http://qs321.pair.com?node_id=381239


in reply to [OT] End of hard disks era, new ways of development

One of the things that make relational databases so efficient is that they are optimized to run on a spinning magnetic disk (like a hard drive). They take time to know just where the head is located physically over the platter and move it in the most efficient manner. This isn't to say that they won't run faster on a solid-state memory device, just that you throw out a lot of hard work if you do.

I don't think this is a reason not to move to a solid-state storage device, just something to keep in mind. If you can justify getting a TB memory array for your application, go ahead.

I don't know if there is any real optimization you can do for solid-state devices, or if those optimizations will even make for a significant speed advantage.

"There is no shame in being self-taught, only in not trying to learn in the first place." -- Atrus, Myst: The Book of D'ni.

  • Comment on Re: [OT] End of hard disks era, new ways of development

Replies are listed 'Best First'.
Re^2: [OT] End of hard disks era, new ways of development
by beable (Friar) on Aug 10, 2004 at 05:52 UTC
    I don't know if there is any real optimization you can do for solid-state devices, or if those optimizations will even make for a significant speed advantage.

    Did you ever study data structures? Have you compared the speed of doing a linear search with a binary tree search? There are optimisations you can do for solid-state devices, it's just like RAM. That's why people spend time trying to work out what the best data structure is. Otherwise, we'd just dump everything in a list and grep for the results we want. As for hardware-specific optimisations, once there is enough hardware around, if people find that it is going too slow, somebody will find a way to do it faster.