Step to the “bright SSD-future”
For several years now, flash memory has entered the market of data storage systems, and not only came, but was fixed on it. However, there is nothing surprising here. It is enough to recall once again the undeniable advantages of flash drives (SSD) over traditional hard drives (HDD). Today we will talk about a completely new storage system in the Fujitsu line - ETERNUS AF, using only SSD.
First of all, it's speed. The response time of modern SSD is 500 times less than that of HDD, which means instant access to data. EMLC corporate-class flash drives are 100 times more reliable than hard drives, they consume significantly less power, and by the number of I / O operations per second (IOPS), flash drives outperform hard drives by even several orders of magnitude. A significant drawback, which until recently slowed down the onset of the “bright SSD-future” on the data storage market, was the high cost of flash memory. But today, businesses need high-performance storage and access solutions. Data must not only be stored securely, but also provide instant access to them. But if the infrastructure of the data center does not meet the high requirements, the business will be idle, lose time, customers and money. First of all, such requirements place on enterprise-level transactional databases and analytical applications, virtualized server environments and workstations, as well as various cloud services. Accordingly, flash-based storage systems require less administration, configuration, and performance tuning from the customer, while providing a higher level of Quality of Services. In a word, they save time resources and money.
As we said above, until recently, the relative high cost of flash storage systems prevented their widespread distribution in the market, but today the cost of SSD drives has come close to the cost of high-speed hard drives, so manufacturers produce models of this class, which can be positioned as systems primary and secondary level. These include the new Fujitsu products - the ETERNUS AF250 and ETERNUS AF650 storage systems, based on Intel Xeon processors. These storage systems were created by the same development team that created the ETERNUS DX disk arrays, which are already well known to customers, including Russian ones. The new line of ETERNUS AF has the same set of capabilities as in DX - creating snapshots, cloning, mirroring, deduplication and compression, integration with VMware solutions, creating fully automatic disaster-resistant systems (Storage Cluster), automatic level of Quality of Services, allows you to work in heterogeneous environments with guaranteed performance, and much more. As for performance comparison, if the classic Fujitsu ETERNUS DX600 DSS with SAS disks in internal tests of the company shows 152 thousand I / O operations per second (IOPS) and shows a delay level of up to 10 milliseconds, the new ETERNUS AF650 gives a multiple increase in speed work - 750 thousand IOPS and 0.2 milliseconds of delay.
')
At the same time, it should not be forgotten that today customers increasingly pay money for IOPS. The developers of Fujitsu managed to achieve such a speed thanks to an improved algorithm for writing data to an SSD carrier, an optimized Garbage Collection procedure that prevents writing any data to a disk at the time of garbage collection, as well as reducing the load on the controllers due to the possibility of writing data immediately on SSD, thereby unloading system memory.
An important parameter of SSD drive reliability is the number of disk overwrites per day, which can be performed for 5 years (DWPD). Many manufacturers of arrays consider exceeding the maximum number of records on SSD carriers as a non-guaranteed case requiring a paid replacement. At the same time, Fujitsu always recognizes the failure of the SSD-drive, in addition to obvious mechanical damage, a warranty case. The current standard or extended warranty for ETERNUS AF storage systems means that all failed SSD drives will be replaced by the manufacturer for free. This allows you to make the cost of maintaining the storage infrastructure more transparent and predictable.
The ETERNUS AF equipment is packaged in the base package with all the necessary software that allows you to manage licenses. Since Fujitsu has long abandoned licensing for terabytes, customers will be able to predict the cost of ownership with a high degree of accuracy.
A few words about the hardware configuration of ETERNUS AF250 and ETERNUS AF650. The junior model supports work from 2 to 24 SSD-disks, is equipped with system memory up to 64 GB, and the eldest - from 2 to 96 SSD-disks and equipped with system memory up to 256 GB. The ETERNUS AF250 model runs under control of a 6-core Intel Xeon processor with a frequency of 2 GHz, and ETERNUS AF650 with a 10-core processor with a frequency of 3 GHz. The maximum amount of SSD-drives that can be installed today in the data storage - 4 TB. But next year, Fujitsu promises to release larger drives.
