Business-oriented storage systems ETERNUS DX S3
Earlier we already talked about updating the Fujitsu server line , which happened this fall. Also in early November of the outgoing year, Fujitsu released an updated family of storage systems ETERNUS DX S3. This update can be called revolutionary, as the performance of some models has increased 5 times. But first things first.
The new line of Fujitsu ETERNUS DX S3 is truly business oriented. Its development was made based on the real requirements of the business to such products. And only then, on the basis of these requirements, were the appropriate technologies and solutions selected or developed. In fact, it can be said that it is the business requirements that formed the basis of the design specification for the design of the updated family.
')
Thus, for the first time in the market, a new interface has been used for connecting SAS3 disks, providing data rates up to 12 Gb / s, which allows up to two times faster access to data placed on disks.
The next important change is the flexible and fine-tuning of the system of priorities and automatic quality of service (QoS), which allow, depending on the business objectives, to provide guaranteed levels of service for various processes. So a new generation of ETERNUS DX family allows you to allocate a certain amount of cache memory, a certain bandwidth, to guarantee the required number of input / output operations (IOPS) for a particular application, for a particular group of users, or for a particular logical volume. In other words, you just need to determine the priorities and response time for the tasks and applications used, depending on their importance to the business, ETERNUS DX S3 will do the rest. Also, based on the results of the work, billing of the use of resources by certain tasks can be done, which may be applicable, for example, to adjust the established priorities or to correlate the costs of operating the system to a particular project.
The main idea of ​​a single storage family from one vendor is to have a flexible architecture to support growing data volumes. What does this mean in reality in our case? First, the ETERNUS DX S3 offers unique scalability within a single model. Secondly, it supports scaling by data-in-place transition to a higher family model without loss of data and the need for a very short time.
For example, in the new family there are two entry-level models: DX100 and DX200. At the same time, the first can be turned into the second by simply replacing the RAID controllers.
It should also be noted that the new family of new high-density disk shelves, in which you can install up to 60 disks.
And also, as in the previous generation of storage systems (ETERNUS DX S2), there were options available for additional disk shelves for 12 3.5 ”disks and 24 for 2.5” disks.
In addition, due to the use of the new SAS3 interface in new models, the limitation on the number of SSDs used has been removed, and the unified storage management interface allows you to switch from low-end models to older ones without retraining personnel.
All models of this family are the so-called Real Unified Storage. This means that they can be used as SAN and as NAS-storage simultaneously. In this case, in systems that use replication or increase in reliability, the same licenses will work for both block (SAN) and file (NAS) access. Possible integration options for the considered storage systems in the IT infrastructure of enterprises via Fiber Channel, FCoE, iSCSI, SAS, Infiniband, 1 and 10 Gb Ethernet CIFS / NFS. It is also expected that access to DSS via HTTP and FTP protocols will appear in the near future.
The ETERNUS DX S3 family uses the VxWorks real-time operating system developed by Wind River Systems. Its effectiveness is confirmed by the fact that it is used in the NASA space program, in particular, under its control , Curiosity was landed on the surface of Mars. And the fact that Wind River is a division of Intel, whose multi-core Intel Xeon processors are used in the storage itself, speaks about the quality optimization for the processors used and allows for achieving the highest possible performance of the system as a whole.
At the very beginning it was already said that the update that occurred could be called revolutionary. To ensure that this statement is not unsubstantiated, a table is presented below with various coefficients demonstrating the changes that have occurred in comparison with the corresponding storage systems of the previous generation Fujitsu ETERNUS DX family (S2):
The twofold increase in bus speed is explained by the transition from SAS2 to SAS3, which was noted earlier. The 2nd level cache, the so-called Extreme Cache, can be delivered only to mid-level systems by installing PCIe SSD drives, the maximum cache size is 5.6 TB. The use of Extreme Cache will reduce the number of used hard drives to 80%, provided that their number was due to the need to obtain a specified number of IOPS. So, if the necessary data will be in Extreme Cache in 30% of cases, the specified 150,000 IOPS for reading can be obtained on 64 HDD (+ 4 SSD Extreme Cache), while to achieve the same value on hard drives alone 360 HDD.
If this amount of disk cache is not enough, with the next version of the controller microcode, it will be possible to allocate a part of the SSD to the cache directly from the disk shelves. As for IOPS, the comparison of current ETERNUS DX models for this indicator can be seen below:
Moreover, these figures do not necessarily show the maximum possible result for a particular storage system. For example, at the end of December, ETERNUS DX200 S3 with just 14 SSD disks showed a result of more than 200k IOPS .
