In a 4 300 GB disk raid 5 setup, the usable space is 900GB. So in the event that one of the hard disk fail, is it that the space available for write will drop to 500GB? It is right to said that in order to maintain write safe for my storage server in Production environment before fail disk recovering, I shall have buffer of > 400GB, ~45-50% of free space available?
The usable space will remain the same, but you won't have any redundancy left after the failure so if a disk fails during the rebuild or even before you replace the faulty disk, all your data will be lost. With a RAID6, you would only have 600GB of usable space, but two disks could fail before you loose data.
See the wikipedia entry on RAID, it explains this quite nice.
Don't think about usable space, think about immeidately stopping use of the raid array if you experience a disk failure. The reason is that you are very vulnerable to a second failure during reconstruction, which would destroy all your data.
There are other raid configurations like raid-6 that someone reduce this risk. However, the fact remains that a raid array with a dead drive is in a very precarious position. You need to do whatever you can to stop writes and get the rebuild done as fast as possible.
Active writes will also slow down the rebuild, causing it to take even longer.
Don't think about write safety - think about immediate recovery plans and stopping all writes until you have recovered.