I've recently been told by a y2k technician that the software will override the Real Time Clock inside the embedded chip and that if the software is fixed, the actual chip wont be a problem. He said that many chips with RTC's will just continue on if their date function isn't utilized by the software that it relates to. I have no technical training and would like to know if this is true.

-- Anonymous, September 09, 1998

Answers

Technically, software can't "override" the RTC; however, it can compensate for the any anamolies in a given RTC. You can read more about this issue in an InfoWorld article dated Sept. 5th at

http://www.infoworld.com/cgi-bin/displayStory.pl?98095.ehrtc.htm

Hope this helps.

-- Anonymous, September 09, 1998

Typically the Real Time Clock(RTC) in embedded systems is a signal that is derived from dividing down the frequency of the main microprocessor clock. This signal will generate an interrupt to the processor at a configurable rate(usually 1 or 10 ms). It is then up to the software to accumulate these interrupts into seconds, minutes, hours, etc. in whatever format the application requires.

A large percentage of these applications operate off of incremental time and could care less about actual clock time. Many of these applications are based on 8-bit micro controllers such as the Intel 8085, Zilog Z80, Motorola 6800 and their derivitives. Newer designs may use external clock chips from companies such as Dallas Semiconductor, which perform the RTC functions external from the main system processor. These vendors should be contacted directly concerning the compliance or their RTC chips.

-- Anonymous, September 10, 1998

My answer will focus on IBM PC compatible embedded systems. 100 percent of these systems have the Y2K bug in hardware. These systems are made by the following companies: Radisys, Ziatech, Texas Micro, Octagon, Win Systems, Ampro, and many many more. Many hundreds of millions of these systems have been deployed. All of them have the potential for Y2K related problems. Even if the embedded system is using an off-the-shelf OS that is Y2K compliant and a recent BIOS that is also Y2K compliant this does not guarentee that the embedded system is Y2K compliant. I have seen it with my own eyes. The engineer wrote his own piece of code to read the RTC CMOS directly and was not aware of the Y2K bug and hence did not account for it. Fall down go boom.

My conservative estimate is that there have been at least 2 billion IBM PC compatible embedded systems installed since 1985. Of those, a minimum of 1% will have malfunctions that cause the system to fail or shutdown. That's 20 million random black boxes controlling everything from ATMs to robots to ICU medical systems to nuclear reactors.

--AJ

-- Anonymous, September 10, 1998

This is an update to revise my previous post. I have since discovered that Dallas Semiconductor has been shipping a compliant IBM PC compatible RTC for a year or two now. So the hardware bug is not in 100% of all PCs, just a large majority. I'd still put it at roughly 90% if you take into account all the "legacy" (186, 286, 386 and 486) embedded systems out there. Also, I think my "2 billion" number might be a tad high. I'll say at least one billion and stick to my 1% will fail number. So the number is now around ten million failures if not fixed.

I'd also like to point out that most of these systems will not be fixed. It's difficult even to find them all. What's more, only a small percentage will fail on January 1, 2000 (providing the power stays on). They will fail the next time they are rebooted.

--AJ

-- Anonymous, September 21, 1998

I was wondering if anyone knew what the normal failure rate is for embedded systems? I know this can't be as high as some of the estimated failure rates due to Y2K but everything has a failure rate.

Also, how are the estimates (some as high as a 7% failure rate) arrived at?

-- Anonymous, September 23, 1998

I don't have any hard numbers for general failure rates. But excluding "infant mortalities" once an embedded system is installed it can go on merrily doing its job for at least ten years. There are plenty of Intel 286 based controllers out there. All of them have the hardware RTC bug, >90% of them have the software RTC bug.

As far as I have been able to discover the 7% number has been arrived at empirically. For my own purposes of trying to get my head wrapped around the scope of the problem I always try to underestimate the problem (and it's still plenty scary).

The rule of thumb I have been using is: for any medium to large plant -- more than one or two hundred PLCs, black boxes, etc. -- the chances are that 4% of the devices will have a Y2K problem (but not necessarily service effecting), and 1% will cause a serious failure (plant shutdown).

Given the time it takes do conduct a full remediation of medium to large plant, I believe that for any such plant that is only now starting its inventory and assessment phase, it is too late.

I have been focusing my energy and research on the power utilities and the telecommunications industry. If these two utilities stay up then I think we'll be ok. We'll have a pretty severe recession but we'll get everything back together in a few years. If either of these two utilities have major failures then we're in for a wild ride.

I know I'm getting way off topic, but here are a couple of thoughts that have been troubling me of late (I'll post these both as separate threads too).

1> All indications are that the power utilities are only now really getting rolling with their Y2K programs. For the sake of argument let's assume they all finish their assessment by 1/1/1999. Whether the plant is a power generating station, energy management system or HVDC bridge they all discover that the problem isn't that bad. They need to upgrade some PC BIOSs, replace some PCs, upgrade and remediate some mainframe software, replace a few PLCs, upgrade some industry specific black boxes and replace some industry specific black boxes. My thought is this, when 7800+ plants place their modest POs for a few pieces of equipment from ABB, Honeywell, Bently-Nevada, Elsag-Bailey, et al., are these companies going to be able to deliver in a timely fashion? Will each plant be able to deploy, implement and test the newly configured system in time? [the feeling in my stomach is not a good one].

2> With the previous point as a backdrop, what are the chances that all the "fixes" are going to be correct? It is well known within the software industry that everytime you fix a bug you run the risk of introducing a new bug or bugs.

------------------------ Here are a few anecdotal cases from the IEE in the UK.

--------------------- http://www.iee.org.uk/2000risk/updates/update02.htm

1.5.6 Cases

1.5.6.1 The following is provided for the information of those who want sample case material.

1.5.6.2 In a meeting in September 1997, an Energy Industry Y2K Special Interest Group which has 19 member companies, most of which are either pursuing or starting to pursue inventory and assessment of their field systems, gave the following sample results:

(a) Four Alcoa Steel Plants

50% of control systems will fail

(b) North Sea Expro (Shell-Exxon JV)

Platform, Pipeline and Gas Plants

1200 systems identified, 12% failure rate

(c) BP Refinery, Grangemouth, UK

94 systems identified

Couldn't find vendor for 20

74 assessed

3 will fail

2 will cause shutdown

1.5.6.3 [From a Healthcare Discussion Group: At a hospital] we educated our Biomedical Department about Year 2000 and they decided to test a ventilator that had been delivered that week. When they set up the ventilator for date and time (12/31/99 11:00) and let it tick-tock down, the ventilator failed. When they called the manufacturer, they were able to instruct our technicians to power down the unit, remove the cover, and flip some internal switch, reset and power the machine up again. The ventilator worked fine at this point. Comment: obviously this vendor had at least made some preparation for year 2000. But how would we have known about the hidden switch to set, and how many other biomedical vendors haven't even heard of Year 2000 yet?

--------------------------

Regards, --AJ

-- Anonymous, September 24, 1998