Due to the use of high-density disk shelves and more efficient use of resources, we are trying to increase the utilization rate of the average disk array from the current standard for the industry 40-50% to 80-90%.
Bidirectional connection of disk shelves ETERNUS DX S3 even in case of complete failure or the absence of one disk shelf ensures the availability of the rest, increasing the reliability of storage. Schematically, such a connection can be represented as follows:
where CE is Controller Enclosure, DE is Disk Enclosure. As can be seen from the diagram, if a unidirectional connection is used, the disk shelves following the failed one would also be inaccessible. Bidirectional connectivity solves this problem.
Another interesting solution that will be available in June of this year is the Fast Rebuild technology developed by Fujitsu. It is designed to solve the bottleneck problem that occurs if one of the disks fails. In this case, the disk on which data recovery is performed becomes “super hot”, since until the completion of the rebuild, data will be written to it at once from all disks of the array. Usually the distribution of I / O operations in this case is 1 to 16, i.e. at 16 I / O, which will serve the hosts, only one will serve the process of restoring the array. Thus, the rebuild process when using large disks (2TB and above) can last a week or even more.
It is to solve this problem and technology was developed Fast Rebuild. Its essence lies in the fact that free backup blocks are initially placed on each of the disks in the array (Reserved - in the diagram below). Thus, data recovery is performed immediately to all disks in these areas, which solves the bottleneck problem and speeds up the recovery procedure at times.
At the same time, it also remains possible to change the priority of a rebuild, and its change can be configured according to a specific schedule. For example, so that the priority of a rebild is increased during off-hours, significantly speeding up the recovery process.
Summing up the above, we note that all Fujitsu ETERNUS DX S3 models of the primary and secondary levels have a unified functionality and control environment. Block and file access to the data is produced within the same storage controller, which enables a higher level of storage consolidation and simplifies their operation. Specifications of all current models of the ETERNUS DX family, taking into account the updated models, can be found in the table below:
In conclusion, we note that any model can be ordered with unified functionality or simply as a block storage system with the possibility of upgrading in the future. In this case, all Fujitsu storage systems supplied to the markets of Russia and the CIS countries are manufactured in Germany and have a three-year factory warranty, Fujitsu offers a wide range of service programs for them, including 24x7 programs and a fixed recovery time.
The new line of Fujitsu ETERNUS DX S3 is truly business oriented. Its development was made based on the real requirements of the business to such products. And only then, on the basis of these requirements, were the appropriate technologies and solutions selected or developed. In fact, it can be said that it is the business requirements that formed the basis of the design specification for the design of the updated family.
')
Thus, for the first time in the market, a new interface has been used for connecting SAS3 disks, providing data rates up to 12 Gb / s, which allows up to two times faster access to data placed on disks.
The next important change is the flexible and fine-tuning of the system of priorities and automatic quality of service (QoS), which allow, depending on the business objectives, to provide guaranteed levels of service for various processes. So a new generation of ETERNUS DX family allows you to allocate a certain amount of cache memory, a certain bandwidth, to guarantee the required number of input / output operations (IOPS) for a particular application, for a particular group of users, or for a particular logical volume. In other words, you just need to determine the priorities and response time for the tasks and applications used, depending on their importance to the business, ETERNUS DX S3 will do the rest. Also, based on the results of the work, billing of the use of resources by certain tasks can be done, which may be applicable, for example, to adjust the established priorities or to correlate the costs of operating the system to a particular project.
The main idea of ​​a single storage family from one vendor is to have a flexible architecture to support growing data volumes. What does this mean in reality in our case? First, the ETERNUS DX S3 offers unique scalability within a single model. Secondly, it supports scaling by data-in-place transition to a higher family model without loss of data and the need for a very short time.
For example, in the new family there are two entry-level models: DX100 and DX200. At the same time, the first can be turned into the second by simply replacing the RAID controllers.
It should also be noted that the new family of new high-density disk shelves, in which you can install up to 60 disks.
And also, as in the previous generation of storage systems (ETERNUS DX S2), there were options available for additional disk shelves for 12 3.5 ”disks and 24 for 2.5” disks.
In addition, due to the use of the new SAS3 interface in new models, the limitation on the number of SSDs used has been removed, and the unified storage management interface allows you to switch from low-end models to older ones without retraining personnel.
