Re: iostat - define Kilobits per transfer

2003-01-22 Thread Noah Garrett Wallach

  I currently have an IDE drive that has the capacity to do 128KB/t and
  a SCSI drive 64KB/t.  Are these stats in fact showing me that there
  is a limitation with the SCSI drive?  Are my file transfering
  capaibilities less with the SCSI drive?  I suppose what do I need to
  look for in the spcifications when choosing new drives so this does
  not happen again?

 FreeBSD's SCSI layer has a cap of 64k per transaction (apparently
 because ancient ISA adapters could not do more than 64k), and the ATA
 layer has a cap of 128k.  You won't see a difference using regular
 disks.  A 20MB/sec transfer rate comes out to ~300 64K
 transactions/sec, which most systems should be able to handle with no
 problems.


so what exactly does KB per transaction mean? what happens if I am
handling 300 concurrent users with 160Kbit encoded audio streams - could I
in fact do this on this machine?  or will I be limited by the 64KB/t
issue?

- Noah




 --
   Dan Nelson
   [EMAIL PROTECTED]


To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-questions in the body of the message



Re: iostat - define Kilobits per transfer

2003-01-22 Thread Dan Nelson
In the last episode (Jan 22), Noah Garrett Wallach said:
 can somebody give me a better understanding of what the iostat output
 is decribing in the KB/t column.  It might be really simple but
 figured I';d ask, if in fact further clarification can be given.
 
 typhoon% iostat 1
   tty da0  da1 acd0
  tin tout  KB/t tps  MB/s   KB/t tps  MB/s   KB/t tps  MB/s  us ni sy in
0   21  0.00   0  0.00   0.00   0  0.00   0.00   0  0.00   1  0  1  0
 
 do drive specifications generally contain KB/t information or statistics?

When there is disk activity, yes.  Try running a couple du's or extract
a couple ports, then run iostat in another window.
 
 I currently have an IDE drive that has the capacity to do 128KB/t and
 a SCSI drive 64KB/t.  Are these stats in fact showing me that there
 is a limitation with the SCSI drive?  Are my file transfering
 capaibilities less with the SCSI drive?  I suppose what do I need to
 look for in the spcifications when choosing new drives so this does
 not happen again?

FreeBSD's SCSI layer has a cap of 64k per transaction (apparently
because ancient ISA adapters could not do more than 64k), and the ATA
layer has a cap of 128k.  You won't see a difference using regular
disks.  A 20MB/sec transfer rate comes out to ~300 64K
transactions/sec, which most systems should be able to handle with no
problems.

-- 
Dan Nelson
[EMAIL PROTECTED]

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-questions in the body of the message



Re: iostat - define Kilobits per transfer

2003-01-22 Thread Dax Eckenberg
   I currently have an IDE drive that has the capacity to do 128KB/t and
   a SCSI drive 64KB/t.  Are these stats in fact showing me that there
   is a limitation with the SCSI drive?  Are my file transfering
   capaibilities less with the SCSI drive?  I suppose what do I need to
   look for in the spcifications when choosing new drives so this does
   not happen again?
 
  FreeBSD's SCSI layer has a cap of 64k per transaction (apparently
  because ancient ISA adapters could not do more than 64k), and the ATA
  layer has a cap of 128k.  You won't see a difference using regular
  disks.  A 20MB/sec transfer rate comes out to ~300 64K
  transactions/sec, which most systems should be able to handle with no
  problems.
 
 
 so what exactly does KB per transaction mean? what happens if I am
 handling 300 concurrent users with 160Kbit encoded audio streams - could I
 in fact do this on this machine?  or will I be limited by the 64KB/t
 issue?
 

300 x 160Kbit = approx. 46Mbit/sec.  
A new-ish SCSI drive should be able to easily pump out in excess of 200 Mbit/sec.
Your bottleneck will be your ethernet adapter long before your local storage.
Unless your app is designed very poorly.

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-questions in the body of the message



Re: iostat - define Kilobits per transfer

2003-01-22 Thread Noah Garrett Wallach
On Wed, 22 Jan 2003, Dax Eckenberg wrote:

 
 
  so what exactly does KB per transaction mean? what happens if I am
  handling 300 concurrent users with 160Kbit encoded audio streams - could I
  in fact do this on this machine?  or will I be limited by the 64KB/t
  issue?
 

 300 x 160Kbit = approx. 46Mbit/sec.
 A new-ish SCSI drive should be able to easily pump out in excess of 200 Mbit/sec.
 Your bottleneck will be your ethernet adapter long before your local storage.
 Unless your app is designed very poorly.



okay things are getting clearer over here.  what exactly does KB
per transaction mean?  I dont understand what this describes?

- Noah

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-questions in the body of the message



Re: iostat - define Kilobits per transfer

2003-01-22 Thread Mark Rowlands
On Thursday 23 January 2003 5:40 am, Noah Garrett Wallach wrote:
 On Wed, 22 Jan 2003, Dax Eckenberg wrote:
   so what exactly does KB per transaction mean? what happens if I am
   handling 300 concurrent users with 160Kbit encoded audio streams -
   could I in fact do this on this machine?  or will I be limited by the
   64KB/t issue?
 
  300 x 160Kbit = approx. 46Mbit/sec.
  A new-ish SCSI drive should be able to easily pump out in excess of 200
  Mbit/sec. Your bottleneck will be your ethernet adapter long before your
  local storage. Unless your app is designed very poorly.

 okay things are getting clearer over here.  what exactly does KB
 per transaction mean?  I dont understand what this describes?

 - Noah

 To Unsubscribe: send mail to [EMAIL PROTECTED]
 with unsubscribe freebsd-questions in the body of the message

my guess :- 

iostat -Iw 60 -t da

 ad0
KB/tXfrs MB
8.42710.58

so over the period,  71 transfers occurred totalling 0.58MB for an average 
KB/t of (0.58*1024) / /71   = 8.37 KB/t   but maybe the actual avg transfer 
size is recorded and summarized giving that slight variation?.  Try a longer 
period and see, or read the code..   (the math starts around line 
600) 

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-questions in the body of the message