Fujitsu Improves Protection and Quality Management of Disk Arrays
In late July, Fujitsu introduced a new functionality for its disk arrays ETERNUS DX S3. These storage enhancements provide an additional level of data protection against disk failures and major crashes, and simplify the monitoring of response times for applications with different priorities.
In the traditional approach to fault tolerance in a sixth-level RAID array in ETERNUS DX, the allocation of one of the array's disks to “hot spare” is used. The main purpose of this backup disk is to write to it the contents of a failed disk restored by RAID mechanisms. After the recording is completed, the failed disk can be replaced with a new one and the recovered data from the hot spare disk can be copied to it. After performing all these data recovery operations and writing them to a hot spare disk and then to a new disk, the RAID array will again be protected from hardware failures due to RAID redundancy.
')
The problem is that as the capacity of modern hard drives grows, the time it takes to overwrite all data from a failed disk to a hot spare disk increases, so when using terabyte drives, this operation can take several hours. As a result, if the failed disk is restored, the fault tolerance of the array is reduced, because if another disk fails before the end of writing the recovered data to the hot spare disk, then some or even all data stored on the array may be irretrievably lost.
The ETERNUS Fast Recovery feature introduced at the end of July, instead of a separate hot spare disk, uses Fast Hot Spare (FHS) space allocation for hot backups on all drives in a RAID array. Due to this distribution of the hot reserve across the disks of the array, while restoring the contents of the failed disk, write operations are performed in parallel on dozens and even hundreds of drives, so the procedure is much faster. For example, when using a RAID-6 array with single-terabyte drives, it accelerates six times - from nine to one and a half hours.
Thus, when using Fast Recovery, the recovery time of a failed disk is significantly reduced, and with it the risk of data loss if another disk fails.
The second new function of ETERNUS storage systems - ETERNUS Storage Cluster - ensures the continuous availability of information stored on the array, and even in the event of a major accident that can damage not only the array itself, but the entire data center, the data will remain intact. The ETERNUS Storage Cluster technology, using the well-known owners of ETERNUS DX family of synchronous remote “equivalent” copying REC (Remote Equivalent Copy), implements a business-friendly failover procedure for other storage systems when the main storage fails. Storage or shutdown for scheduled maintenance.
The system administrator can configure the automatic start of switching to a backup system in case of a failure or manually start it before performing array maintenance (for example, when updating the microcode) or in the case of such minor failures as a planned outage. Thanks to real-time mirroring between the primary and backup arrays and transparent switching, business process downtime is almost completely eliminated due to the temporary unavailability of the storage system. This technology is supported in the entire line of ETERNUS DX S3, including the low-end models DX100 S3 and DX200 S3.
To build a Storage Cluster, in addition to the two arrays connected via Fiber Channel, a server is needed that acts as a Storage Cluster Controller cluster controller. In normal cluster operation mode, data from the main array is continuously copied to the backup using synchronous REC, and the controller constantly monitors the status of the main array. If the controller detects a failure on the main array, it will initiate a failover procedure with transfer to the backup array of the basic parameters of the main array (LUN numbers, WWN addresses, etc.), so the switch will be transparent for the application servers storing their data on main array.
Storage Cluster can be used in two scenarios - a cluster can consist of arrays installed in one data center, and arrays separated by up to 100 kilometers to protect against major disasters. Unlike similar disaster-proof solutions, a storage virtualizer does not use a virtualizing set-top box for mirroring data between arrays, which complicates management and maintenance.
Consolidated storage on a single system of large amounts of data can lead to competition for storage I / O resources between different servers and their business applications. The Auto QoS (Auto Quality of Service) feature implemented in the new version of the ETERNUS SF V16.1 management software allows you to achieve guaranteed response times for the most important business applications, even in situations where the storage system is at its maximum capacity due to high load. . Using Auto QoS, a system administrator can select for each application one of three priority levels (high, medium, or low) and set a target response time (Target Response Time) for data volumes allocated to specific applications.
ETERNUS SF monitors the I / O performance of volumes and dynamically adjusts them to ensure compliance with the Target Response Time. If the response time exceeds the target value, then additional I / O subsystem resources are automatically allocated to the data or it moves to the storage level with a higher data access speed, for example, to SSD. Auto QoS, unlike traditional QoS mechanisms, does not require the system administrator to perform complex I / O performance calculations (IOPS) to ensure the required quality of storage service.
