CST 334 - Week 7

This week I learned about Persistence, with a focus on hard drives and IO devices. The first half of the lessons taught how to calculate variables like rotational delays and calculating the time it takes to read or write data. Running the calculations show that sequential reading of data performs much better than random reads and is the preferred strategy. Disk schedulers decide the order that disk requests are processed. I also learned about how the main goals for persistent storage is to be high performing, reliable, protected, and follow a naming scheme where it makes sense. At the lowest level, the name of a file is called the "inode number" and for now, we are to assume that every file has an inode number associated with it. While directories are also associated with an inode number, the contents within it are more specific, such as the user-readable name that it would have to map to the inode number.

I also finished researching with my team about how IO scheduling might be with today's ULL technology. We learned that despite its necessity in the early day of operating systems, with modern technology, it actually hinders performance, efficiency, and even worsens latency. The authors of the article prepared to prove this point through the use of macro and microbenchmarks for different read/writes while using several test equipment to collect the data. The test conducted by the authors showed that the use of IO schedulers on most of today's technology even consumes more energy than the tests conducted with no IO scheduling. It really shows that we must be flexible and constantly adjust as technology continues to grow and change.