Chapter 7. GridStor Technology. Adding Data Paths. Data Paths for Global Deduplication. Data Path Properties

Transcription

1 Chapter 7 GridStor Technology GridStor technology provides the ability to configure multiple data paths to storage within a storage policy copy. Having multiple data paths enables the administrator to route data from multiple clients to storage on the best path available, re-route data should a data path resource fail or load balance data movement across multiple paths for performance. By defining multiple paths, the number of storage policies needed within a CommCell environment can be reduced. This will make it easier to manage backup data and help consolidate media usage. Adding Data Paths Data paths can be added in the Data Path tab of the policy copy. How multiple data paths will be used is determined in the Data Path Configuration tab. Multiple data paths can be used in a round-robin, failover, or choose best path (LAN-Free Preferred path) if available. Note that depending on how GridStor technology is configured; any data path available may be used so plan your data paths before implementing them If the existing path for the copy is to a shared library, you can configure the copy to automatically add additional paths. This is typically used with shared disk or shared tape libraries when configuring GridStor to round-robin data paths. Select the option Automatically add new data path in the Data Path Configuration tab of the policy copy. Data Paths for Global Deduplication A new Simpana 9 feature is the use of global deduplication storage policies. This allows multiple storage policies to use a single deduplication store for better deduplication ratios. If a global deduplication policy is being used all data paths will be configured in the global dedupe policy. Policies that link to the global policy cannot have their data paths modified. However, the method in which the paths will be used (LAN-Free, round-robin, or failover) can be specified for each policy that links to the global policy. This allows for a more flexible configuration. Data Path Properties One of the many powerful features of GridStor technology is the ability to independently configure data path properties within the storage policy copy. This allows custom settings at the logical level and not bound to the physical path. For example, financial data backing up to an LTO5 tape drive might require encryption. This can be enabled at the data path level of the storage policy managing the financial data. Other data managed by a different storage policy may not require encryption so it would be disabled in the LTO 5 data path for that policy copy. Even though the same physical drive is being used, the encryption will be turned on or off based on the data path configuration for a storage policy copy. The following settings can be customized for a data path:

2 Hardware compression Hardware encryption Chunk size Block size Hardware Compression If a tape drive supports hardware compression, it can be enabled in the data path properties. If no other compression has been performed on the data, enabling this option will make more efficient use of tape media. For data paths defined to write to tape libraries this option will be enabled by default. Some applications will perform compression on data as it is being backed up. If this is the case, compression should be disabled for the data path. When CommVault s data deduplication features is enabled, by default the software will compress data on the client before it is backed up to media. This will make data movement more efficient and deduplication ratios better. If the data is being copied from deduplicated disk to tape, CommVault recommends disabling compression for the tape data path. If using other third party deduplication, check with the vendor to see where compression is taking place and whether the data is decompressed by their hardware. Set the data path properties based on whether the data remains compressed or not. Hardware Encryption For tape drives that support hardware encryption, CommVault can manage configure settings and manage keys. Keys will be stored in the CommServe database. Keys can optionally be placed on the media to allow recovery of data if the CommServe database is not available at time of recovery. The data path option Via Media Password will put the keys on the media. The option No Access will only store the keys in the CommServe database. Note: If you choose the Via Media Password option it is absolutely essential that a Media Password be configured or the encrypted data can be recovered without entering any password during the recovery process. A global Media Password can be set in the System Settings in the Control Panel applet. Optionally a storage policy level password can be set in the Advanced tab of the Storage Policy Properties. Chunk Size Chunk sizes define the size of data chunks that are written to media. The default size for disk is 2GB. The default size for tape is 4GB for indexed based operations, or 16GB for non-indexed database backups. The data path Chunk Size setting can override the default settings. A higher chunk size will result in a more efficient data movement process. In highly reliable networks, increasing chunk size can improve performance. However for unreliable networks, any failed chunks will have to be rewritten, so a larger chunk size could have a negative effect on performance. Block Size The default block size CommVault uses to move and write data to media is 64k. This setting can be set from k. Like chunk size, a higher block size can increase performance. However, block size is hardware dependent. Before modifying this setting ensure all hardware being used at your production and DR sites support the higher block size. If you are not sure, don t change this value. When writing to tape media, changing the block size will only become affective when CommVault rewrites the OML header on the tape. This is done when new media is added to the library, or existing media is recycled into a scratch pool. Media with existing jobs will continue to use the block size established by its OML setting. When writing to disk, it is important to match the block size data path setting to the formatted block size of the disk. Matching block sizes can greatly improve disk performance. The default block sizes operating systems use to format disks is usually much smaller than the default setting in CommVault. It is strongly recommended to format disks to the block size being used in CommVault. Consult with your hardware vendors documentation and operating system settings to properly format disks.

3 Data Path Configurations Data Path Configuration allows you to specify how multiple data paths will be used. This provides a simple means to cluster media agents for scalability in large environments. There are several methods to configure and customize data paths: Preferred path (LAN-Free) Default path (LAN based) Round-Robin load balancing Failover Data path override Default Data Path When creating a storage policy, a default path is defined during the policy creation wizard. If additional paths are added, the default path can be changed. Given no other guidance, all LAN based subclients associated with the storage policy will use the default path. Example: You have just added a new disk library in your CommCell environment. You have an existing storage policy with a default path pointing to a tape library and you want to redirect all backup data to the new disk storage. Solution: In the storage policy copy, add the data path to the disk library and select the new path as the default path.

