AnsweredAssumed Answered

HUS - How is total reclaimable calculated?

Question asked by Paul Hutchings on Jan 2, 2015
Latest reply on Jan 22, 2015 by Anahad Dhillon

Like • Show 1 Like1
Comment • 6

We have a HUS110 with tiering enabled (2 tiers).

We have volumes assigned to VMware (5.5 U2), all volumes are 5TB and all our VMDKs are thin provisioned via VMware.

I'm confused at how "Total Reclaimable" is calculated under DP Optimisation.

For example, I have a 5TB LUN formatted as VMFS and on that VMFS I create a single thin 2TB VMDK.

After some time, when I look the volumes under DP Optimisation that volume will show approx 2.1TB used with 300GB reclaimable - how can the usage on the volume exceed the maximum size of the VMDK when the VMDK is the only thing on it?

Similarly I have volumes where VMware shows there is 900GB in use but under DP Optimisation I see 1.3TB used with 400GB reclaimable - we definitely haven't created or deleted 400GB worth of VM's.

I suspect I'm missing something fundamental but I've read the HUS guide and just don't see it so before I open an SR...

Thanks,

Chuck Strickland Jan 20, 2015 5:07 AM

Correct Answer

I fought this battle a few months ago.

when a page is moved to a different tier the page the "from" page is not released but remains assigned to the lun. Yep, that is right. thus the "assigned" capacity of the luns grows and grows and grows.

There is a process to reclaim this space but by default it doesn't work well. a lun must have a reclaimable space of 500GB before the reclaim process is run for that lun.

go into the dp optimization and turn on Auto DP optimization for each lun with a trigger size of 10GB. A process runs at midnight and any luns meeting this reclaimable space requirements will have their space reclaimed.

you will still see the space grow during the day.

you can also slow down the page migration rate but that has the side effect of not moving pages as quickly.

BTW, none of this is adequately documented.

See the reply in context

No one else had this question

Outcomes

Chuck Strickland Jan 20, 2015 5:07 AM

I fought this battle a few months ago.

There is a process to reclaim this space but by default it doesn't work well. a lun must have a reclaimable space of 500GB before the reclaim process is run for that lun.

you will still see the space grow during the day.

you can also slow down the page migration rate but that has the side effect of not moving pages as quickly.

BTW, none of this is adequately documented.

Actions

Paul Hutchings @ Chuck Strickland on Jan 20, 2015 8:04 AM

Chuck thanks, that at least makes sense - do you have any more info before I open an SR to try to find out something official?

Like yourself I've scoured the documentation and KB and can't find anything and this seems like a flaw/bug or "something" that just doesn't quite seem right.

Actions

Chuck Strickland @ Paul Hutchings on Jan 20, 2015 8:14 AM

Unfortunately there isn’t any addition information available.

Support didn’t have anything and since this “is working as designed” it is not considered break fix or a bug.

We had to use the few pieces of information I found scattered around the various HDS sites, many phone calls and “experimentation” on the system to figure this out.

Bottom line, setup the auto dp optimization on each volume and let the array handle it.

I’ve done this at several sites now and it works well.

chuck

Actions

Anahad Dhillon

@ Chuck Strickland on Jan 20, 2015 12:45 PM

Hello Folks,

There are a few clarifications that need be made here. It is true that what you see here (increase in ‘reclaimable’ space) is due to the way HDT is designed for the HUS family and to automatically release back (reclaim) this space, HUS is equipped with a feature called ‘AUTO DP OPTIMIZE’. This feature runs automatically at midnight once enabled. To give user more control there is a trigger setting within this feature, user can set a minimum value ‘reclaimable space’ available before this feature runs. This trigger can be set anywhere from 1 GB to 128TB. So basically, if you don’t care about reclaiming a small amount of capacity you can set this trigger to an amount that makes sense for your environment and this feature won’t run until your reclaimable space gets to that level. Or you can set this trigger to 1GB and it will run every night when a single Chunk can be reclaimned.

