About Virtual Partition Manager
Your storage system can connect to multiple hosts and can be shared by multiple users, which can result in conflicts among users. For example, if a host issues many I/O requests or reads or writes a large amount of data, the I/O performance of other hosts may be affected. Virtual Partition Manager allows you to create multiple virtual cache memories called cache logical partitions (CLPRs), each allocated to different hosts, to prevent contention for cache memory and improve I/O performance.
A cache logical partition is a pool of the cache and parity groups in the storage system. Partitioning cache into one or more CLPRs allows storage administrators to dedicate individual CLPRs to a different host, preventing I/O contention for cache memory. Partitioning cache dedicates cache resources for exclusive use by specific applications to maintain priority and quality of service for business-critical applications. Storage administrators can secure or restrict access to storage resources to ensure confidentiality for specific applications. By dedicating resources to each partition as needed, a high quality of service can be maintained for all users.
You can allocate one of the following for a CLPR:
- A parity group
- An external volume
- A virtual volume of Dynamic Provisioning or Thin Image
CLPR names
CLPR0 is the default CLPR in the storage system. If you have not yet created any cache logical partitions, all cache belongs to CLPR0.
The following table lists the reserved CLPR names and corresponding CLPR IDs. These CLPR names are reserved with each CLPR ID. You cannot use another CLPR ID. For example, if the CLPR ID is 1, you cannot change the CLPR name to CLPR2.
CLPR ID |
CLPR name |
CLPR ID |
CLPR name |
0 |
CLPR0 |
16 |
CLPR16 |
1 |
CLPR1 |
17 |
CLPR17 |
2 |
CLPR2 |
18 |
CLPR18 |
3 |
CLPR3 |
19 |
CLPR19 |
4 |
CLPR4 |
20 |
CLPR20 |
5 |
CLPR5 |
21 |
CLPR21 |
6 |
CLPR6 |
22 |
CLPR22 |
7 |
CLPR7 |
23 |
CLPR23 |
8 |
CLPR8 |
24 |
CLPR24 |
9 |
CLPR9 |
25 |
CLPR25 |
10 |
CLPR10 |
26 |
CLPR26 |
11 |
CLPR11 |
27 |
CLPR27 |
12 |
CLPR12 |
28 |
CLPR28 |
13 |
CLPR13 |
29 |
CLPR29 |
14 |
CLPR14 |
30 |
CLPR30 |
15 |
CLPR15 |
31 |
CLPR31 |
Requirements for using performance functions
- Performance management products: The following software products are required for performance management operations.
- Hitachi Performance Monitor
- Hitachi Server Priority Manager
- Cache Residency Manager
- Hitachi Virtual Partition Manager
- Access: Administrator access for
Hitachi Device Manager - Storage Navigator or write access for the performance management software products is required to perform operations. Users without Administrator access or write access can only view the performance management information and settings. You need the specific administrator roles for each function:
- Performance Monitor: Storage Administrator (Performance Management)
- Server Priority Manager, Cache Residency Manager, Virtual Partition Manager: Storage Administrator (System Resource Management)
- License keys: The license keys for the performance management software products must be installed. For details about installing license keys, see the System Administrator Guide.
- Java: Java is required for Server Priority Manager and Cache Residency Manager. For details about installing Java and configuring Device Manager - Storage Navigator, see the System Administrator Guide.
- Hitachi Device Manager - Storage Navigator secondary windows: Secondary windows must be enabled before you use Server Priority Manager or Cache Residency Manager. Java and some settings of Hitachi Device Manager - Storage Navigator are required for the Hitachi Device Manager - Storage Navigator secondary window. For details, see the System Administrator Guide.
- Cache memory for Cache Residency Manager: Use of Cache Residency Manager might require additional cache memory in your storage system. For details, see Estimating cache size for Cache Residency Manager.
- Cache memory for Virtual Partition Manager: Use of Virtual Partition Manager might require additional cache memory in your storage system.
System Option Modes for Virtual Partition Manager
To provide greater flexibility, the storage systems have additional operational parameters called system option modes (SOMs) that allow you to tailor the storage system to your unique operating requirements. The SOMs are set on the service processor (SVP) by your service representative. Review the SOMs for your storage system, and work with your service representative to ensure that the appropriate SOMs for your operational environment are configured on your storage system.
