Data in Place allows the current generation of VSP 5000 controllers to be upgraded to next generation of controllers without having to do a data migration.
In a traditional data migration scenario the new generation controllers have to be installed, new drive boxes also have to be installed, then the new controllers and drives have to be configured, and finally the data migration has to be done. These activities are identified as activities 1 2 3 4 on the data migration scenario (Figure 1 left side)
While in a Data in Place scenario (Figure 1 rifgt side) the process is less expensive and much simpler as new drives boxes are not needed, no data migration is needed, and only one activity takes place: replacing the controllers.
Figure 1. Data in Place migration (DIP)
What is Data in Place replacement
The Data in Place upgrade from VSP 5500 series to VSP 5600 series uses the “controller type change replacement” method (Figure 2).
For the procedure to take place the microcode supporting “controller type change replacement” has to be installed in the VSP 5500 and the DKCPS power supplies of the VSP 5500 have to be replaced with the ones of the VSP 5600.
Figure 2. Controller type change replacement
It is important to mention here that with the VSP 5600 models the data compression and decompression functionalities can be offloaded to the Compression Accelerator Module. The Compression Accelerator Module is the ASIC module that assists the microcode in performing the data compression and decompression.
The algorithm of the Compression Accelerator Module is more efficient than the existing software compression performed by the controller of the VSP 5500 models and by offloading the compression and decompression functions the Compression Accelerator Module reduces the utilization of the micro processing units in the VSP 5600 models.
To have accelerated compression in the VSP 5600 units, on each controller two backup modules (BKMFs, also referred as fan modules) are replaced by accelerator compression modules (ACLF). With the ACLFs installed then compression acceleration can be enabled on a per volume basis or on a per pool basis.
Figure 3. Compression Accelerator Module
How to Perform a Non Disruptive DIP replacement
An evaluation test was conducted in Hitachi Vantara labs to verify the non‑disruptive characteristic of this procedure. A VSP 5500-2N was upgraded to VSP 5600-2N using the “DIP controller type change replacement” methodology.
- The test began with the VSP 5500 DP volumes with deduplication and SVOS‑compression enabled
- After the controllers replacement was completed, then compression acceleration was enabled
The three steps of the process are represented in the illustration in Figure 4 as steps ① ② ③. During the process a workload was generating I/O load on the 256 DP volumes of the configuration.
Figure 4. Evaluation test environment
The following sections discuss the detailed step by step procedure that was followed.
1. Non Disruptive Evaluation — Test Procedure Overview
Figure 5 summarizes what were the steps that were followed to run the test. We will be covering the details of each step in a few seconds.
Figure 5. Evaluation test procedure
2. Test Procedure — Preparation / Initial Requirements
On the VSP 5500 the preparation activities were the following (Figure 6):
- Installing the microcode supporting DIP upgrade, in this case it was SVOS 9.8
- Replacing the DKCs power supplies for those of the VSP 5600 model
Figure 6. Test procedure: preparation and initial requirements
3. Test Procedure — Disable Controller 1 paths to DP volumes
For the replacement process of each controller the first thing to do is to disable the paths that the controller has to the active DP volumes.
This is the reason you want to have at least two paths configured from different controllers to each LUN. This way when a controller is blocked the I/O activity continues uninterrupted to the otherwise affected volumes.
Figure 7 illustrates the case when controller Controller 1 is going to be replaced. In this case access to DP volumes 0 to 127 via ports 1A and 5A will be affected so before any other thing those two paths have to be logically disabled.
The example shown in the screen is a case where two paths from different controllers reach each DP volume. As mentioned, two paths is the minimum required for the I/O operation to continue uninterrupted. Nevertheless, and to avoid a single point of failure situation while a controller is been replaced, the recommendation is to have more than two paths to each DP volume all of them using different controllers.
Figure 7. Keeping I/O active while an alternate path is disabled.
4. Test Procedure — Perform controller replacement (part 1/4)
Once the paths are offline the controller can be replaced. The first thing to do is to block the controller. For this, in the Maintenance Utility menu you (Figure 8):
- Select the target controller and click [Replace (Type Change)]
- Then confirm whether hosts are shutdown or alternate paths are available
- Select Accelerator Module
- And finally click [Block]
Figure 8. Controller replacement (part 1 of 4).
5. Test Procedure — Perform controller replacement (part 2/4)
After you click [Block] you are prompted for the password and once you click [OK] the blocking process begins (Figure 9).
Figure 9. Controller replacement (part 2 of 4).
6. Test Procedure — Perform controller replacement (part 3/4)
With the controller blocked now the physical replacement of the controllers can take place. This is the process (Figure 10):
- Remove the batteries from the two backup modules in the old controller and install them in the corresponding backup modules of the new controller
- Remove the DIMMs from the old controller and install them in the new controller
- In the new controller remove the two fan modules and replace them with the ACLF modules. This activity can also be made initially on all four controllers as part of the preparation process or it can also be done later when all the controllers have been replaced.
- Finally install the new controller in the controller chassis
Figure 10. Controller replacement (part 3 of 4).
7. Test Procedure — Restore Controller 1 and enable paths (part 4/4)
Once the controller is replaced then select [Restore]. Once the restore completes then the paths that were disabled can now be brought online (Figure 11).
Figure 11. Controller replacement (part 4 of 4).
8. Enable Compression Acceleration
In the evaluation test the 256 DP volumes in the VSP 5500 had SVOS deduplication and compression enabled. Once the replacement of the four controllers was completed the DP volumes in the upgraded VSP 5600 still were with SVOS deduplication and compression enabled.
Now it was time to switch from SVOS compression to accelerated compression. To enable accelerated compression the CCI raidcom modify ldev command was used.
The completion of the transition could be verified either with Storage Navigator to verify when volumes transitioned from Converting to Enabled, and it could also be verified with the CCI raidcom get ldev command checking that COMPRESSION_ACCELERATION and COMPRESSION_ACCELERATION_STATUS both appeared as ENABLED.
Figure 12. Enabling Compression Acceleration.
9. Ops Center Administrator GUI
In case you are using Ops Center then the Administrator GUI will reflect the compression acceleration enabled status in two of its menus:
- On the Storage System detail page (Figure 13)
- On the Volumes table page (Figure 14)
Figure 13. Ops Center Storage System detail page.
Figure 14. Ops Center Volumes Table page
10. Key features of DIP controller type change replacement
Two key features of the “Data in Place controller type change replacement” new function are:
- DIP controller type change replacement is not disruptive to the I/O operation if the DP volumes are configured with two or more paths from different controllers
- There is no need to disable SVOS deduplication and compression to do the replacement
Figure 15 illustrates the non-disruptive characteristic of the DIP controller type change replacement, showing how the I/O to the volumes was not interrupted during the procedure in our evaluation test.
Figure 15. I/O to the volumes is not interrupted during the DIP replacement.