From the HUS Provisioning Guide

Auto DP optimization mode

Auto DP optimization mode monitors unused capacity of each tier in a DP volume and performs DP optimization automatically. In the DP volume where auto DP optimization mode is set, the unused capacity of each tier is monitored at midnight every day and DP optimization starts when the unused capacity equals or exceeds the auto DP optimization threshold value. If the tier usage rate equals or exceeds the depletion alert threshold value, excluding the Buffer Space for New Page Assignment and the Buffer Space for Tier Relocation, DP optimization mode starts if the unused capacity of the tier equals or exceeds 1 GB. Table 10-3 shows the conditions when auto DP optimization is performed.

As you might be aware that HDT on HUS family assigns capacity in terms of Chunks. These 1 GB chunks consist of 32 x 32 MB pages (except SSD, where a Chunk is 32MB). HDT relocation moves in unit of 32MB pages (not the 1GB Chunks). So if a page is moved from this 1GB chunk to a different tier, there is an opportunity to rebalance pages within that tier to release back chunks that are not completely full. This is done by using DP Optimization (when run manually, includes ZPR) and Auto DP Optimization (automatic version no ZPR).

From the HUS Provisioning Guide

When a host issues a data write command to a Hitachi Dynamic Provisioning pool comprised of SSDs: 1. Hitachi Dynamic Provisioning initially allocates a 1 GB chunk of the Dynamic Provisioning pool.

2. Thereafter, Hitachi Dynamic Provisioning allocates a 32 MB page into which the data is written.

As more write operations occur, additional 32 MB pages are allocated. Eventually, another 1 GB chunk of the pool is allocated and more 32 MB pages are used as necessary.

Under Optimizing a DP pool

Optimizing can also eliminate capacity taken by binary zero data patterns and combine partially used chunks assigned to a DP-VOL so capacity can be released from DP-VOLs. The released capacity will increase the free capacity in the DP pool.

Hope this helps. Feel free to reach back if you need more clarifications.

Best of Regards,

1 person found this helpful

Actions

Paul Hutchings @ Anahad Dhillon on Jan 21, 2015 9:23 AM

Hi Anahad,

Thanks for the response and info - I guess this may not be a "one size fits all" question but is there any kind of best practise/guidance on what is a sensible default?

It looks like it has to be enabled per volume - I guess there isn't a way to enable it globally for all volumes automatically?

Thanks.

Actions

Anahad Dhillon

@ Paul Hutchings on Jan 22, 2015 7:31 AM

Hi Paul,

You can start around 15GB as the trigger. Note that if your pool is on the full side, meaning your reach your depletion threshold then Auto DP optimize will ignore this trigger and run even if only 1GB can be reclaimed.

The optimize setting is per DP VOL but i would imagine that you can select multiple volumes at the same time. I apologize, I do not have access to a machine so confirm this step.

Regards,

1 person found this helpful

Actions

6 Replies

Chuck Strickland Jan 20, 2015 5:07 AM
Unmark Correct
Correct Answer
I fought this battle a few months ago.
when a page is moved to a different tier the page the "from" page is not released but remains assigned to the lun. Yep, that is right. thus the "assigned" capacity of the luns grows and grows and grows.
There is a process to reclaim this space but by default it doesn't work well. a lun must have a reclaimable space of 500GB before the reclaim process is run for that lun.
go into the dp optimization and turn on Auto DP optimization for each lun with a trigger size of 10GB. A process runs at midnight and any luns meeting this reclaimable space requirements will have their space reclaimed.
you will still see the space grow during the day.
you can also slow down the page migration rate but that has the side effect of not moving pages as quickly.
BTW, none of this is adequately documented.
Like • Show 1 Like1
Actions
- Paul Hutchings @ Chuck Strickland on Jan 20, 2015 8:04 AM
  Mark Correct
  Correct Answer
  Chuck thanks, that at least makes sense - do you have any more info before I open an SR to try to find out something official?
  
  Like yourself I've scoured the documentation and KB and can't find anything and this seems like a flaw/bug or "something" that just doesn't quite seem right.
  Like • Show 0 Likes0
  Actions
  - Chuck Strickland @ Paul Hutchings on Jan 20, 2015 8:14 AM
    Mark Correct
    Correct Answer
    Unfortunately there isn’t any addition information available.
    Support didn’t have anything and since this “is working as designed” it is not considered break fix or a bug.
    
    We had to use the few pieces of information I found scattered around the various HDS sites, many phone calls and “experimentation” on the system to figure this out.
    