Mode |
Category |
Description |
Default |
MCU/RCU |
454 |
Virtual Partition Manager |
CLPR (function of Virtual Partition Manager) partitions the cache memory in the storage system into multiple virtual cache and assigns the partitioned virtual cache for each use. If a large amount of cache is required for a specific use, it can minimize the impact on other uses. The CLPR function works as follows depending on whether SOM 454 is set to ON or OFF. Mode 454 = OFF (default): The amount of the entire destage processing is periodically determined by using the highest workload of all CLPRs (*a). (The larger the workload is, the larger the amount of the entire destage processing becomes.) *a: (Write Pending capacity of CLPR#x of concerned MPB) ÷ (Cache capacity of CLPR#x of concerned MPB), x=0 to 31 CLPR whose value above is the highest of all CLPRs Because the destage processing would be accelerated depending on CLPR with high workload, when the workload in a specific CLPR increases, the risk of host I/O halt would be reduced. Therefore, set SOM 454 to OFF in most cases. Mode 454 = ON: The amount of the entire destage processing is periodically determined by using the workload of the entire system (*b). (The larger the workload is, the larger the amount of the entire destage processing becomes.) *b: (Write Pending capacity of the entire system of concerned MPB) ÷ (Cache capacity of the entire system of concerned MPB) Caution: Because the destage processing would not be accelerated even if CLPR has high workload, when the workload in a specific CLPR increases, the risk of host I/O halt would be increased. Therefore, set SOM 454 to ON only when a CLPR has constant high workload and the I/O performance in a CLPR with low workload has higher priority than host I/O halt in the CLPR with high workload. Notes:
|
OFF |
- |
Cautions and restrictions for Virtual Partition Manager
- Increased cache memory requirements
Install additional cache memory before partitioning cache. Adding cache memory after CLPRs have been created is difficult and can be problematic. For details, see Calculating cache capacity.
- Impact on host performance
Creating and changing cache logical partitions can significantly degrade host performance and should only be performed during the initial installation and setup of the storage system or during maintenance. If you must perform these changes on a production machine, use Hitachi Performance Monitor to verify that the write pending rate and sidefile occupancy rate of the CLPR and CLPR0 satisfy the following formulas on all MP blade or unit:
- For CLPR with decreased cache capacity
Write pending rate x cache capacity before operation / cache capacity after operation < 30%
Sidefile occupancy rate x cache capacity before operation / cache capacity after operation < sleep wait threshold x 50%
- For CLPR with increased cache capacity
Sidefile occupancy rate < sleep wait threshold x 50%
- For CLPR with decreased cache capacity
- When you move resources in CLPRs, make sure that the writing pending rates of the MP blade or unit for each CLPR are less than 50 percent at both the source and the destination. The process for changing the CLPR settings will be suspended until the write pending rates of each MP blade or unit become less than 50 percent.
- Duration of cache partition operations
Adding or changing CLPR definitions or configurations can take hours to complete. After a CLPR operation has been started, you cannot cancel or modify the process until the operation is complete.
- Interoperability restrictions for cache partitioning
Application
Behaviors
ShadowImage/ ShadowImage for Mainframe
You cannot use ShadowImage or ShadowImage for Mainframe Quick Restore functions that affect multiple CLPRs.
Volume Migration
You cannot use Volume Migration when it affects multiple CLPRs.
Cache Residency Manager (VSP G1x00 and VSP F1500)
- A parity group containing LDEVs assigned to Cache Residency Manager cache areas cannot be migrated to another CLPR.
- When you decrease the cache capacity of the CLPR to which the cache area of Cache Residency Manager is allocated, perform the followings steps:
- Cancel the Cache Residency Manager bind mode or priority mode setting.
- Change the cache capacity of the CLPR.
- Set the Cache Residency Manager bind mode or priority mode again.
Universal Replicator
Universal Replicator data volumes and journal volumes can belong to different CLPRs. All journal volumes in the same journal must belong to the same CLPR. If not, an error occurs.