We have already mentioned the Storage Cluster disaster recovery mechanism above. Let's tell about it in more detail. It allows you to switch between two physical sites in a fully automatic mode, in case of equipment failure on one of them. This feature is already well known to customers of hybrid and disk storage systems ETERNUS. A synchronous replica is built between the two arrays. In case of failure of one of the arrays, the same addresses and WWN ports that worked on the first array automatically rise on the second one. For applications and systems working with this array, such a switch will be absolutely transparent and will happen in just a few seconds. At the request of the customer can be combined in Storage Cluster flash and a regular disk array, although the performance of this solution, of course, will give the combination of two flash arrays. This feature can also be used for normal office purposes when it is necessary to “transfer” data from one array to another.
Storage Cluster Work Pattern
A good way to optimize the use of storage space is deduplication and compression, which are also supported in ETERNUS AF. Deduplication is done on the fly, and the block size is 4K. Simultaneously with deduplication, the compression process also works. Of course, no one has repealed the laws of physics, so the use of these tools can affect a certain decrease in productivity. As for the data decompression procedure, it is much faster than compression. It should also be noted that the most frequently used blocks will most likely end up in system memory, which will significantly speed up both the compression and unpacking processes. The customer can also manually specify for which data or logical partitions deduplication and compression should be applied, and for which not. For example, for encrypted data, deduplication and compression should not be used at all.
The Quality of Services mechanism has long been used in traditional disk storage systems. The easiest way to configure it is to set priority levels for certain logical volumes or data sets. If the customer requires a guaranteed SLA level of response time for any data, then thanks to the automatic reallocation of resources, this can also be realized.
VMware virtual volume support in ETERNUS AF storage can be implemented using VVOL technology. It transfers mechanisms for creating “snapshots” and clones of virtual machines from a server to a disk array. Using VVOL allows you to create a virtual volume on this array in accordance with the policies set by the administrator. The same volume is used in the backup and recovery process of virtual machines. Also VVOL on ETERNUS AF can create a complete clone of the virtual machine. In addition, VVOL provides visualization and monitoring tools for each individual virtual machine.
So, solutions like ETERNUS AF storage systems, on the one hand, increase storage reliability, and on the other hand, they allow to realize high-speed characteristics that were recently available only for the most powerful Hi-End systems. At the same time, a huge number of background operations can be simplified and even automated, which significantly reduces equipment maintenance costs. Finally, such solutions bring the same “bright SSD-future”, which we spoke about at the very beginning, since the cost of one terabyte and the cost of IOPS are already very close to the classic HDD products.
First of all, it's speed. The response time of modern SSD is 500 times less than that of HDD, which means instant access to data. EMLC corporate-class flash drives are 100 times more reliable than hard drives, they consume significantly less power, and by the number of I / O operations per second (IOPS), flash drives outperform hard drives by even several orders of magnitude. A significant drawback, which until recently slowed down the onset of the “bright SSD-future” on the data storage market, was the high cost of flash memory. But today, businesses need high-performance storage and access solutions. Data must not only be stored securely, but also provide instant access to them. But if the infrastructure of the data center does not meet the high requirements, the business will be idle, lose time, customers and money. First of all, such requirements place on enterprise-level transactional databases and analytical applications, virtualized server environments and workstations, as well as various cloud services. Accordingly, flash-based storage systems require less administration, configuration, and performance tuning from the customer, while providing a higher level of Quality of Services. In a word, they save time resources and money.
As we said above, until recently, the relative high cost of flash storage systems prevented their widespread distribution in the market, but today the cost of SSD drives has come close to the cost of high-speed hard drives, so manufacturers produce models of this class, which can be positioned as systems primary and secondary level. These include the new Fujitsu products - the ETERNUS AF250 and ETERNUS AF650 storage systems, based on Intel Xeon processors. These storage systems were created by the same development team that created the ETERNUS DX disk arrays, which are already well known to customers, including Russian ones. The new line of ETERNUS AF has the same set of capabilities as in DX - creating snapshots, cloning, mirroring, deduplication and compression, integration with VMware solutions, creating fully automatic disaster-resistant systems (Storage Cluster), automatic level of Quality of Services, allows you to work in heterogeneous environments with guaranteed performance, and much more. As for performance comparison, if the classic Fujitsu ETERNUS DX600 DSS with SAS disks in internal tests of the company shows 152 thousand I / O operations per second (IOPS) and shows a delay level of up to 10 milliseconds, the new ETERNUS AF650 gives a multiple increase in speed work - 750 thousand IOPS and 0.2 milliseconds of delay.
