What theory of program is the basis of virtual storage management system?

The basis of the virtual storage management system is the "locality" theory of the program. Virtual storage technology is based on the principle of program locality, and the principle of program locality is reflected in two aspects: 1. Time locality. Time locality means that after an instruction is executed, it may be executed again soon; 2. Spatial locality means that if a certain storage unit is accessed, the units adjacent to the storage unit may also be accessed quickly.

The operating environment of this tutorial: Windows 7 system, Dell G3 computer.

The basis of the virtual storage management system is the "locality" theory of the program.

The so-called virtual storage is to organically combine memory and external storage to obtain a "memory" with a large capacity. This is called virtual storage. It can be said that the overall performance of the storage network platform will directly affect the normal operation of the entire system.

Virtual storage technology is based on the principle of program locality, which is reflected in two aspects: time locality and space locality. Temporal locality means that after an instruction is executed, it may be executed again soon; spatial locality means that if a certain storage unit is accessed, then the units adjacent to the storage unit may also be accessed soon. .

Classification of virtual storage

There is no unified standard for the development of virtual storage.

• From There are two main methods of virtualized storage topology: symmetric and asymmetric.

  Symmetric virtual storage technology refers to the virtual storage control device, storage software system, and switching device integrated into a whole, embedded in the network data transmission path; asymmetric virtual storage technology refers to the virtual storage control device Independent of the data transmission path.

• In terms of the implementation principle of virtualized storage, there are two ways: data block virtualization and virtual file system.

The details are as follows:

Symmetric virtual storage

Storage control device High Speed ​​Traffic Directors (HSTD) are integrated with the storage pool subsystem Storage Pool to form a SAN Appliance. It can be seen that in this solution, the storage control device HSTD plays a core role in the process of data exchange between the host and the storage pool. The virtual storage process of this solution is as follows: the HSTD embedded storage management system virtualizes the physical hard disks in the storage pool into logical storage units (LUNs), and performs port mapping (specifying which ports a certain LUN can be seen by) , the host side maps each visible storage unit to a drive letter recognized by the operating system. When the host writes data to the SAN Appliance, the user only needs to specify the data writing location as its mapped drive letter (LUN). The data passes through the high-speed parallel port of the HSTD and is first written into the cache. The storage management system in the HSTD automatically Complete the conversion of the target location from LUN to physical hard disk. During this process, the user only sees the virtual logical unit and does not care about the specific physical organizational structure of each LUN. This solution has the following main features:

(1) Using a large-capacity cache to significantly increase the data transmission speed.

Cache is an intermediate medium widely used in storage systems on the I/O path between the host and the storage device. When the host reads data from the storage device, it will read the data connected to the current data storage location into the cache, and keep the data called multiple times in the cache; when the host reads the data, there is a high probability that it can Find the required data from the cache. Read directly from cache. The speed when reading data from the cache is only affected by the propagation speed of electrical signals (equal to the speed of light), so it is much higher than the speed of mechanical rotation of the disk when reading data from the hard disk. When the host writes data to the storage device, it first writes the data into the cache. After the host-side writing operation stops, the data is then written from the cache to the hard disk. This is also faster than directly writing to the hard disk.

(2) Multi-port parallel technology eliminates I/O bottlenecks.

In traditional FC storage devices, there is a fixed relationship between the control port and the logical disk. Access to a hard disk can only be through the controller port that controls it. In a symmetric virtual storage device, the relationship between the storage port of the SAN Appliance and the LUN is virtual, which means that multiple hosts can concurrently access the same LUN through multiple storage ports (up to 8); in Fiber Channel 100MB/bandwidth Under the general premise, the more ports working in parallel, the higher the data bandwidth.

(3) The logical storage unit provides high-speed disk access speed.

