OCZ ZD-XL SQL Accelerator. Enlarge your SQL
Bore the queen in the night
Not the son, not the daughter;
Not a little mouse, not a frog,
And an unknowable little animal.
Alexander Sergeevich Pushkin, if he lived in our time and worked in the IT industry, it would have announced the new version of the device with the beautiful name OCZ ZD-XL SQL Accelerator . But we are technical people, and we understand that if miracles of technology are created, it means that someone needs it. The software and hardware complex from OCZ, first of all, is designed to make life easier for users of Microsoft SQL servers (2014, 2012 and 2008 R2), but not only for them. How this wonderful unit works will be dealt with under the cut.
In fact, the engineers at OCZ Storage Soultions have done a tremendous job in bringing the product to market. You need to understand that the niche in which it is aimed is potentially huge, and getting into the client is 100%.
')
At the hardware level, this is the OCZ Z-Drive 4500 SSD (known to users for a long time) with minor changes, and at the software level it is a host application that can simultaneously accelerate volumes or separate SQL databases located on block disk storage systems by caching all or part of the SQL database directly on the server side, moreover, in close proximity to the CPU and RAM, on the PCIe bus.
Let's dive deep. The iron part, as I said, is similar to the PCIe SSD OCZ Z-Drive 4500 - capacitors are soldered on this drive to prevent data loss during an emergency power off. The drive uses 8 SandForce SF-2500 series controllers. The minimum NAND MLC-memory resource is 680 TB for the 800GB model, 1300 TB for the 1600GB model and 2500TB for the 3200GB model.
By the way, when buying such units, OCZ, in addition to the warranty, offers the customer a dedicated support engineer for the entire warranty period.
What is interesting is proposed to database administrators? After installing the drive and software, a special assistant will offer to divide the drive into two parts - one for that and the other for the cache. This separation is possible thanks to two visualization processors located on the drive and controlling the distribution of NAND, the workload of each NAND controller. This allows you to place on the TempDB volume, and sometimes logs, which significantly optimizes the performance of the SQL server, reduces traffic on the SAN network and reduces the load on the disk storage systems. Also, the configuration will provide a choice of how to optimize caching policies - between OLAP (for complex analytical applications) and OLTP (for small discrete transactions).
The software allows you to set analysis parameters and intervals for each volume or base under the ZD-XL acceleration, which will track activity for a given period of time, and accumulate statistics on the “hottest” data, which can later be loaded into the cache by requirements for specialized, regular operations, such as data analysis or the construction of periodic reports. The function of pre-filling the cache based on statistics or the complete transfer of a volume or database into a cache allows you to quickly bring system performance to a maximum after a reboot or before performing specific database operations. Script support allows you to fully automate these processes.
The administrator can place a whole volume or database of comparable size in the cache, and then the increase is very significant and there is no need to analyze the volume in advance. The procedure is very useful after server reboots, since after that the synchronization of data in the cache and on the storage system can be broken.
The corporate market is significantly different from the consumer market in that it is very difficult to deceive the corporate client. Any statements about high performance are checked significantly more carefully than end users.
Let's talk about performance. Probably the easiest option is to compare a RAID 6 disk array of 16 SAS disks with a 10K rotational speed and two options for using ZD-XL SQL — both as a cache and as a flash.
The picture is as follows:
When working with a small number of users - the difference will be barely perceptible, but as soon as the border of 5000 users is passed, the gap begins to grow sharply. Server performance (transactions per second) when working with a database of 20,000 users grows up to 2.5 times when using OCZ ZD-XL SQL as a cache, and up to 9 times when working as a data store.
The average waiting time when using the same number of users when using the cache is improved almost 3 times and a huge number of times (the difference reaches 6 orders of magnitude) when the database is fully stored in the device’s memory
And the above are considered not the worst options, where the storage systems have a direct connection. In cases where consolidation is used, the SAN is actively used, the total storage system is overloaded, the SAN network itself is overloaded, the SQL servers can benefit from ZD-XL up to 20 times, and the rest of the network receive a breath of “fresh air”, since significantly reduced traffic from SQL servers.
The advantage of this solution is that you can use both scenarios at the same time, and the cache and flash are a serious difference from other similar devices.
Pure performance for connoisseurs of beauty:
» At the same time reading 4K blocks in 16 streams and a depth of 16 read requests, the indicators exceed 171K IOPS, and the write exceeds 10K IOPS.
" Sequential read / write blocks of 128K is more than 2150 and 1120 gigabytes per second, respectively.
Why am I telling all this? Yes, in fact, I brag about what OCZ can do - this is a solution for a wide market segment right out of the box, not requiring a radical reconfiguration or reconfiguration of the existing system, with an instant result that maximizes the company's experience in developing simple to use, however, complex in essence and quality products.
A truly unique combination of software and hardware is already working in real life and will only improve its performance over time. Small drives (800 gigabytes) from the ZD-XL series should be primarily used as caching, but the big brothers 1.6 and 3.2 terabytes, of course, better serve as a fast storage for medium-sized databases, while speeding up the cache in parallel bulk network resources.
ZD-XL can accelerate not only MS SQL databases, but also MS Exchange servers and virtually any OLTP-like loads. Simply, maximum optimization of caching policies is done for MS SQL.
Given the fact that the deployment does not require special skills and is easily configured, ZD-XL is an excellent option to speed up existing database servers that run on classic hard drives. The ability to have both a volume and a cache on one device will be especially useful for speeding up 1C servers.
For more information about the product can be on the official website . And under this link you can order white paper .
Our previous publications:
» OCZ plans to release new SSDs for 2015
» SSD SATA vs. SSD PCI-E | Part One: Theoretical
» OCZ. We returned
Thanks for attention!
