How to modify a SOAP request?

For Linux, once the I/O requests are submitted to the block I/O layer, will the I/O scheduler be able to tell whether an I/O request comes from a sequential or random read?

• As a follow-up question to . When reading about CFQ I/O scheduler, I found that CFQ I/O scheduler maintains several different service trees and queues for I/O requests. I wonder how I/O scheduler can differentiate the I/O requests and put them into different queue. It seems that the I/O request can be marked with http://lxr.free-electrons.com/ident?i=REQ_NOIDLE from reading cfq-iosched.txt. struct http://lxr.free-electrons.com/ident?i=request has a field called cmd_flags, it is compared with http://lxr.free-electrons.com/ident?i=REQ_NOIDLE in http://lxr.free-electrons.com/ident?i=cfq_update_idle_window. From http://lwn.net/Articles/201682/ The flags field of struct request has been split into two new fields: cmd_type and cmd_flags. The former contains a value describing the type of request (filesystem request, sense, power management, etc.) while the latter has the flags which modify the way the command works (read/write, barriers, etc.).

• Answer:

  Quoted from Kernel Korner - I/O Schedulers By http://www.linuxjournal.com/user/800995 http://www.linuxjournal.com/article/6931?page=0%2C0 This magic is accomplished through two main actions, sorting and merging. First, the I/O scheduler keeps the list of pending I/O requests sorted by block number. When a new I/O request is issued, it is inserted, block-wise, into the list of pending requests. This prevents the drive head from seeking all around the disk to service I/O requests. Instead, by keeping the list sorted, the disk head moves in a straight line around the disk. If the hard drive is busy servicing a request at one part of the disk, and a new request comes in to the same part of the disk, that request can be serviced before moving off to other parts of the disk. Merging occurs when an I/O request is issued to an identical or adjacent region of the disk. Instead of issuing the new request on its own, it is merged into the identical or adjacent request. This minimizes the number of outstanding requests. Let's look at an example. Consider the case where two applications issue requests to the following block numbers, such that they arrived in the kernel in this order: 10, 500, 12, 502, 14, 504 and 12. The I/O scheduler-less approach would service these blocks in the given order. That is seven long seeks, back and forth between two parts of the disk. What a waste! If the kernel sorted and merged these requests, however, and serviced them in that order, the results would be much different: 10, 12, 14, 500, 502 and 504. Only a single far-off seek and one less request overall.