Use an optical fiber disk array for storage sharing ()

Disk arrays are increasingly used in various application systems. Initially, they are simply used as external storage devices attached to a host or server, it is mainly used to expand the permanent storage space of a single host or server and is generally directly connected to the host through SCSI or other interfaces. Later, with the storage network technology, especially the Fiber Channel) with the development of technology, the disk array is connected to the storage area network (SAN) through the fiber channel interface, providing shared storage space for multiple hosts.

At present, on the one hand, people are dedicated to developing more interface technologies (such as iSCSI and InfiniBand) to connect disk arrays to lower-cost storage networks (such as IP networks ), or a higher-performance, more fully functional storage network (such as InfiniBand Network); on the other hand, we are committed to improving the utilization of disk arrays through storage virtualization technology and global file system technology.

In fact, the definition of disk arrays is not accurate. According to the definition of snia, disk array) it is a series of disks that are combined by a set of control software on one or more accessible disk subsystems; the control software provides the storage space of these disks to the host in the form of one or more Virtual Disks. the control software running on the controller is generally referred to as firmware or Microcode ); A volume manager is usually run on a host. The disk array subsystem is usually called a disk array, that is, a disk subsystem with control software that can organize its disks. In subsequent discussions, we will still use the familiar disk array term to replace the obscure disk array sub-system unified term.

What is an optical fiber disk array? This disk array uses fiber channel technology. The use of fiber channel technology has two meanings: one layer refers to the external, that is, the use of fiber channel interface connection mode for the host; the other layer refers to the use of fiber channel technology to connect internal disks. In general, an optical fiber disk array refers to the latter. Initially, when an optical fiber disk array was launched, storage interfaces such as SCSI and SSA were often used internally, and external storage interfaces were the Optical Fiber Channel interfaces.

Nowadays, more and more fiber-optic disk arrays are gradually developing towards both internal and external Fiber Channel interfaces. This disk array is discussed here. For internal use of IDE, SCSI, SSA and other interface technologies, external use of fiber channel technology, or internal use of fiber channel technology, external Disk Arrays Using SCSI and other interface technologies (although this is against common sense, but this disk array does exist), although it is also an optical disk array, it is not covered in this article.


Composition of an optical fiber disk array


From the definition of an optical fiber disk array, we can see that the hardware structure should be composed of a bunch of disks, controllers, and internal and external interfaces. This structure is also common for mid-and low-end Optical Fiber Disk Arrays: it consists of one or more Cabinet cabinets, two array controllers, array backplane, several power supplies, fans, and other hardware components for storing a large number of disks. The most important component is the array controller and Cabinet. The controller can manage the entire array through its built-in control software. Generally, the interface of the array to the host is on the array controller. Generally, each controller has at least one host interface, and some controllers provide more host interfaces. These host interfaces can be directly connected to the host or through an optical fiber switch. In addition, various management interfaces (serial port, Ethernet port, etc.) are also on the controller.

The two controllers are mainly used in terms of high availability, improved performance, and load balancing. Many Arrays can switch between these two controllers to prevent single point of failure of controllers, connection cables, network devices (such as fiber channel switches and hubs), and host hbas. Some Arrays can achieve multi-channel data access and load balancing between channels through the host or array software.

It can be said that the array controller is the core of the mid-and low-end disk array, which is equivalent to the PC motherboard, memory and CPU. The cabinet where the hard disk is placed is the place where the array actually stores data, which is equivalent to a PC hard disk. The main characteristics of the Cabinet is that the internal usually at least adopt redundant double FC-AL arbitration ring road structure, the internal hard disk is actually connected to the two arbitration ring at the same time.

The mid-range disk array supports more loops, up to 4, 8, and 16. The main purpose of this multi-redundancy arbitration ring structure is to ensure high availability and prevent the failure of the entire array due to the failure of a single line or interface. In addition, each loop uses the bypass technology to prevent the impact of hard disk access and hard disk failure on loop communication. The structure of the high-end Optical Fiber disk array is similar, but it is also unique. For example, EMC's DMX structure and HDS's hi-star switched architecture are designed for high-end disk arrays to provide higher performance, reliability, availability, and scalability, and more advanced functions (such as support for business continuity ).


