On Thu, Jun 12, 2003 at 03:19:35PM -0400, Michael Packer wrote:
> 
> I've got amanda running on one machine and it seems to be running fine.
> 
> I've got another server (Dell 2500) Dual 1.1g 512k cache.
> 
> Raid card (Adaptec AHA274x/284x/294x)
> 
> DLT Drive BNCHMARK Model DLT1
> 
> I've got about 28Gig used.  It was taking about 4.5 hours before it started
> to write to tape and about 3.5 hours to write to tape.  I'm trying to dump
> ArcServeIT by CA.  ArcserveIT can write all of the data in less than 2
> hours, but I'd rather use amanda (like I am on the other servers) ... if i

At maximum rate, what 5MB/sec (18GB/hr) 2 hrs would be about right if nothing
else were coming into play.

> can get it to a "better" time usage and without overloading the machine.  It
> is currently running the load up to around 4.   I'd appreciate any ideas on
> settings I may have incorrect.

How many DLE's?
Do you have those on the same drive set to the same spindle number?

> Here are some of my conf settings:
> 
> inparallel 2
> dumporder  "sssS"
> netusage 1500Kbps
> dumpcycle 1 day
> runspercycle 5
> tapecycle 1 tapes

HUH?  five times per day and only one tape?

> bumpsize 20Mb
> bumpdays 1
> bumpmult 4
> etimeout 300
> dtimeout 1800
> ctimeout 40
> tapebufs 20
> runtapes 1
> tapedev "/dev/nst0"
> tapetype DLT
> 
> holding disk is "use 500Mb, chunksize 1GB"

Kinda tough to create "chunks" of 1GB in a 0.5GB space.
That means you are sending directly to tape and there is
no parallelism going on.  I.e.  one DLE dumping to
holding disk while another is being transfered at full
speed from the holding disk to the tape.  Going directly
to tape may also be greatly degrading the speed of the
tape writing as every time the collection of dump data
from the disk falls behind the tape drive stops streaming,
backs up, waits for data, starts up, and resumes writing.

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)