble found.’ In some cases, the board is returned to service and the trouble escalates from there.

Q: How so?

Alan: Well, no manufacturer wants to develop a reputation for unreliable systems. But, in many instances, a NTF circuit board can be tested and diagnosed, and then returned to service. Later it can exhibit the same problems all over again. The cycle, complete with dissatisfied users, can be repeated several times until the unit is scrapped. The cost to the manufacturer in warranty returns and to their customer in terms of wasted effort and system downtime is huge. At the core of the NTF problem is the need for more thorough and comprehensive embedded diagnostics – either in the lab or in the field or both – that will uncover and diagnose the root cause of the problem so it can be eliminated in the unit under test itself and, more importantly, in all of the other circuit boards coming off the manufacturing line.

Q: So how would embedded diagnostics help?

Alan: Let me give you a hypothetical example. A typical high-availability x86-based mid-range server system has a service processor that monitors a heartbeat from the main x86 processor in the system. If the heartbeat from the CPU skips a beat or several beats, the service processor assumes that the system is hung up or it’s hit a fault of some sort like a double-bit memory error. Currently, the only solution is for the service processor to initiate a reboot of the compute node – not a very elegant or effective solution for high-availability systems. And the user or manufacturer learns nothing about the nature of the failure. With embedded diagnostics like our ScanWorks embedded intellectual property (IP), the service processor could take control of the CPU and do a kernel dump. With this information, technicians gain insight into the source of the fault and fix it. Of course, there are several steps in between and lots of refinement that can be done, but generally speaking, that’s how it works.

Q: Is embedded diagnostics a fairly new concept?

Alan: The concept is not really new. As I said, it’s been done at the high-end for some time. What is new is the fact that manufacturers are looking to outsource their embedded diagnostic IP to ASSET because when they outsource the IP they can focus on functionality that gives their systems competitive advantage in the market. Embedding ScanWorks IP for diagnostics means that a system manufacturer doesn’t have to develop and maintain and constantly upgrade its diagnostic capabilities. We take care of that for them. The other new aspect of embedded diagnostics is the increased interest we’re seeing in high-availability computing and mid-range server applications, as I mentioned earlier. Intel’s Intelligent Platform Management Interface (IPMI) generally provides only in-band access, and some out-of-band control comes via vPro/AMT on x86-based systems, but this isn’t available on all of Intel’s platforms.

Q: Are there other trends in the industry that you’ve observed?

Alan: Yes, it’s been somewhat surprising to see how quickly manufacturers have deployed Intel®’s new QuickPath Interconnect (QPI) architecture for high-speed input/output (I/O). We’ve seen QPI designs from companies like Xilinx, Convey Computing, Nallatech, XtremeData and others. It’s a fantastic bus and I anticipate more growth of this in the future.

Q: You’ve talked a lot about embedded diagnostics, but ASSET is all about embedded instrumentation. What’s going on there?

Alan: There’s a lot going on. Too much to talk about in detail, but one thing that’s particularly exciting for the future is the decision by chip maker PLX Technologies to implement our IEEE P1687 synthesis tools for the instrumentation they are embedding on their chips. Of course, we’re excited that PLX has chosen to implement P1687, but we’re even more excited about the implications of the decision. This means that PLX sees the value in conforming to the new P1687 IJTAG standard. We believe this is part of a growing trend and that other chip makers and independent IP developers will conform their embedded instruments to P1687 and that will facilitate a thriving tools market for embedded instrumentation.

Q: Thanks, Alan for your insights. Are there any other thoughts you’d like to leave us with?

I’d just remind Connect’s readers to check out the ASSET web site. We’ve added some really informative features, like a video that shows how non-intrusive embedded instrumentation test technologies can achieve comprehensive test coverage on an Intel® GreenCity board design featuring the Xeon® Processor 5500 Series. Just go to the home page and look in the “Latest News” section to find this video. And the “Customer Focus” feature on the home page has a link to a slide presentation that EMC gave at the recent Board Test Workshop. It’s very interesting because it shows how EMC is turning to non-intrusive embedded instrumentation test technologies for board test because they are getting better results – better coverage, faster test development, faster test times – than they can get from legacy intrusive test technologies like in-circuit test (ICT).

Q: Well, thanks for your responses, Alan. You’ve been very informative.