Data Protection

 View Only

Thin Image Advanced Best Practices

By Dang Luong posted 09-05-2024 19:40

  

Author: @Malaya Acharya

Introduction

The Thin Image Advanced snapshot software enables you to rapidly create point-in-time copies of mission-critical information within the storage system or virtualized storage pool without impacting host service or performance levels.

Compared to Thin Image, Thin Image Advanced significantly improves Split state performance and copy operations elapsed time. The enhancements are because of the change in how data is written after a snapshot pair is split.

For Thin Image Advanced, Data Reduction Shared (DRS) volumes with Adaptive Data Reduction (ADR) enabled are used. Because ADR stores data in 8 KB units, Thin Image Advanced also stores data in 8 KB units. This is a significant change from Thin Image, which stores snapshot data in 256 KB units. The smaller units allow Thin Image Advanced to perform better and be more space-efficient.

After a snapshot pair is split, new data is written to the P-VOL without overwriting the original data. Then, metadata is created for the secondary volume (S-VOL) that points to the original data. This is called Redirect-on-Write. The removal of the “Copy” step brings a significant improvement in performance. Figure 1 shows the sequence of operations:

Figure 1: Redirect-on-Write

Figure 1: Redirect-on-Write

Apart from the performance improvements, Thin Image Advanced delivers important enhancements in pool capacity management. After a snapshot is deleted, the pool capacity is automatically reclaimed asynchronously by garbage collection. This is done even if the P-VOL has other snapshots in the same snapshot tree. This is different from Thin Image, where you must delete the entire snapshot tree to reclaim the snapshot capacity. Otherwise, you must manually run defragmentation after deleting an individual snapshot to reclaim the used pool capacity.

Prerequisites

  • Thin Image Advanced is only supported with Data Reduction Shared volumes with the capacity-saving function enabled.
  • Thin Image Advanced requires the P-VOL and corresponding snapshots to be in the same Dynamic Provisioning (DP) pool. This is different from Thin Image, which allows separate pools to be used.
  • Thin Image Advanced requires the pool to consist of flash media (of either SSD, FMD, or SCM).

Best Practices

  • The following types of pools cannot be used for Thin Image Advanced:
    • A pool for the S-VOL that is different from the pool for the P-VOL.
    • A pool that contains external pool volumes. (Note: This is technically possible to do, but not recommended.)
    • A pool for which the Data Direct Mapping attribute is enabled.
  • A Thin Image Advanced pair cannot be created when the used pool capacity exceeds the warning threshold. In this scenario, increase the pool capacity first, and then create the pair.
  • Use consistency groups to split the Thin Image Advanced pairs that are defined in the group. Splitting the pairs using the consistency group assures data consistency at the time the storage system receives the request for storing snapshot data (splitting the pair).
  • Thin Image Advanced does not support the manual assignment of consistency group ID. Consistency group ID is automatically assigned while creating pairs.
  • An MU number can be specified while pairs are created. If an MU number is not specified, it is automatically assigned in the order of 3 to 1,023, followed by 0 to 2. Note that ShadowImage uses MU numbers 0 to 2. If Thin Image Advanced is already using MU numbers 0 to 2, Thin Image Advanced volumes cannot be shared with ShadowImage volumes.
  • Thin Image Advanced first-layer (L1) S-VOLs can also be paired with secondary-layer (L2) S-VOLs. L2 to L64 pairs are called cascaded pairs. A maximum of 64 layers can be created, and a maximum of 1,024 S-VOLs can be used for a P-VOL.
  • If two or more pairs in a snapshot group share one P-VOL, split pairs cannot be performed at the same time for the snapshot group. In this scenario, split pairs one by one.
  • Secondary volumes can be assigned to Thin Image Advanced pairs even after creating the pair. Secondary volume assignments can be released or changed as part of management activities.
  • Snapshot data of all layers in the snapshot tree cannot be deleted with the root volume by specifying -range tree.
  • Enabling storage-based encryption does not impact Thin Image Advanced performance.
  • User intervention for space reclaim is not required. After Thin Image Advanced snapshots are deleted, garbage collection is automatically performed in the background to reclaim space in the pool.
  • The performance of Thin Image Advanced in Split state can be better than Simplex because in Simplex state, Thin Image Advanced must invalidate existing data on update-write operations, which causes extra work. In Split state, Thin Image Advanced must maintain existing data for the snapshots; therefore, the invalidation step is skipped.
  • The performance of Thin Image Advanced in Split state is significantly better than that of Thin Image in Split state because host update write operations for Thin Image Advanced do not require copying the original data to the snapshot pool. The writes are added to the capacity of the P-VOL.
  • With Thin Image Advanced, Resync, Restore, and Delete operations take the same time to finish, regardless of the pair sync rate. This is because of the copy-less nature of the copy-program product. Only metadata pointers are updated during the copy operations; therefore, the amount of changed data does not impact the elapsed time.
  • The elapsed time of Thin Image Advanced Map and Unmap operations increases with P-VOLs used capacity and changed data.
  • Multi-generation snapshots and snap-on-snap have no performance impact.

