1. Select 32-bit or 64-bit RISC CPU or normal Intel 586 CPU?
SCSI is developed in the following order: SCSI2 (narrowband, 8-bit, 10MB / S) → SCSI3 (Broadband, 16-bit, 20MB / S) → Ultra Wide (16-bit, 40MB / s) → Ultra2 (Ultra Ultra Wide 80MB / s) → Ultra3 (Ultra Ultra Wide, 160MB / s) → Ultra3 (Ultra Ultra Wide, 320MB / s). When using the Ultra Wide SCSI disk array, the requirements for the CPU do not need to be too fast, because the SCSI itself is not very fast. However, when SCSI is developed to ULTRA2, the requirements for the CPU are very critical, and the general CPU (ie, 586 CPU) must be changed to a high speed RISC CPU.
The structure of the server has been changed from a conventional I / O structure to an I2O structure, which is to reduce the burden on the CPU in the server, separated the I / O load of the system to the server CPU. I2O is a RISC CPU to be responsible for I / O work. The server has been used on the server, and the disk array must also be used to form a bottleneck with the RISC CPU. In addition, we are now mostly 32-bit or 64-bit network operating systems. When the operating system has turned from 32-bit to 64 bits, the CPU on the disk array must be a RISC CPU to meet the requirements. 2. Is there a sequential requirement within the disk array? That is, can the hard disk can be inserted into the array in the original order, and the data still can access it normally? Many people want to certainly think that there should be no order requirements at all, in fact. We have used an array that must be accessed normally according to the original order. Now suppose such a situation, we are ready to clean the hard disk arrays, put all the hard drives together, and don't remember the order, in order to access data, we only have one-stop, and for 8 hard drives For the worst case, I have to try 88 times. The disk array product has now occurred with a function that does not require a hard disk sequence. In order to prevent the above event, the array of order is not required. 3. Is the hardware disk array or a software disk array?
The software disk array refers to the use of a SCSI card with disk, and the hardware disk array refers to an array of backplanes in the array cabinet, which is very different from the software disk array. Hardware disk arrays are a complete disk array system to interface with the system, built-in CPU, with the host and act, all I / O is completed in the disk array, reducing the host's burden, increase system overall performance, SCSI bus master With the DMA channel to accelerate the access and transmission of data. The software disk array is a program that is executed on the host. The maximum disadvantage is greatly increased by an array of SCSI cards, which greatly increases the burden of the host. It is easy to make the system for a large amount of input and output. Obviously, the hardware disk array should be selected as much as possible. 4. Is it an IDE disk array or a SCSI disk array? There is a IDE disk array in the recent market. They are very fast, such as the transmission rate of the enhanced IDE can reach 66MB / s under the PCI bus, the price is much cheaper than the SCSI disk array; and the SCSI Ultra3 rate Close to 160MB / s. However, from the actual application, the IDE disk array is fast than the SCSI disk array; when multiple tasks, the SCSI disk array is much more than the IDE disk array. But the IDE disk array has a fatal disadvantage: no electric heat swapping. This shortcoming makes it destined to the IDE disk array only to non-important occasions. If your application can't stop, you must select the SCSI disk array. 5. Is it a single controller or redundant controller? The disk array is generally connected to the host and disk in a controller, and the integrity of the data is achieved under the fault tolerance function of the disk array. However, the disk array controller will fail, under this situation, data may be lost. In order to solve this problem, two controllers can be connected to each other, and back up each other. However, the connection between the two independent controllers in the chassis means that once the failure must open the chassis controller, you must stop, this is not possible in many applications, so we should choose hot-swappable dual control redundancy. Architecture. Now some disk array new products use the way to get memory and memory mirroring, to ensure that the data is lost in the event of a fault, and after the controller is replaced, the work setting before the fault is automatically recovered, and the workload is dispersed to each other. Controller to achieve load balancing, which can provide high performance and high security that can not be achieved by single controllers. 66. SCSI interface or fiber channel interface?
SCSI's perfect specifications, mature technology and high performance have always attracted small systems, but from the current situation, the fiber channel has formed a market, the double ring can reach 200Mb / s, and the transmission distance is 10km, which can be connected 126 devices. . Fiber Channels use the bus to the network, which is the foundation of the storage network, and its replacement SCSI is a general trend. Therefore, in order to ensure the vitality of the system, the fiber channel interface should be selected. However, the fiber network network is particularly high, approximately 4 to 5 times the SCSI interface network, and from the actual situation, the fiber channel is still a weak place, and the software of the client is high, so in choice When the actual situation should be selected.
The disk array is still a new product for most users, and there are not many users. However, from the current situation, people have more strong demand for large-capacity storage and data security, and the choice of disk arrays is inevitably. This article is only aware of the design of disk arrays from macro aspects. Please carefully Read product manual.