Not the son, not the daughter;
Not a little mouse, not a frog,
And an unknowable little animal.
Alexander Sergeevich Pushkin, if he lived in our time and worked in the IT industry, it would have announced the new version of the device with the beautiful name OCZ ZD-XL SQL Accelerator . But we are technical people, and we understand that if miracles of technology are created, it means that someone needs it. The software and hardware complex from OCZ, first of all, is designed to make life easier for users of Microsoft SQL servers (2014, 2012 and 2008 R2), but not only for them. How this wonderful unit works will be dealt with under the cut.
In fact, the engineers at OCZ Storage Soultions have done a tremendous job in bringing the product to market. You need to understand that the niche in which it is aimed is potentially huge, and getting into the client is 100%.
')
At the hardware level, this is the OCZ Z-Drive 4500 SSD (known to users for a long time) with minor changes, and at the software level it is a host application that can simultaneously accelerate volumes or separate SQL databases located on block disk storage systems by caching all or part of the SQL database directly on the server side, moreover, in close proximity to the CPU and RAM, on the PCIe bus.
Let's dive deep. The iron part, as I said, is similar to the PCIe SSD OCZ Z-Drive 4500 - capacitors are soldered on this drive to prevent data loss during an emergency power off. The drive uses 8 SandForce SF-2500 series controllers. The minimum NAND MLC-memory resource is 680 TB for the 800GB model, 1300 TB for the 1600GB model and 2500TB for the 3200GB model.
By the way, when buying such units, OCZ, in addition to the warranty, offers the customer a dedicated support engineer for the entire warranty period.
What is interesting is proposed to database administrators? After installing the drive and software, a special assistant will offer to divide the drive into two parts - one for that and the other for the cache. This separation is possible thanks to two visualization processors located on the drive and controlling the distribution of NAND, the workload of each NAND controller. This allows you to place on the TempDB volume, and sometimes logs, which significantly optimizes the performance of the SQL server, reduces traffic on the SAN network and reduces the load on the disk storage systems. Also, the configuration will provide a choice of how to optimize caching policies - between OLAP (for complex analytical applications) and OLTP (for small discrete transactions).
The software allows you to set analysis parameters and intervals for each volume or base under the ZD-XL acceleration, which will track activity for a given period of time, and accumulate statistics on the “hottest” data, which can later be loaded into the cache by requirements for specialized, regular operations, such as data analysis or the construction of periodic reports. The function of pre-filling the cache based on statistics or the complete transfer of a volume or database into a cache allows you to quickly bring system performance to a maximum after a reboot or before performing specific database operations. Script support allows you to fully automate these processes.
The administrator can place a whole volume or database of comparable size in the cache, and then the increase is very significant and there is no need to analyze the volume in advance. The procedure is very useful after server reboots, since after that the synchronization of data in the cache and on the storage system can be broken.
The corporate market is significantly different from the consumer market in that it is very difficult to deceive the corporate client. Any statements about high performance are checked significantly more carefully than end users.
Let's talk about performance. Probably the easiest option is to compare a RAID 6 disk array of 16 SAS disks with a 10K rotational speed and two options for using ZD-XL SQL — both as a cache and as a flash.
The picture is as follows:
When working with a small number of users - the difference will be barely perceptible, but as soon as the border of 5000 users is passed, the gap begins to grow sharply. Server performance (transactions per second) when working with a database of 20,000 users grows up to 2.5 times when using OCZ ZD-XL SQL as a cache, and up to 9 times when working as a data store.
The average waiting time when using the same number of users when using the cache is improved almost 3 times and a huge number of times (the difference reaches 6 orders of magnitude) when the database is fully stored in the device’s memory
And the above are considered not the worst options, where the storage systems have a direct connection. In cases where consolidation is used, the SAN is actively used, the total storage system is overloaded, the SAN network itself is overloaded, the SQL servers can benefit from ZD-XL up to 20 times, and the rest of the network receive a breath of “fresh air”, since significantly reduced traffic from SQL servers.
The advantage of this solution is that you can use both scenarios at the same time, and the cache and flash are a serious difference from other similar devices.
Pure performance for connoisseurs of beauty:
» At the same time reading 4K blocks in 16 streams and a depth of 16 read requests, the indicators exceed 171K IOPS, and the write exceeds 10K IOPS.
" Sequential read / write blocks of 128K is more than 2150 and 1120 gigabytes per second, respectively.
Why am I telling all this? Yes, in fact, I brag about what OCZ can do - this is a solution for a wide market segment right out of the box, not requiring a radical reconfiguration or reconfiguration of the existing system, with an instant result that maximizes the company's experience in developing simple to use, however, complex in essence and quality products.
A truly unique combination of software and hardware is already working in real life and will only improve its performance over time. Small drives (800 gigabytes) from the ZD-XL series should be primarily used as caching, but the big brothers 1.6 and 3.2 terabytes, of course, better serve as a fast storage for medium-sized databases, while speeding up the cache in parallel bulk network resources.
ZD-XL can accelerate not only MS SQL databases, but also MS Exchange servers and virtually any OLTP-like loads. Simply, maximum optimization of caching policies is done for MS SQL.
Given the fact that the deployment does not require special skills and is easily configured, ZD-XL is an excellent option to speed up existing database servers that run on classic hard drives. The ability to have both a volume and a cache on one device will be especially useful for speeding up 1C servers.
For more information about the product can be on the official website . And under this link you can order white paper .
Our previous publications:
» OCZ plans to release new SSDs for 2015
» SSD SATA vs. SSD PCI-E | Part One: Theoretical
» OCZ. We returned
Thanks for attention!
Source: https://habr.com/ru/post/376223/
All Articles