High performance and high availability


The structure of the optical fiber disk array shows that its most prominent advantage is storage sharing. Unlike other arrays, such as SCSI arrays, an optical fiber disk array can be connected to a storage area network. Multiple hosts can access one or more optical fiber disk arrays at the same time through the storage area network, this provides the most flexible hardware and network platform for centralized and shared storage.

Another advantage of an optical fiber disk array is its high availability. Optical fiber disk arrays not only provide various high availability functions supported by general disk arrays, such as raid support, hot-Spare hard disks, raid automatic reconstruction, background online reconstruction, online raid resizing, hot swapping of hard disks, support for concurrent I/O and command queues, disk array configuration backup, cache battery protection, automatic hard disk fault detection, etc, the dual-controller, multi-redundancy loop, and multi-host interface redundancy configurations ensure the availability of the Local Machine and prevent the impact of other device faults on data access in the storage area.

After storage sharing and high availability, we also need to mention the high performance of the disk array. Fiber Channel supports higher performance than traditional storage technologies such as SCSI. Currently, the optical fiber disk array supports full-duplex read/write of 200 Mb/s for both internal and external use. In the near future, 1 Gbit/s products will also be available.

In addition, another advantage of an optical fiber disk array is its high scalability. On the one hand, for the same disk array, due to the arbitration ring structure, theoretically, up to 126 hard disks can be connected to one ring, this is much higher than the capacity of up to 15 hard disks on the SCSI bus. You can also increase the number of supported hard disks by adding the number of loops supported by the same disk array, in a storage network composed of optical fiber channels, because the optical fiber disk array can be shared, when a host accesses a hard disk installed on an array and the capacity cannot be expanded, you can expand the storage space by sharing the storage space on the other array to the host.


Suitable for San


Based on its advantages and features, fiber optic disk arrays are mainly used in industries or application environments that require high data sharing, high availability, high reliability, high performance, and high scalability. For domestic users, it is necessary to use a shared disk array to store data in data centers of key business departments such as finance, telecom, electric power, taxation, chemical, and metallurgy, it can meet the storage requirements of these industries. Industry Data Centers that require a large volume of data, such as media, books and archives, scientific and technological research, and monitoring. the use of fiber-optic disk arrays can meet the requirements of large-capacity storage and continuous expansion.

Note that when selecting an optical fiber disk array, San is generally selected as the basic architecture of the entire IT Information System. In the SAN architecture, a storage network consists of servers or hosts, optical fiber switches, optical fiber disk arrays, and optical fiber tape libraries, clients and servers with low storage requirements communicate with servers in the storage network through common IP networks.

In addition, in high-end NAS applications, the NAS head is often used as the control end of the NAS server, and the optical fiber disk array is directly connected to the NAS head through San or, it provides a high-performance, large-capacity, high-availability storage backend.


Typical SAN architecture with an optical fiber disk array


Product purchase

Perfect


What kind of optical fiber disk array can meet your own needs? In addition to the price factors, we mainly consider the following 10 aspects.

1. storage sharing capability

If the SAN architecture is used, if many servers or hosts share an optical fiber disk array, consider the maximum number of servers or hosts supported by the array. Each vendor provides this parameter, but the number of supported disk arrays varies.

2. availability and reliability

If you want to design a scheme with high availability requirements, consider the availability of the array. The availability of an optical fiber disk array mainly depends on the redundancy switching capability of the controller, including the number of host interfaces. The more host interfaces are provided, the more connection paths available for redundant connections; many arrays require dedicated path switching software to support path switching. others can be implemented using third-party software similar to veritas dmp, in addition, the selection of arrays must be considered based on the actual situation for different environments with different levels of support; switching time, I/O read/write can have a certain delay, generally not shorter than the switching time, otherwise, data will be lost.

3. interoperability and compatibility

When selecting a disk array, check the interoperability. Generally, each supplier will provide a list of compatibility of its arrays to determine whether it supports existing or possibly used network storage devices, including OS, switches, HbA, and SFP.

4. Capacity and scalability

The initial disk capacity, capacity expansion method, and maximum scalable capacity of each disk array are different, and the prices for different configurations are different. select the most suitable disk array.

5. Performance

Currently, all disk arrays support 200 Mbit/s full-duplex bandwidth interfaces, but the actual read/write capabilities are different. Some Arrays can use host interfaces and controller load balancing to provide very high read/write bandwidth and iops. If the application requires high performance, consider using this disk array.

