Read vs. Write Time

Question

Which operation requires more time - reading from disk or writing to disk for the same amount of data and the same place in memory?

Answer 1

This is actually a pretty tricky question, and it requires an understanding of how your I / O system is configured. The simple example you are quoting (reading / writing a fixed amount of data to a specific location on disk) is not as realistic as you think. Here is a brief description of things that can affect I / O performance.

• Disk speed
  Hard disk speed is usually expressed in terms of rotational speed (rpm or revolutions per minute), which indicates how quickly the plates rotate inside the drive. Typical values ​​are from 5400 to 10,000 rpm. Typical transfer rates range from 1 to 6 Gb / s and can support transfer rates of up to 125 MB / s.

  Keep in mind that there is a difference between latency and throughput . If you write very small pieces of data in different places on your disk, you depend on the reception delay ( search time , rotation delay and access time). But, if you transfer a large amount of data at the same time, you are probably more dependent on bandwidth. Your file system determines how files are laid out on disk, and may try to optimize such things (see below).

  Another thing to keep in mind is that you (and most enterprises) can get faster transfer speeds using multiple drives in a RAID configuration. RAID performance depends on the combination of striping, mirroring, and parity you choose. Check out the Wikipedia article for all the intricacies. There are too many parameters to explain here.

• Caching
  Modern operating systems carefully plan when they interact with the drive. There may be several cache levels between your program and a physical disk, so the performance that you see as an application programmer may depend more on how your operating system processes the data, and not on the actual performance of your disk.

  Most operating systems today use cache memory so that data from the disk can be stored in memory, and the OS can schedule when it is talking to the disk. Application entries will look fast, as they can go directly into memory, and the OS can wait to flush the buffer while it has nothing to do. In practice, the OS will attempt to write the flash in a fairly timely manner, so that a power failure does not destroy all your data. Thus, although there is free space in the buffer, the entries will look fast. If you fill up the buffer cache, or if the operating system has little free memory to work with, you can see that I / O performance is deteriorating because the OS has to flush buffers more often.

  Read speed, as well as write speed, depends on how the data is cached. Most hard drives today have hardware caches that can increase the speed of data transfer. Similarly, the OS uses a buffer cache to store the files you accessed. Any of them can use some prefetching strategy to pre-retrieve data from disk, if it seems necessary. When caching, initial reads can be slow, but subsequent reads (especially reading the same one) will take less time if the data to be read is already in the cache.

• Filesystem
  Finally, you need to consider the file system . A large recording may not all go to the same place, so you cannot just consider your transfer rate when estimating how long it will take. Files are not always contiguous on the disk, and your file system must figure out how to arrange them, which can dramatically affect performance depending on how much free space is available and how fragmented it is on your disk.

Read / write performance is reduced to the totality of all the above effects plus the characteristics of the workload that you load onto the disk (data size, read and write frequency, etc.). As in most cases, you need to experiment with your application, the OS that you are going to run, and your specific disk configuration to get a realistic idea of ​​how it runs.

Answer 2

Short answer: it depends a lot.

From the application level, entries usually appear faster. since you are actually only requesting that the OS write data, and the OS can quickly return to you and write data at your leisure. When reading, you need to wait until the OS returns to you with the data you want.

The file system can dramatically affect the speed of reading and writing ... often a lot of work happens when writing, but if you add a file that can speed up.

Most SSDs are much slower to write than to read.

Answer 3

Buffers will affect the time for reading and writing in large numbers. Buffers can be supported by the operating system in RAM, and many drives also contain internal buffers, which are part of the disk controller.

Note that the operating system can cache part of the file in RAM, so reading from these parts can complete very quickly. In addition, the operating system can cache writes to RAM until there is a sufficient amount to write to disk. The call to the "write" function may return after copying data only to another memory area.

In short, to summarize, if you need to write bits to a disk (using a flash operation or something like that), this operation will be at least until it is read from the disk, most likely longer.