Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I understand that performance degrades with single disk volumes and small RAID volumes as the volumes fill up. Some admins say that they perceive slowdowns at 80% and use that as a general rule of thumb to increase available storage. Others say 90%. I'll also accept that different file types present different experiences, too — say, video files, version small database files, to general Office type documents, partition table and filesystem format. There are, I'm sure, other contributing factors that affect performance, like RAID vs SAN vs virtualized storage, RAID block size, NFS vs CIFS and other aspects.

My question is: in large, direct-attached, non-SAN, RAID 60 HFS+J volumes, of, say, 50TB, attached via fibre channel and made available to an office 100 concurrent users using typical network filesharing protocols -- does that same rule of thumb (80% or 90% used) hold true, or can administrators expect performance degradation to occur much later, at, say, 95% volume usage? In other words, if my 50TB RAID volume is at 80% now, at what point am I pressed to increase storage? When should I be alarmed?,

share|improve this question
1  
Can you provide citations for these rules of thumb you've heard? I'm dubious. –  Mark Wagner Apr 21 '11 at 22:10
    
Well, as "rules of thumb" it's more folk wisdom, peer anecdotes and personal observation than something I can cite in the Big Book of System Administration. Probably because there are so many variables in play, and it's tough to have hard and fast rules. My point in asking is that I lack that direct experience with large filesystems like this. –  flumignan Apr 21 '11 at 23:05
add comment

1 Answer

up vote 1 down vote accepted

I guess that degradation you mention is caused by the filesystem and not the storage device. In that case, any block device (single disks, RAIDs, whatever) would behave similarly.

The only exception that comes to my mind is SDDs; on one hand, the lack of significant seek times would make irrelevant the level of fragmentation and non-locality that plague mostly-full filesystems, and on the other hand, non-enterprise SDDs tend to have little (if any) non-accessible storage cells that are crucial for the wear-leveling and erase-cycle-hiding algorithms, so it might start to behave more and more like cheap USB flashdrives.

share|improve this answer
    
+1; The file system affects this the most, and most modern file systems do not experience adverse issues until they are basically full. Older file systems the case was quite different. Good SSDs have more reserve, and the TRIM command helps hide the erase cycle substantially. Cheaper SSDs and those not using TRIM will noticeably have less performance. –  Chris S Apr 22 '11 at 3:10
    
Leaving aside the peculiarities of SSDs, is it fair to say that the rate of performance degradation on a single HDD volume tracks closely as a single large RAID volume, provided they have the same filesystem? –  flumignan Apr 22 '11 at 21:19
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.