All models of this family are the so-called Real Unified Storage. This means that they can be used as SAN and as NAS-storage simultaneously. In this case, in systems that use replication or increase in reliability, the same licenses will work for both block (SAN) and file (NAS) access. Possible integration options for the considered storage systems in the IT infrastructure of enterprises via Fiber Channel, FCoE, iSCSI, SAS, Infiniband, 1 and 10 Gb Ethernet CIFS / NFS. It is also expected that access to DSS via HTTP and FTP protocols will appear in the near future.
The ETERNUS DX S3 family uses the VxWorks real-time operating system developed by Wind River Systems. Its effectiveness is confirmed by the fact that it is used in the NASA space program, in particular, under its control , Curiosity was landed on the surface of Mars. And the fact that Wind River is a division of Intel, whose multi-core Intel Xeon processors are used in the storage itself, speaks about the quality optimization for the processors used and allows for achieving the highest possible performance of the system as a whole.
At the very beginning it was already said that the update that occurred could be called revolutionary. To ensure that this statement is not unsubstantiated, a table is presented below with various coefficients demonstrating the changes that have occurred in comparison with the corresponding storage systems of the previous generation Fujitsu ETERNUS DX family (S2):
The twofold increase in bus speed is explained by the transition from SAS2 to SAS3, which was noted earlier. The 2nd level cache, the so-called Extreme Cache, can be delivered only to mid-level systems by installing PCIe SSD drives, the maximum cache size is 5.6 TB. The use of Extreme Cache will reduce the number of used hard drives to 80%, provided that their number was due to the need to obtain a specified number of IOPS. So, if the necessary data will be in Extreme Cache in 30% of cases, the specified 150,000 IOPS for reading can be obtained on 64 HDD (+ 4 SSD Extreme Cache), while to achieve the same value on hard drives alone 360 HDD.
If this amount of disk cache is not enough, with the next version of the controller microcode, it will be possible to allocate a part of the SSD to the cache directly from the disk shelves. As for IOPS, the comparison of current ETERNUS DX models for this indicator can be seen below:
Moreover, these figures do not necessarily show the maximum possible result for a particular storage system. For example, at the end of December, ETERNUS DX200 S3 with just 14 SSD disks showed a result of more than 200k IOPS .
Due to the use of high-density disk shelves and more efficient use of resources, we are trying to increase the utilization rate of the average disk array from the current standard for the industry 40-50% to 80-90%.
Bidirectional connection of disk shelves ETERNUS DX S3 even in case of complete failure or the absence of one disk shelf ensures the availability of the rest, increasing the reliability of storage. Schematically, such a connection can be represented as follows:
where CE is Controller Enclosure, DE is Disk Enclosure. As can be seen from the diagram, if a unidirectional connection is used, the disk shelves following the failed one would also be inaccessible. Bidirectional connectivity solves this problem.
Another interesting solution that will be available in June of this year is the Fast Rebuild technology developed by Fujitsu. It is designed to solve the bottleneck problem that occurs if one of the disks fails. In this case, the disk on which data recovery is performed becomes “super hot”, since until the completion of the rebuild, data will be written to it at once from all disks of the array. Usually the distribution of I / O operations in this case is 1 to 16, i.e. at 16 I / O, which will serve the hosts, only one will serve the process of restoring the array. Thus, the rebuild process when using large disks (2TB and above) can last a week or even more.
It is to solve this problem and technology was developed Fast Rebuild. Its essence lies in the fact that free backup blocks are initially placed on each of the disks in the array (Reserved - in the diagram below). Thus, data recovery is performed immediately to all disks in these areas, which solves the bottleneck problem and speeds up the recovery procedure at times.
At the same time, it also remains possible to change the priority of a rebuild, and its change can be configured according to a specific schedule. For example, so that the priority of a rebild is increased during off-hours, significantly speeding up the recovery process.
Summing up the above, we note that all Fujitsu ETERNUS DX S3 models of the primary and secondary levels have a unified functionality and control environment. Block and file access to the data is produced within the same storage controller, which enables a higher level of storage consolidation and simplifies their operation. Specifications of all current models of the ETERNUS DX family, taking into account the updated models, can be found in the table below:
In conclusion, we note that any model can be ordered with unified functionality or simply as a block storage system with the possibility of upgrading in the future. In this case, all Fujitsu storage systems supplied to the markets of Russia and the CIS countries are manufactured in Germany and have a three-year factory warranty, Fujitsu offers a wide range of service programs for them, including 24x7 programs and a fixed recovery time.
Source: https://habr.com/ru/post/208852/
All Articles