Path: utzoo!attcan!uunet!husc6!uwvax!umn-d-ub!nic.MR.NET!shamash!nis!ems!bungia!orbit!pnet51!shawn From: shawn@pnet51.cts.com (Shawn Stanley) Newsgroups: comp.sys.apple Subject: Re: GS/OS Message-ID: <158@orbit.UUCP> Date: 7 Oct 88 15:05:07 GMT Sender: root@orbit.UUCP Organization: People-Net [pnet51], Minneapolis, MN. Lines: 34 shankar@src.honeywell.COM (Subash Shankar) writes: >Does anybody know what the benefits of the disk caching NDA under GS/OS? >I have tried several different values for the cache size, but all of them >(including 0K) seem to have the same performance when I tested their speeds >at disk copying. The same disk copying, when run with the Applied Engineering >GS/RAM cache, runs significantly faster. So what is the NDA cache used for? As you know, disk caching involves keeping the blocks recently read in memory for later use, and juggling them such that old blocks are replaced by new blocks. [In some cases, old means the "usage" counter is lower and it hasn't been in the buffer for a certain number of accesses.] When you access files and directories, the file and directory sectors are placed in the cache. Any re-reads of these blocks will come from the cache and not from the disk. Any writes modify the cache copy and, in turn, are immediately output to the disk. This is known as a write-through cache. It preserves disk integrity and allows disk copying without removing the cache. When you copy a disk, you access blocks that are probably not in the cache. You start at block zero and work your way up to block 1599 (with an 800k disk). Previous cached blocks are replaced by the huge number of newly read blocks, thus the chances that a disk copy will access normal file and directory blocks before they're "aged" out of the buffer is rather low. When you write to the target disk, there is no time saved. The write HAS to occur, whether the cache is invoked or not. The cache is designed to speed up reading blocks that are accessed often. If you had your cache set to }isay 800k and then accessed your drive a lot, and then copied the current disk in the drive, I would guess there's a good chance it might be a bit faster when reading from the source disk. UUCP: {rosevax, crash}!orbit!pnet51!shawn INET: shawn@pnet51.cts.com