AutoBalanceLin and LIN

[scott]

Can anyone explain exactly what a LIN is, and the AutoBalanceLin job?

 

I have a 9 node - 280TB cluster.  It seems like my jobs are running for excessively log periods of time - (weeks). I'm currently 10 days and about 1/2 way through a AutoBalance and notice the processed LIN count is around 140 million. 

 

I'm guessing a LIN is something like a global inode?   Is there a LIN per Node ratio I should be planning for?

 

Thanks

 

-Scott

 

 

[Luc Simard]

LINs are Logical iNODES, Isilon unlike other *NIX platforms are not tied to a finite number of iNodes. You are  correct in your statement this is a cluster wide value since Isilon uses a Global Name Space system.

There is no defined ratio of Logical iNodes per nodes, it is dynamically balance per pool and throughout the cluster. I suspect your workflow in based onto a lot of smaller files and if you have directory structure that are shallower, with large amount of files in them , the tree walk is the effect seen here. Like any Filesystem out there, direcotry and file layout will influence overall performance. It's hard to say without knowing more about your environment.

If that becomes a concern, you should discuss with your Technical advisers or sales engineer, isilon offers a Global Namespace acceleration based on SSD Drives, where a cache entry of the LIN is store on SSD  for responsiveness. See their web site, S200+SSD/X200+SSD/X400+SSD.

The option is not available on older nodes, unless you already have 32000x or IQ10000S/IQ5000S nodes. If you find the performance not adequate, i recommend you contact their support team. 

Luc 

 

-Scott

 

 

--
You received this message because you are subscribed to the Google Groups "Isilon Technical User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isilon-user-gr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Blake Golliher]

I am not an expert, but I have it in my head that a LIN is like a OneFS block.  If you have a autobalancelin job running, it'll balance lins across the cluster.  Using autobalance, it'll balance based on file objects.  In my experience, autobalancelin is much faster for balancing out data across a cluster.  Especially with a very dense number of files in OneFS, and have added some number of new nodes.

On Thu, May 30, 2013 at 9:55 AM, scott <huntm...@gmail.com> wrote:

 

-Scott

 

 

--

[Luc Simard]

Blake is spot on on the Speed aspect , AutoBalanceLIN vs Autobalance