Start a Conversation

Unsolved

This post is more than 5 years old

43691

August 6th, 2013 13:00

Ask the Expert: Isilon Performance Analysis

Welcome to this EMC

YOU MAY ALSO BE INTERESTED ON THESE ATE EVENTS...

Ask the Expert: EMC Isilon Scale-out Data Lake

Ask the Expert: Are you ready to manage deep archiving workloads with Isilon’s HD400 node and OneFS 7.2.0? Find out more…

Ask the Expert: SMB Protocol on an Isilon Cluster

Support Community Ask the Expert conversation. This is an opportunity to learn about and discuss the best practices for Isilon performance analysis, including:

  • Client work-flow considerations
  • Network considerations
  • OneFS identify and break down causes of latency or contention
  • Tools that help identify work-flow bottle-necks
  • Sizing and tuning protocol OPS to disk IOPS

This discussion begins on August 12 and concludes on August 23. Get ready by bookmarking this page or signing up to receive email notifications.

Your host:

profile-image-display.jspa?imageID=8218&size=350

John Cassidy has spent three decades developing, supporting complex solutions and simplifying problems; He brings the following to bear in OneFS Performance or Complex work-flow issues:

* Work-Flow profiling

* Simplification methods

* Measurement tools

* Deterministic Results

2 Intern

 • 

165 Posts

December 18th, 2013 09:00

Hi John

This is a very nice discussion and information provided is really helpful.

I have a question on NVRAM. Is there a way to check the NVRAM usage stats.

Thanks

Yoga

1.2K Posts

January 7th, 2014 04:00

Hello Yoga,

the are a multitude of sysctl calls related to the NVRAM coalescer,

but I'm not aware of (publicly available) documentation other

than the general caching description in the OneFS White Paper.

You might check out (and see what appears self-explaining to you):

sysctl efs.bam.coalescer_stats

sysctl efs.bam.dump_coalescer

sysctl efs.bam.coalescer.coalescers

sysctl efs.bam.coalescer

Full list of related sysctl calls:

sysctl -a -d | grep coalescer

If you find some particular useful stuff for general use, please share.

Cheers

-- Peter

2 Intern

 • 

165 Posts

January 7th, 2014 07:00

Thank you for the information Peter. I tried it out on my virtual cluster and , may be i should do it on a real one which would make more sense to me and i will keep it posted if i figure out something.

Is there a document which explains all this Peter ?

2 Intern

 • 

165 Posts

January 21st, 2014 14:00

--> "As you perform buffered write operations from smb, nfs or protocol writes. These writes will be coalesced in a write buffer e.g. If there are a sequence of proto.nfs write operations @32KB at next offsets to each other we will buffer this upto 1MB or 2MB (concurrent / stream optimized)"


Peter, I guess i am confused with this part. Please explain on this.

Until data buffered (1MB or 2MB) goes to NVRAM, the ACK is not sent back to user. If the ACK for the last 32KB data is not sent, then it will be re-transmitted. How does it work here ?



1.2K Posts

January 21st, 2014 20:00

Damal, there is no need for re-transmitting. How asynchronous NFS3 writes actually work has been defined in "RFC 1813 - NFS Version 3 Protocol Specification" (google for it; can't post links here). The key is the COMMIT mechanism.

Cheers

-- Peter

2 Intern

 • 

165 Posts

January 23rd, 2014 13:00

Found the RFC in IETF website and I guess I am clear with that.

Thank you so much Peter

20 Posts

February 10th, 2014 03:00

Dear John,

           I would request you to please explain me the following which are related to Coherent cache and number of disks in disk pool in Isilon :

1. How Isilons/OneFS forms coherent caches (READ/WRITE) across all the nodes which are part of the same cluster

2. Why the number of disks in a (disk pool = 6disks per node, per node pool up to 40 nodes ) disk pool is limited with 6 disk drives only.

     Thanks in advance.

2 Intern

 • 

165 Posts

February 10th, 2014 14:00

Hi John,

Can you take a look at the attached output for the isi statistics drive --top --long -nall --orderby=OpsIn. I see the Queued I/O is >1. Does it mention we need to add few more nodes to pool, or any other point you want to add on the output

Its a NL400 node running on OneFS 7.0.2.4

stats drive.JPG.jpg

DC1-cluster% sudo isi statistics query -snode.ifs.cache.oldest_page_age

  NodeID node.ifs.cache.oldest_page_age

       2                           8128

average                    8128

Thanks,

Damal

1 Message

December 20th, 2018 01:00

thanks for sharing such a brief and useful information

character count

No Events found!

Top