6. ease of use

The simpler and more user-friendly the disk array configuration is, the more difficult the configuration is to see errors, the higher the priority should be given.

7. manageability

Generally, disk arrays can be managed in two ways: in-band and out-of-band. The two methods have their own advantages and disadvantages: in-band management can obtain the topology of the entire storage network structure and does not require the support of other networks (such as Ethernet) connections, however, if the network fails (for example, the host interface of the array fails), the management cannot be performed. The management of out-of-band needs to connect to other networks, however, the out-of-band management is more reliable because it cannot be managed due to a fault such as an array host interface failure, in addition, when the disk array encounters a problem, you can change the configuration, hot start, and other operations to restore its normal operation status. We recommend that you select a disk array that is supported by both management methods.

8. business continuity capability

Many mid-and high-end Optical Fiber disk arrays provide the ability to ensure business continuity, but this is generally achieved through mirroring or copying between two or more arrays, therefore, the cost is often greatly increased.

9. Security

The security of the optical fiber disk array is mainly manifested in whether the host partition or LUN Masking function is supported. You can set which Lun the host can only access, or which host port the host can only access. This can prevent data from being accessed by irrelevant hosts and reduce mutual interference of intra-san communication.

10. Support for basic functions of the array

 

 

 

The explosive growth of information in today's world not only brings greater impetus to the development of science and technology, but also brings huge challenges to enterprises' data storage. However, as the most critical part of an enterprise's information storage system-disk arrays, many may not be clear.

The disk array technology was born in 1987 and proposed by the University of California, Berkeley. The core design concept of this technology is RAID technology. The original name is "Redundant Array of Inexpensive Disk". It was initially developed to combine small cheap disks to replace large expensive disks to reduce the cost of mass data storage. At the same time, we also hope that data will not be lost when a single disk fails by means of redundant information. Therefore, we have developed raid data protection technologies of different levels and are committed to improving the data access speed. This name was later changed to "Redundant Array of independent disk", but it is still called "raid ".

After years of development, the value of data in enterprises is getting higher and higher, and the disk array carrying the data is becoming more and more valued by users. From the market distribution, we can see that the proportion of storage and server is increasing year by year. The powerful requirements of users also create huge business opportunities for storage system vendors. At present, there are not only a variety of products provided by old manufacturers, but also a variety of new systems pushed by start-ups. Naturally, the current disk array on the market is also a splendid scene. When users have many choices, they are also confused about their choices. Therefore, we will briefly analyze the differences between the current disk arrays from the perspective of architecture, and hope to give users a reference when selecting disk arrays.

Currently, disk interfaces include IDE, SATA, SCSI, SAS, and FC. The IDE interface disk is being replaced by the SATA interface hard disk and will gradually exit the historical stage. The two are mainly used for desktops. The SAS interface disk is also gradually eliminating the SCSI interface, it will soon occupy the low-end market of enterprise applications, while the FC (Fiber Channel, optical fiber) interface hard disk was born specifically for High-reliability, high-availability, high-performance enterprise storage applications, not only is the interface fast, but it also supports dual-port access and is strictly controlled by the production process, with good reliability. Due to these inherent advantages, the FC interface hard disk occupies an absolute advantage among enterprise users, especially key data storage applications, and is also the preferred disk for high-end storage applications.

 

 

 

We have seen a lot of disk arrays based on SATA and SCSI interfaces. Here we will not repeat them here, especially the disk arrays with all optical fiber interfaces. Fiber Optic Disk Arrays can be further subdivided into three categories: jbod disk arrays, dual-controller disk arrays, and multi-controller disk arrays.

Strictly speaking, jbod cannot be called an array ". Jbod is the abbreviation of just bundle of disk, meaning it is just a combination of disks. Such a "disk array" is also called a silly disk array, because jbod has neither a controller nor a cache, nor any means to improve performance and security between disks. Each disk receives data access from the host independently. To implement raid-level protection, the host not only needs to perform disk read/write operations, but also performs raid algorithm processing. The usage of Host resources is high, seriously affecting the overall system performance.

