[OpenIndiana-discuss] Interesting question about L2ARC

[Dan Swartzendruber]

At the moment, 20GB.  Here is the hit/miss info from later in
arc_summary.pl:

        CACHE HITS BY DATA TYPE:
          Demand Data:                45%        8649309
          Prefetch Data:               1%        232747
          Demand Metadata:            31%        5979043
          Prefetch Metadata:          22%        4286704
        CACHE MISSES BY DATA TYPE:
          Demand Data:                56%        996300
          Prefetch Data:               8%        140398
          Demand Metadata:            35%        615346
          Prefetch Metadata:           0%        1196 

-----Original Message-----
From: Sašo Kiselkov [mailto:skiselkov.ml at gmail.com] 
Sent: Tuesday, September 11, 2012 9:31 AM
To: Dan Swartzendruber
Cc: zfs at lists.illumos.org; 'Discussion list for OpenIndiana'
Subject: Re: [OpenIndiana-discuss] Interesting question about L2ARC

On 09/11/2012 03:27 PM, Dan Swartzendruber wrote:
> 
> Saso, I think you might be on to something here.  This is a handful of 
> VM's, none doing anything all that disk intensive, so any disk I/O 
> will tend to be somewhat random.  I'm not so worried about premature 
> death of the device, since it's a brand-new device, and we're not 
> talking hundreds of terabytes per month or anything.  My pool is 
> definitely NOT throughput stressed (rarely exceeding a 20-30 MB/sec at 
> most), so latency is more of a priority to me.  I'll give this a try,
thanks!
> 

What is your ARC size? Chances are your dataset fits entirely in the ARC.
Just a quick guess.

Cheers,
--
Saso