Safe Snap

Safe Snap is a function of Thin Image Advanced that allows you to create immutable snapshots. You enable it by setting a retention period on a new or existing snapshot, snapshot group, or snapshot consistency group. The snapshot type can be floating (also known as volume-less) or volume-based (also known as S-VOL).

The retention period can range from one hour to 12,288 hours (or 512 days). When a retention period is set, the data on a snapshot, snapshot group, or snapshot consistency group cannot be changed. This includes disallowing changes by mounting the S-VOL to a host and writing to the volume, as well as resyncing or deleting the snapshot. Furthermore, the retention period cannot be reduced easily. This requires a one-time password from HV Support.

Safe Snap works similarly to Data Retention Utility (DRU), with the main difference being that Safe Snap enables you to set a retention period on a volume-less snapshot. DRU can only be used on volumes with a Logical Device ID (LDEV ID). As a result, Safe Snap enables you to protect many volume-less snapshots without consuming LDEV IDs.

Note: Safe Snap cannot be combined with DRU on the same snapshot volume, only one feature can be used at a time.

The following is an example of creating a snapshot pair and then setting a one-hour retention period:

raidcom add snapshot -ldev_id 1 2 -pool 0 -snapshotgroup MP0 -snap_mode cascade

raidcom modify snapshot -snapshotgroup MP0 -snapshot_data split -retention 1

You can additionally set retention on an already split snapshot, snapshot group, or snapshot consistency group, as follows:

raidcom modify snapshot -snapshotgroup MP0 –snapshot_data renew_retention -retention 1

Safe Snap Best Practices

  • The granularity of the snapshot data retention period is in hours.
  • The snapshot data retention period cannot be shortened or disabled for a pair for which the snapshot data retention period is already enabled.
  • Because the retention period is checked every 10 minutes, it can be up to 10 minutes shorter.
  • Safe Snap cannot be combined with Data Retention Utility.
  • Safe Snap does not impact performance.
  • The retention period can be extended for snapshot data in which retention is currently enabled.
  • For information about the retention period, “-key details” is added during pair display:
    raidcom get snapshot <PVOL> -key details
  • Safe Snap is only available with Thin Image Advanced and not Thin Image.
  • Increase retention period does not impact host I/O performance.

Summary

In addition to the general best practices, you must properly size a Thin Image Advanced solution based on your I/O requirements and your VSP storage system and configuration. This can be done using the CPK sizing tool and reviewing the performance test results from Global Product and Solutions Enablement (GPSE). For more details, contact your Hitachi Vantara representative.

References

Hitachi Virtual Storage Platform One Block Thin Image Advanced User Guide MK-23VSP1B015-00.

0 comments
25 views

Permalink