    Bottom line, setup the auto dp optimization on each volume and let the array handle it.
    I’ve done this at several sites now and it works well.
    chuck
    Like • Show 0 Likes0
    Actions
    - Anahad Dhillon @ Chuck Strickland on Jan 20, 2015 12:45 PM
      Mark Correct
      Correct Answer
      Hello Folks,
      There are a few clarifications that need be made here. It is true that what you see here (increase in ‘reclaimable’ space) is due to the way HDT is designed for the HUS family and to automatically release back (reclaim) this space, HUS is equipped with a feature called ‘AUTO DP OPTIMIZE’. This feature runs automatically at midnight once enabled. To give user more control there is a trigger setting within this feature, user can set a minimum value ‘reclaimable space’ available before this feature runs. This trigger can be set anywhere from 1 GB to 128TB. So basically, if you don’t care about reclaiming a small amount of capacity you can set this trigger to an amount that makes sense for your environment and this feature won’t run until your reclaimable space gets to that level. Or you can set this trigger to 1GB and it will run every night when a single Chunk can be reclaimned.
      
      From the HUS Provisioning Guide
      Auto DP optimization mode
      Auto DP optimization mode monitors unused capacity of each tier in a DP volume and performs DP optimization automatically. In the DP volume where auto DP optimization mode is set, the unused capacity of each tier is monitored at midnight every day and DP optimization starts when the unused capacity equals or exceeds the auto DP optimization threshold value. If the tier usage rate equals or exceeds the depletion alert threshold value, excluding the Buffer Space for New Page Assignment and the Buffer Space for Tier Relocation, DP optimization mode starts if the unused capacity of the tier equals or exceeds 1 GB. Table 10-3 shows the conditions when auto DP optimization is performed.
      
      As you might be aware that HDT on HUS family assigns capacity in terms of Chunks. These 1 GB chunks consist of 32 x 32 MB pages (except SSD, where a Chunk is 32MB). HDT relocation moves in unit of 32MB pages (not the 1GB Chunks). So if a page is moved from this 1GB chunk to a different tier, there is an opportunity to rebalance pages within that tier to release back chunks that are not completely full. This is done by using DP Optimization (when run manually, includes ZPR) and Auto DP Optimization (automatic version no ZPR).
      
      From the HUS Provisioning Guide
      When a host issues a data write command to a Hitachi Dynamic Provisioning pool comprised of SSDs: 1. Hitachi Dynamic Provisioning initially allocates a 1 GB chunk of the Dynamic Provisioning pool.
      2. Thereafter, Hitachi Dynamic Provisioning allocates a 32 MB page into which the data is written.
      As more write operations occur, additional 32 MB pages are allocated. Eventually, another 1 GB chunk of the pool is allocated and more 32 MB pages are used as necessary.
      
      Under Optimizing a DP pool
      Optimizing can also eliminate capacity taken by binary zero data patterns and combine partially used chunks assigned to a DP-VOL so capacity can be released from DP-VOLs. The released capacity will increase the free capacity in the DP pool.
      Hope this helps. Feel free to reach back if you need more clarifications.
      Best of Regards,
      
      1 person found this helpful
      Like • Show 3 Likes3
      Actions
      - Paul Hutchings @ Anahad Dhillon on Jan 21, 2015 9:23 AM
        Mark Correct
        Correct Answer
        Hi Anahad,
        
        Thanks for the response and info - I guess this may not be a "one size fits all" question but is there any kind of best practise/guidance on what is a sensible default?
        
        It looks like it has to be enabled per volume - I guess there isn't a way to enable it globally for all volumes automatically?
        
        Thanks.
        Like • Show 0 Likes0
        Actions
        Anahad Dhillon @ Paul Hutchings on Jan 22, 2015 7:31 AM
        Mark Correct
        Correct Answer
        Hi Paul,
        
        You can start around 15GB as the trigger. Note that if your pool is on the full side, meaning your reach your depletion threshold then Auto DP optimize will ignore this trigger and run even if only 1GB can be reclaimed.
        
        The optimize setting is per DP VOL but i would imagine that you can select multiple volumes at the same time. I apologize, I do not have access to a machine so confirm this step.
        
        Regards,
        1 person found this helpful
        Like • Show 1 Like1
        Actions

Jive

Hitachi Vantara Community

HUS - How is total reclaimable calculated?

Outcomes

Related Content

Recommended Content