Volatile Yard 挥洒自如 Write anything I want to write...

I roughly timed the Captive NTFS operation a few days ago. Since there were a lot of noises in the experiment, I did a similar measurement again last weekend. I have changed the setup a bit to eliminate some noises, although some were still there.

This time, instead of a harddisk partition, I used a RAM disk as the source. Copying from a RAM disk to a harddisk partition makes the total distance travelled by the harddisk's read-write arm more consistent. There are 512MB ram on my computer. To avoid using the swap memory, I use a RAM disk of only 160MB (formatted with ext3) and measure the copying of only 3 files that occupy 150,000,000 bytes. I repeated the experiment for a few times and selected the minimum time for each reading. The table below summarizes the elapsed time required for each operation:

By calculating the transfer rates and compare them with my previous results (with slightly different experimental setup), you will find that the transfer rates for copying to ext3 almost double here. The reason being that reading an ext3 partition from a RAM disk is obviously faster than reading it from a harddisk. From the figures, copying with the Captive NTFS is about 60 times slower than that with the native ext3 module.

Again, the measurement here is still rough. For example, the experiment didn't account for the fragmentation issues which can affect the writing rates. Nonetheless, such factors shouldn't affect the measurement much. All in all, I'm still ok with the Captive NTFS as it works well for me and I hardly need to access to the NTFS partitions.