Advocacy
Dojo (HowTo)
Reference
Markets
Museum
News
Other
Myths
Press
General
Hack
Hardware
Interface
Software
Standards
People
Forensics
Web
CodeNames
Easter Eggs
History
Innovation
Sightings
Opinion
Martial Arts
ITIL
Thought
|
Why HD performance is so critical
Computer systems are starting to lean so heavily on hard drives for their tasks that hard drives performance is now becoming critical to measuring a systems performance. When systems don't have enough RAM, they use virtual [pretend] memory (RAM) - by saving a block of Memory to the hard drive, which frees up that memory for other things. Later when the user needs to access that first block of virtual RAM - the system quickly saves out some other RAM to the HD, and reads in the first block - making it appear to the system (and user) that the machine has more RAM than it really does. It works pretty well but is slow (compared to real RAM). The more real RAM you have, the less virtual RAM you will need, and the faster your system will perform (in those conditions).
It is this Virtual Memory that makes systems much more dependant on Hard Drive performance than ever before, because they are constantly reading and writing to the hard drive. Windows is more dependant on virtual memory (than the Mac) and so is more dependant on hard drive performance(1).
(1) To be fair, the MacOS has a worse VM implementation than Windows -- but the MacOS needs VM less because of better space efficiency, larger ROM, and more RAM expandability. Apple is expected to fix their VM deficiencies in late '97 or early '98. Rhapsody (Mach/Unix) exceeds WindowsVM performance.
Understanding HD performance
There are a few aspects of hard drive performance. Generally bigger drives are faster because they have more data crammed into the same space. So if all the mechanisms move at the same speed, then they are able to access more data in the same amount of time if that data is more closely packed. So bigger is often better. Furthermore, the larger the drives then usually the more cutting edge technology is used (to get that data into the same space), so companies often use better components and the drives are often faster.
Think of a hard drive as a spinning vinyl record album (for those of you that remember them) but with concentric rings instead of one spiral track, and you have to move the needle to the correct ring (called track) constantly. From this context I will explain the different aspects of drive speed -
- RPM's - how fast the platter is spinning, somewhere between 4500 and 7600 RPM's (9,600 and 10,000 RPM drives are starting to be shown).
- Seek time (access time) - how much time it takes to move the needle (actuator arm) from one track to the next. This is often measured in "average" seek, which is how long it would take to go 1/2 way across the whole platter. Some people only measure seek as from one track to the next - but that is an ideal condition.
- Latency - once the actuator arm (needle) is transfered to the right track, you must wait for the platter to spin to the exact part of the data that you want to read (hear). This time delay is what is usually referred to when talking about drive latency. The faster the drives RPM's - the less this latency.
- Transfer rate - once the drive gets to the data you want, how fast can it send the data to the computer. This is related to how fast the connection is (and depends on the type), and how fast the drive is spinning relative to how much data the drive can hold. So the faster the connection and faster the drive is spinning with the most data in the same amount of space - then the better transfer rate you will get.
- Cache - how much of a buffer the drive has built in to make it SEEM like it is faster than it really it. Macs also used another buffer in the system (controlled through the memory control panel) to help the drive out - this was important years ago (with slower drives with less cache) but is less important today.
Now drive performance is relative to many of those parameters, and varries depending on how you are using the computer. When you want to send large amounts of data (creating videos or doing desktop publishing or scanning) - then transfer rate (through-put) is the most important spec. If you are working with lots of little files (compiling), then the access time/latency is more important.
Because of modern OS's and VM, computers are becoming more I/O bound (to hard drive performance). This means that Intel machines are more harmed by their inferior IDE design than most Macs. The more RAM you have and the more efficiently you use it, then the less important the drive speed often becomes (in the case of VM). Since Macs have 4-8 megs of ROM (aleviating the need for that much RAM), and Macs are often more memory efficient, the effects are more dramatic than many people know.
Conclusion
Remember that HD performance is only one paramter in System performance, and that System performance is the sum of all of its parts. Since computer systems are complex, users often have a hard time predicting what a system actual performance will be until they use it and measure it (and benchmark it against other system). Most Mac users are pleasantly suprised - most PC users are not.
|