Therefore, when an optical fiber disk array is used, a disk array with a smart disk controller is generally used. The disk controller is a control unit between the host and the disk. It is configured with a processor specially optimized for I/O and a certain number of caches. The CPU and cache on the controller work together to perform operations on I/O requests from the host system and manage raid on the disk array. Compared with the jbod disk array, the Controller disk array releases a large number of host resources. The I/O requests from the host are accepted and processed by the Controller, and the cache on the array is used as the I/O buffer pool, this greatly improves the read/write response speed of the disk array and the performance of the disk array. Because the optical fiber disk has a dual port, the general Optical Fiber disk arrays use dual controllers to make full use of the high availability of the optical fiber disk. Whether configured as active-active or active-standby, the two controllers can provide users with high availability, and most of them support hot swapping, which can achieve simple no single point of failure, uninterrupted services provided to users.

Some of its high-end products can also run Disk Array-based storage software. Therefore, it provides a comprehensive disk array-based solution.

In the current storage market, this type of disk array has a wide variety and a large number, and there is also a huge gap in quality and performance, and the price span is also large. Its representative products include the ibm ds series, hp eva series, EMC clariion series, and HDS thunder 95 series. LSI has also made significant achievements in this category of disk arrays. In particular, the IBM S-4000 series, stk d series, and sgi tp series are all eseries array controllers from oem lsi.

In terms of architecture, such products are mid-range products, but the most striking of them is the IBM shark series products. The shark series of IBM products are typical dual controller structure products, its high-end model DS-8300 products each controller is 4 CPU p570 minicomputer, Dual Controller maximum configuration of 8 CPUs. But the DS-8000 series products have many characteristics of high-end products, such as host port up to 128 2 gb fc, disk interface up to 64, the cache capacity can also reach 256 GB, these features make it comparable to multi-controller storage systems, so the DS-8000 series is also the main weapon IBM participates in the market competition for high-end storage products, in addition, it has obvious price advantages compared with its competitor's high-end products.

 

 

 

 

 

Category 3: Multi-controller Disk Arrays: dual-controller disk arrays can only be configured with two controllers. They cannot be configured with more controllers in the same disk array, which limits their data processing capabilities to a certain extent. The multi-controller disk array came into being. Its architecture is generally divided into three layers:

Channel Controller: manages the I/O between the host and cache and runs a storage-based software solution.

Global cache Controller: a huge non-volatile cache, which is one of the foundations for superior system performance.

Disk controller: Manages I/O between cache and physical disk groups and runs storage-based software solutions.

In actual applications, controllers at each layer must be configured in pairs at least to provide full redundancy to ensure no spof. You can also configure multiple pairs to multiply the performance. For example, EMC's DMX-3 can be configured with up to eight front-end controllers (Channel ctor), eight cache controllers (memory Director), eight back-end controllers (Disk Director ), the total number of processors in the system can reach 130. Multi-level controllers work collaboratively to achieve optimal scalability for the overall system performance. More importantly, it provides many unique storage software solutions based on the multi-controller architecture. This product is the first choice for large key business data centers.

Currently, only EMC and HDS are the manufacturers that have mastered the technology of high-end multi-controller architecture products. From the perspective of architecture, there is no difference between the original EMC and HDS multi-controller disk array products, both of which are based on the bus structure. Later, HDS developed the lightning9000/USP series based on the fully switched architecture, and separated the control flow from the data stream internally, improving the internal transmission bandwidth. Later, EMC introduced the symmetrix DMX system based on the direct matrix structure, which further increased the bandwidth of the internal bus based on the low-latency feature. However, in any case, these systems are high-end storage arrays with high reliability and availability, and they have achieved almost perfect results in these two aspects, it has been widely recognized and applied in fields such as telecommunications and banking.

Many server manufacturers also sell disk arrays. Because of its extensive sales channels and powerful market operation capabilities, it also has a high market share. So far, disk Arrays owned by all host manufacturers are still at the dual-controller disk array level. While HP and Sun are also selling disk arrays with multiple controller architectures, they are all oem hds products.

The concepts of information lifecycle management, virtual storage, and storage resource management are closely related to Hierarchical Storage, which provides a reference for users to carry out comprehensive data management. Users can adopt appropriate disk arrays based on the value of information to develop corresponding storage solutions. This article is just a simple division of the optical fiber disk array from the system structure for your reference.

Use an optical fiber disk array for storage sharing ()