by KaiserJeep » Thu 05 Sep 2013, 17:57:30
$this->bbcode_second_pass_quote('Outcast_Searcher', '')$this->bbcode_second_pass_quote('KaiserJeep', '
')I'm a 61 year old Electrical Engineer working on computer hardware for the largest tech company on Earth.
...
Then somebody bought that company, and a few years later, somebody bought the company that bought ours, and now I am one of over 300,000 employees.
Hi KaiserJeep. From context, I assumed you were talking about IBM.
Looking at the 10 biggest tech companies on Wiki (as of 3/2012), IBM is third in employees, roughly tied for second in market cap, and Fifth in annual revenue. For consistency in being "big", I'd give either IBM or Samsung the nod. (DIsclosure: I spent my 26+ year I/T career at IBM). I'm NOT being pedantic here -- it is no longer obvious at all to me which is the largest tech company.
...
Re your interest in fault tolerant computers, and the large chilled room (presumably mainframes) you cite, it prompts a question.
My expertise after my application programming days was system software programming, specifically DB2 on the IBM mainframe. (Fault tolerance was clearly a BIG deal to me before RAID made single point of failure mainframe hard drive head crashes a thing of the past). With the advent of networking and distributed computing, and robust applications like DB2 (parallel sysplex) that can operate on many mainframes simultaneously which (within distance limits) are on different sites -- I wonder about the relative NEED for serious fault tolerant computing (i.e. one single super-duper fault tolerant computer) any more.
Essentially, cheap redundancy and reliable communication seems to have largely filled the gap. The thing all the beasts need, which is the obvious linchpin over time, is plenty of reliable electric power.
Just wondered if I'm missing something obvious. (I'm a pure software guy -- I know a little physics and math and principles of electricity, but nothing close to formal degrees in EE, etc).
....
For the home user, having a few cheap PC's and a group of memory sticks stored in several relatively secure locations (like safe deposit boxes -- which also serve as a Faraday cage) in a ubiquitous format like PDF seems to have the data redunancy issue covered without complexity or undue expense. (Just diligence in backing up and storing the data).
Cheers, and welcome to the discussion boards.
Thank you. My original company was Tandem Computers Inc. , the remnants of which live on as the NonStop Enterprise Division of HP.
The Tandem product was unique, and the entirely proprietary operating system was based upon processor pairs that communicate over entirely redundant high speed interprocessor busses, using data packets and duplicating each piece of system hardware for redundancy. The "online transaction processing system" also utilized the dual interprocessor busses to scale the server from the minimum two CPU configuration up to 16 CPUs/server. Add the networking layer and the system scaled to up to 254 nodes of 16 CPUs. The system was 100% redundant hardware, and designed for online repair that allowed you to restore redundancy before a second failure occurred.
Tandem grew rapidly until every stock exchange, most banks, most credit card companies, and miscellaneous applications like cellphone billing were saturated by Tandem products. It is still the case that NonStop computers are still the most expensive to acquire but offer the most security possible (0.999994 availability) and the lowest cost per transaction - even when measured against PCs and mobile devices. Our equipment fulfills the "database of record" function in a remote chilled room often hundreds of miles from the users.
Just the upgrade business is $6B to $8B per year, and NonStop still enjoys the highest profit margin of any part of HP. But we have literally saturated our OLTP market, and growth has slowed. The NonStop OS has been ported onto HP Blade Servers and we no longer manufacture unique hardware. However I still sustain our older products, some of which have been in service for over two decades.
I never really left Tandem Computers, it left me, through the two acquisitions I described. Now at HP it seems like full circle, the original Tandem NonStop I computer from the days of ferrite core memory in the late 1970's, was very similar to the HP 3000, plus dual redundant hardware busses to allow the message based operating system to function.