In the video application environment, the application reads and writes data in fixed-size data blocks (from 512byte to 1MB). In order to ensure the bandwidth requirements of applications, storage systems are often designed to achieve their best I/O performance when data block sizes of more than 512 bytes are transmitted. In traditional SAN structures, when capacity requirements increase, the only solution is to bind multiple disks (physical or logical) into stripe sets to achieve large-capacity LUNs. In a symmetrical virtual storage system, the host is provided with truly large-capacity, high-performance LUNs instead of poor-performance logical volumes implemented using stripe sets. Compared with stripe sets, Power LUN has many advantages. For example, large I/O blocks will be truly accepted by the storage system, effectively increasing the data transfer speed; and because there is no stripe set processing, the host CPU can relieve a lot of problems. Large burden, improves the performance of the host.

(4) Fault tolerance performance of paired HSTD system.

In a symmetric virtual storage system, HSTD is the only place for data I/O, and the storage pool is the place where data is stored. Since the data in the storage pool has a fault-tolerant mechanism to ensure security, users will naturally wonder whether HSTD has fault-tolerant protection. Like many large-scale storage systems, in mature symmetrical virtual storage systems, HSTDs are configured in pairs, and each pair of HSTDs achieve cache data consistency and mutual communication through the network management service embedded in the SAN Appliance.

(5) Switching equipment can be easily connected on the SAN Appliance to realize a super-large Fabric structured SAN.

Because the system maintains a standard SAN structure and provides technical support for system expansion and interconnection, switching devices can be easily connected on the SAN Appliance to realize a super-large-scale Fabric structured SAN.

Asymmetric virtual storage system

Every host and virtual storage management device in the network is connected to the disk array , in which the data path of the host reaches the disk array through the FC switching device; the virtual storage device performs virtualization operations on the disk arrays connected to the network, virtualizes the LUNs in each storage array into logical strip sets (Strips), and performs virtualization on the network. Each host specifies access rights (writable, readable, prohibited access) to each Strip. When the host wants to access a Strip, it must first access the virtual storage device, read the Strip information and access permissions, and then access the actual data in the Strip through the switching device. During this process, the host will only recognize the logical strip and not the physical hard disk directly. This solution has the following characteristics:

(1) Logically combine the capacities in different physical hard disk arrays to implement virtual stripe sets, and bind multiple array controller ports to improve the performance to a certain extent. the available bandwidth of the system.

(2) When the number of switch ports is sufficient, two virtual storage devices can be installed in one network to achieve redundancy of Strip information and access rights.

However, this solution has the following shortcomings:

(1) This solution is essentially a stripe set-disk array structure. Once a disk array controller in the stripe set is damaged, Or the copper cable or GBIC on the path from the array to the switch will be damaged, which will cause a virtual LUN to go offline. The stripe set itself has no fault tolerance. The damage of a LUN means the loss of data in the entire Strip.

(2) Since the bandwidth improvement of this solution is achieved through array port binding, and the effective bandwidth of ordinary fiber channel array controllers is only about 40MB/S, it needs to reach a bandwidth of several hundred megabytes. This means calling more than a dozen arrays, which will occupy dozens of switch ports. This is unachievable in small and medium-sized networks with only one or two switches.

(3) Since the performance of disk arrays of various brands and models is not exactly the same, if arrays of different brands and models are bound for the purpose of virtualization, a problem will arise: data The speed of each concurrent data stream is different when writing or reading, which means that the original data packet order is disrupted after the transmission is completed, and the system needs to take up time and resources to re-sort the data packets, which will seriously affect the system performance.

Data block virtualization and virtual file system

The data block virtual storage solution focuses on solving conflicts and delays during data transmission. In a large-scale Fabric structure SAN composed of multiple switches, since multiple hosts access storage devices through multiple switch ports, latency and data block conflicts are very serious problems. The data block virtual storage solution uses virtual multi-port parallel technology to provide extremely high bandwidth for multiple clients, minimizing the occurrence of delays and conflicts. In practical applications, the data block virtual storage solution uses a symmetrical Topology is a representation.

The virtual file system storage solution focuses on solving the security mechanism problem of file sharing in large-scale networks. Ensure the security of network files by specifying different access permissions for different sites. In practical applications, virtual file system storage solutions take the form of asymmetric topology.

For more related knowledge, please visit the FAQ column!

The above is the detailed content of What theory of program is the basis of virtual storage management system?. For more information, please follow other related articles on the PHP Chinese website!