')
At the same time, it should not be forgotten that today customers increasingly pay money for IOPS. The developers of Fujitsu managed to achieve such a speed thanks to an improved algorithm for writing data to an SSD carrier, an optimized Garbage Collection procedure that prevents writing any data to a disk at the time of garbage collection, as well as reducing the load on the controllers due to the possibility of writing data immediately on SSD, thereby unloading system memory.
An important parameter of SSD drive reliability is the number of disk overwrites per day, which can be performed for 5 years (DWPD). Many manufacturers of arrays consider exceeding the maximum number of records on SSD carriers as a non-guaranteed case requiring a paid replacement. At the same time, Fujitsu always recognizes the failure of the SSD-drive, in addition to obvious mechanical damage, a warranty case. The current standard or extended warranty for ETERNUS AF storage systems means that all failed SSD drives will be replaced by the manufacturer for free. This allows you to make the cost of maintaining the storage infrastructure more transparent and predictable.
The ETERNUS AF equipment is packaged in the base package with all the necessary software that allows you to manage licenses. Since Fujitsu has long abandoned licensing for terabytes, customers will be able to predict the cost of ownership with a high degree of accuracy.
A few words about the hardware configuration of ETERNUS AF250 and ETERNUS AF650. The junior model supports work from 2 to 24 SSD-disks, is equipped with system memory up to 64 GB, and the eldest - from 2 to 96 SSD-disks and equipped with system memory up to 256 GB. The ETERNUS AF250 model runs under control of a 6-core Intel Xeon processor with a frequency of 2 GHz, and ETERNUS AF650 with a 10-core processor with a frequency of 3 GHz. The maximum amount of SSD-drives that can be installed today in the data storage - 4 TB. But next year, Fujitsu promises to release larger drives.
We have already mentioned the Storage Cluster disaster recovery mechanism above. Let's tell about it in more detail. It allows you to switch between two physical sites in a fully automatic mode, in case of equipment failure on one of them. This feature is already well known to customers of hybrid and disk storage systems ETERNUS. A synchronous replica is built between the two arrays. In case of failure of one of the arrays, the same addresses and WWN ports that worked on the first array automatically rise on the second one. For applications and systems working with this array, such a switch will be absolutely transparent and will happen in just a few seconds. At the request of the customer can be combined in Storage Cluster flash and a regular disk array, although the performance of this solution, of course, will give the combination of two flash arrays. This feature can also be used for normal office purposes when it is necessary to “transfer” data from one array to another.
Storage Cluster Work Pattern
A good way to optimize the use of storage space is deduplication and compression, which are also supported in ETERNUS AF. Deduplication is done on the fly, and the block size is 4K. Simultaneously with deduplication, the compression process also works. Of course, no one has repealed the laws of physics, so the use of these tools can affect a certain decrease in productivity. As for the data decompression procedure, it is much faster than compression. It should also be noted that the most frequently used blocks will most likely end up in system memory, which will significantly speed up both the compression and unpacking processes. The customer can also manually specify for which data or logical partitions deduplication and compression should be applied, and for which not. For example, for encrypted data, deduplication and compression should not be used at all.
The Quality of Services mechanism has long been used in traditional disk storage systems. The easiest way to configure it is to set priority levels for certain logical volumes or data sets. If the customer requires a guaranteed SLA level of response time for any data, then thanks to the automatic reallocation of resources, this can also be realized.
VMware virtual volume support in ETERNUS AF storage can be implemented using VVOL technology. It transfers mechanisms for creating “snapshots” and clones of virtual machines from a server to a disk array. Using VVOL allows you to create a virtual volume on this array in accordance with the policies set by the administrator. The same volume is used in the backup and recovery process of virtual machines. Also VVOL on ETERNUS AF can create a complete clone of the virtual machine. In addition, VVOL provides visualization and monitoring tools for each individual virtual machine.
So, solutions like ETERNUS AF storage systems, on the one hand, increase storage reliability, and on the other hand, they allow to realize high-speed characteristics that were recently available only for the most powerful Hi-End systems. At the same time, a huge number of background operations can be simplified and even automated, which significantly reduces equipment maintenance costs. Finally, such solutions bring the same “bright SSD-future”, which we spoke about at the very beginning, since the cost of one terabyte and the cost of IOPS are already very close to the classic HDD products.
Source: https://habr.com/ru/post/315202/
All Articles