home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.periphs.scsi
- Path: sparky!uunet!gatech!rpi!uwm.edu!spool.mu.edu!agate!usenet.ins.cwru.edu!magnus.acs.ohio-state.edu!csn!teal.csn.org!milton
- From: milton@teal.csn.org (Milton Scritsmier)
- Subject: Re: Noise like a seek coming from new scsi hard drive.
- Message-ID: <C1G2Gt.CGu@csn.org>
- Sender: news@csn.org (news)
- Nntp-Posting-Host: teal.csn.org
- Organization: Colorado SuperNet, Inc.
- References: <1330@alsys1.aecom.yu.edu>
- Date: Tue, 26 Jan 1993 04:43:41 GMT
- Lines: 79
-
- In article <1330@alsys1.aecom.yu.edu> manaster@yu1.yu.edu (Chaim Manaster) writes:
- >
- >Most (not quite all) the bugs are out, however one thing worries me
- >somewhat. Every so often on intervals on the order of 20 to 30
- >minutes apart, (not too regular) there is a momentary sound from
- >the hard drive that lasts for about one second (by my reckoning)
- >and sound like the drive has been given a seek command. Is this
- >possible? (By the way, the above reffers to periods where the
- >computer has been switched on, but is otherwise idle during that
- >entire period, and not being used by someoneone at the keyboard.
- >The only thing happening is whatever internal loop Dos does while
- >waiting at the Dos prompt, or Windows does while wating for some
- >user input.)
- >
-
- Many people have responded to say that it is done to avoid a stiction
- problem. However, it is much more likely that the drive is going through
- its periodic thermal calibration sweep.
-
- The problem is that the density of data on the media has become so
- great that mechanical variations within the drive and the slightly
- different thermal properties of each platter causes each platter to
- expand and contract differently at different temperatures. If the
- drive uses a dedicated servo platter (at the top of the stack, say),
- the mechanical and thermal differences between the platters actually
- causes misalignments among the heads that are big enough with the
- greater crowding of tracks to make it impossible to seek to a track
- reliably on one platter even when the servo head is right on the
- money. This was not a problem on earlier drives where the tracks
- were wider and the platters not so closely spaced (newer drives
- are much more likely to develop hot spots at various places
- inside the drive, and have less outer cooling surface because
- they're typically smaller).
-
- The trick drive manufacturers use is to do a calibration sweep at
- spin-up time to determine how far each head is from where the drive
- thinks it should be on a given track. This offset for each track is
- stored and any time the drive needs to seek to a track on the platter
- a kind of two stage process is used. There is the first major seek
- which gets the head to about the right location, and then there is a
- minor seek where the previously stored offset (plus some extra servo
- sampling on the track) is used to find the precise center of the track.
- Note that even when a sequential operation is being done and the drive
- only needs to switch heads, the second seek must still be done (your're
- on a new platter with a different offset) even though the first major
- seek can be eliminated. That's why head switch times on newer drives
- are measured in milliseconds now instead of the microseconds they
- used to be measured in. A head switch now requires a small mini-seek.
-
- This is the one performance area where drives have actually gotten
- worse over the last few years. Often a drive manufacturer will put
- things like rotational speed and seek times right up front in their
- drive manual where you can't miss it, but then bury or not even
- report the head switch time. This greater head switch time (which
- can be as much as a quarter or third of a revolution) really does
- have an impact on performance, even when the drive is properly
- formatted with the correct track skew. There is simply no way to
- avoid the delay caused by the head switch mini-seek, unless you
- count read-ahead and caching schemes.
-
- I should also say here that drives that have embedded servos instead
- of a dedicated servo platter also will do a calibration sweep
- at power-on. They suffer from exactly the same thermal and mechanical
- problems and have the same media density as drives with dedicated
- servo platters. Thus these drives must also do a mini-seek during
- head switches.
-
- The reason for periodic thermal calibration sweeps to get new offset
- values is because the temperature of the drive is not constant from
- power-on to running over several hours. The drive starts out relatively
- cold, and warms up after it spins for awhile. Also, the computer it is
- in starts to warm up as well. This changes the thermal effects for
- each platter and since each platter expands differently, new offsets
- must be determined by doing a calibration sweep on each head. This is
- done quite often immediately after power-on, and is gradually reduced
- in frequency as the drive has been on for awhile (every 20 or
- 30 minutes and a second or so duration sounds about right).
-
-
-