Intro to Parity and Memory Checking
Let's talk about Memory Errors! This video is meant to ease into our next topic which is explaining how parity works and how error correcting code works. Thanks for watching =)
My notes are as follows;
Parity computing definition is as follows: a function whose being even (or odd) provides a check on a set of binary values.
This video relates to RAM and how it keeps, communicates and transfers data in a reliable fashion. The idea of parity which is a highlight of today's video can be applied to storage and communications protocols but for now we will look at in the context of RAM and reliability.
RAM can have a hard error which means something broke and it won’t work again. BREAK SOME RAM Then there are soft errors. This is caused during runtime. The system can keep working although with faulty data about it doesn’t mean it should.
Memory errors occur due to a multitude of things; static, timing issues, excess heat, glitches in controllers, speed of the ram not to be confused with timings and radio waves from wifi/mouse etc.
In the early days, think 70s, RAM was badly sourced and was found to have errors caused by alpha particles, which are released by radioactive materials. It was a short lived problem and was very quickly fixed. Thorium/Uranium. 70’s Zeigler. RAM issues still happen. Although instead of alpha particles we have cosmic rays. These are high energy particles given off by the sun everytime it has a hiccup. They are able to pass through most materials such as your house and metal case around your computer. What these will do is take a bit and flip it. I.e. making a 1 a 0 or making a 0 a 1. These particles have become a fact of life in computing. Fun fact there was a study of putting computers under about 50 feet of rock and there was a lot less memory errors (IBM Volume 40) There is still some speculation as to why Memory errors occur in computers. There are also other things such as cross talk, bad programming, and bad quality sticks being made.
Three different solutions have been put in place. First is to not fix it all. This is great for cheap devices that we don’t put a lot of worries on. Think portables, mp3 players, mobile phones and consumer desktop computers. It is okay for these devices to crash once every 4 months or so due to a memory error. They will probably crash way more often to bad programming on the devices. The memory can also be slightly changed and there is a small chance anyone will lose a fortune either. This is called non-parity RAM and is probably what you are running in your computer right now.
The second type of memory is called parity RAM and isn’t seen much nowadays due to its inability to actually fix any errors. It is only able to detect errors and not fix them. This memory was standard on computers for the longest time, think 1970s to 1990s. These machines when they encountered a parity error would inform the user and the user could save the potentially corrupted data into another file and restart the machine.
The third type of memory is error correcting code memory. This is memory that is designed to be fault tolerant and correcting. The memory is able to detect errors on the fly and fix them as they happen.