4 In the following diagram, the default tape data path was changed to a disk path. All subclient data associated with the storage policy will now use the disk data path for subsequent data protection jobs. Preferred Data Path While multiple paths are defined, the preferred path is determined by the CommVault software. A preferred path is one that uses a locally hosted (non-lan) Media Agent. When a preferred path is available, it will always be used to move data to storage. In a preferred path configuration where the preferred path becomes unavailable, alternate paths will not be used. You can accomplish LAN-Free backups by co-locating Media Agent software on Client enabled servers. Example: You have several servers which will be backing up over the LAN and several servers attached to the SAN. You want to use a dedicated Media Agent for LAN based backups and you want to backup any SAN attached servers using a LAN-Free backup. Solution: By installing Media Agent software on each SAN attached server you can perform LAN-Free backups. All LAN backups will use the dedicated Media Agent. The following diagram shows LAN based clients using a LAN based Media Agent data path and clients with Media Agents directly installed using LAN-Free data paths to storage.

5 Data Path Load Balancing (Round-Robin) In a small backup environment using one Media Agent may be adequate for all backup needs. As an environment grows multiple Media Agents may be needed to distribute or route data movement for better performance. You can define multiple paths within a storage policy copy and configure those paths to Round-Robin load balance In order to use multiple media agents in a round robin method, they must all have access to the same index cache location. Prior to Simpana Version 9 that required configuring a shared index cache. A shared index cache is configured by a hosting media agent presenting the index cache location and other media agents connecting to the index cache over UNC paths. This created a single point of failure for the index cache location. A new feature of Simpana V9 is the use of an Index Cache Server (ICS). An Index Cache Server is a dedicated media agent that hosts backup copies of the index cache. The media agent will host a local copy of the index files and send a copy of the index files to the ICS. All media agent that are associated with the ICS can access the index files and eliminates the single point of failure of a shared index cache. Though a shared index cache is still supported, the preferred method for index sharing is through the use of the Index Cache Server. Example: You have several servers backing up over the LAN. The Media Agent currently handling the backups is no longer able to move all data to the library within your backup window. Solution: You install another Media Agent and add the data path to the storage policy copy. You configure GridStor technology to Round-Robin (load balance) between the paths. The following diagram shows multiple LAN based clients using Round-Robin data paths with two Media Agents. This method provides greater scalability to meet protection windows.

6 Data Path Failover In some cases you may also want to define another path if the default path is not available. Data path failover can be used to failover to another path if the default path is unavailable or if a resource in the path is offline or unavailable. In order to configure data path failover you must have a shared index cache, an index cache server, or the media agent must be able to see all libraries configured in the data path failover. Example: You have a Media Agent with two direct attached libraries. You want a file server to backup to the disk library. However, if the library goes offline you want to be able to use the tape library to ensure the data is protected. Solution: Define the data path to both libraries in the storage policy. Define the disk library path as the default path. Configure the storage policy copy to failover to the other path if resources are off-line. The following diagram shows a server backing up to a disk library. The storage policy is configured with a tape library data path. If the disk library cannot be written to the tape library will be used as a failover path.

7 Media Agent Access to indexes In order to use GridStor Round-Robin or data path failover between different Media Agents, all Media Agents must have access to the same index cache location. This is required because dependent data protection jobs (incremental or differential) require updating the indexes that were created when the full backup was run. In Simpana v9 an Index Cache Server is the preferred method of providing shared access to indexes. Prior to v9 a shared index cache was used. Both of these concepts are covered in the CommVault Data Management Concepts chapter. Data Path Override Storage policies allow multiple data paths to be defined within each copy. When configured in the round-robin or failover mode all the subclients associated with the policy copy will use all the paths defined within the policy based on the policy configuration. The data path override option, configured in the subclient properties, can be used to determine specific paths within the policy that the subclient will use. This method of specifying data paths can be useful when trying to consolidate the number of storage policies being used. Example: You want to backup several servers over the production network and load balance them through two Media Agents. You also have a Database server you want to perform LAN free backups for. For off-site disaster recovery you want to consolidate all three servers to one tape. Solution: Create a single storage policy defining all three paths for load balancing. The database server will always use the LAN free path. In each subclient of the LAN backup servers configure data path override not to use the database Media Agent. Now they will only load balance between the two dedicated Media Agents. All subclients on LAN servers are configured to only use dedicated media agents by specifying the paths in the Data Path Override configuration.

8 Note: This solution adds additional complexity to a CommCell environment. Use this storage policy consolidation method when data from LAN based and LAN free clients require consolidation to secondary copy media. Using GridStor Technology for Storage Policy Consolidation Since storage policies create logical divisions for business data it is important not to configure too many policies as each policy will require its own set of media. If you have ten storage policies assigned to a tape library and want to send data off-site every day you will be sending a minimum of ten tapes off-site daily. Using GridStor features to define multiple data paths allows you to consolidate the number of storage policies needed within your environment. Data can be managed centrally through a storage policy but data paths can be configured to distribute backups to the appropriate storage locations. Since the data is managed by the same policy, backup jobs can be consolidated on media for more efficient media usage. GridStor Technology and Simpana Deduplication With Simpana Deduplication using the GridStor feature is not always the preferred configuration method. When using shared disk libraries with multiple Media Agents, data blocks can become scattered to different mount paths making auxiliary copy and restore performance slow. If multiple Media Agents data paths are added to a dedupe enabled storage policy then database access to all but the hosting Media Agent will require network paths to access and compare signatures. For best performance and scalability it is recommended to use dedicated storage policies when using a shared disk library. For resiliency in data protection operations, additional data paths can be configured as failover paths, but Round Robin is not recommended.