The new ETENUS DX S3 functionality will be especially in demand by CIOs of business, who need the highest possible level of data availability and effective management of the quality of storage services for the most important business data.
In the traditional approach to fault tolerance in a sixth-level RAID array in ETERNUS DX, the allocation of one of the array's disks to “hot spare” is used. The main purpose of this backup disk is to write to it the contents of a failed disk restored by RAID mechanisms. After the recording is completed, the failed disk can be replaced with a new one and the recovered data from the hot spare disk can be copied to it. After performing all these data recovery operations and writing them to a hot spare disk and then to a new disk, the RAID array will again be protected from hardware failures due to RAID redundancy.
')
The problem is that as the capacity of modern hard drives grows, the time it takes to overwrite all data from a failed disk to a hot spare disk increases, so when using terabyte drives, this operation can take several hours. As a result, if the failed disk is restored, the fault tolerance of the array is reduced, because if another disk fails before the end of writing the recovered data to the hot spare disk, then some or even all data stored on the array may be irretrievably lost.
The ETERNUS Fast Recovery feature introduced at the end of July, instead of a separate hot spare disk, uses Fast Hot Spare (FHS) space allocation for hot backups on all drives in a RAID array. Due to this distribution of the hot reserve across the disks of the array, while restoring the contents of the failed disk, write operations are performed in parallel on dozens and even hundreds of drives, so the procedure is much faster. For example, when using a RAID-6 array with single-terabyte drives, it accelerates six times - from nine to one and a half hours.
Thus, when using Fast Recovery, the recovery time of a failed disk is significantly reduced, and with it the risk of data loss if another disk fails.
The second new function of ETERNUS storage systems - ETERNUS Storage Cluster - ensures the continuous availability of information stored on the array, and even in the event of a major accident that can damage not only the array itself, but the entire data center, the data will remain intact. The ETERNUS Storage Cluster technology, using the well-known owners of ETERNUS DX family of synchronous remote “equivalent” copying REC (Remote Equivalent Copy), implements a business-friendly failover procedure for other storage systems when the main storage fails. Storage or shutdown for scheduled maintenance.
The system administrator can configure the automatic start of switching to a backup system in case of a failure or manually start it before performing array maintenance (for example, when updating the microcode) or in the case of such minor failures as a planned outage. Thanks to real-time mirroring between the primary and backup arrays and transparent switching, business process downtime is almost completely eliminated due to the temporary unavailability of the storage system. This technology is supported in the entire line of ETERNUS DX S3, including the low-end models DX100 S3 and DX200 S3.
To build a Storage Cluster, in addition to the two arrays connected via Fiber Channel, a server is needed that acts as a Storage Cluster Controller cluster controller. In normal cluster operation mode, data from the main array is continuously copied to the backup using synchronous REC, and the controller constantly monitors the status of the main array. If the controller detects a failure on the main array, it will initiate a failover procedure with transfer to the backup array of the basic parameters of the main array (LUN numbers, WWN addresses, etc.), so the switch will be transparent for the application servers storing their data on main array.
Storage Cluster can be used in two scenarios - a cluster can consist of arrays installed in one data center, and arrays separated by up to 100 kilometers to protect against major disasters. Unlike similar disaster-proof solutions, a storage virtualizer does not use a virtualizing set-top box for mirroring data between arrays, which complicates management and maintenance.
Consolidated storage on a single system of large amounts of data can lead to competition for storage I / O resources between different servers and their business applications. The Auto QoS (Auto Quality of Service) feature implemented in the new version of the ETERNUS SF V16.1 management software allows you to achieve guaranteed response times for the most important business applications, even in situations where the storage system is at its maximum capacity due to high load. . Using Auto QoS, a system administrator can select for each application one of three priority levels (high, medium, or low) and set a target response time (Target Response Time) for data volumes allocated to specific applications.
ETERNUS SF monitors the I / O performance of volumes and dynamically adjusts them to ensure compliance with the Target Response Time. If the response time exceeds the target value, then additional I / O subsystem resources are automatically allocated to the data or it moves to the storage level with a higher data access speed, for example, to SSD. Auto QoS, unlike traditional QoS mechanisms, does not require the system administrator to perform complex I / O performance calculations (IOPS) to ensure the required quality of storage service.
The new ETENUS DX S3 functionality will be especially in demand by CIOs of business, who need the highest possible level of data availability and effective management of the quality of storage services for the most important business data.
Source: https://habr.com/ru/post/236023/
All Articles