1. Select 32-bit or 64-bit RISC CPU or ordinary Intel 586 CPU?
SCSI is developed in the following order: SCSI2 (narrowband, 8-bit, 10MB/s)→SCSI3 (broadband, 16-bit, 20MB/s)→UltraWide (16-bit, 40MB/s)→Ultra2 (UltraUltraWide, 80MB/s) → Ultra3 (Ultra Ultra Wide, 160 MB/s)→ Ultra3 (Ultra Ultra Wide, 320 MB/s). When using UltraWideSCSI disk arrays in the past, the CPU requirements do not need to be too fast, because SCSI itself is not very fast. But when SCSI evolves to Ultra2, the requirement for CPU is very critical, and the general CPU (ie, 586-level CPU) must be changed to a high-speed RISC CPU.
The structure of the server has changed from the traditional I/O structure to the I2O structure. The purpose of the server is to reduce the burden on the CPU in the server and to separate the system I/O from the server CPU load. I2O is a RISC CPU responsible for I/O work. The RISC CPU has been used on the server. Of course, the RISC CPU must also be used on the disk array. In addition, most of the commonly used network operating systems are 32-bit or 64-bit. When the operating system is switched from 32-bit to 64-bit, the CPU on the disk array must be a RISC CPU to meet the requirements.
2. Is there any order requirement for the hard disk in the disk array?
In Other words, can the hard disk be inserted back into the array in the original order and the data can still be accessed normally? Many people take it for granted that there should not be order requirements at all. It is not. We have used an array that must be in the correct order to access data. Now suppose that such a situation, we are ready to clean up the hard disk array, put all the hard drives together, the results can not remember the order, in order to access the data, we only have one test, and for an array of eight hard drives In the worst case, 88 attempts will be required. Now that the disk array product has a function that does not require the hard disk sequence, in order to prevent the occurrence of the above events, an array that does not require an order should be selected.
3. Is it a hardware disk array or a software disk array?
A software disk array refers to the connection of a SCSI card to a disk. A hardware disk array refers to an array with a backplane in an array cabinet. It is very different from a software disk array. Hardware disk array is a complete disk array system connected to the system, built-in CPU, parallel operation with the host, all the I / O is completed in the disk array, reduce the burden on the host, increase the overall system performance, SCSI bus master With DMA channels to accelerate data access and transmission. The software disk array is a program that is executed on the host and forms an array by connecting a SCSI card to the disk. Its biggest drawback is that it greatly increases the burden on the host. For a large number of input and output systems, it is easy to paralyze the system. Obviously, hardware disk arrays should be chosen as much as possible.
4. Is IDE disk array or SCSI disk array?
IDE disk arrays have recently appeared on the market. They are very fast. For example, the enhanced IDE has a transfer rate of 66 MB/s under the PCI bus. The price is much lower than that of SCSI disk arrays. The SCSI Ultra3 speed is close to 160 MB/sec. s. However, from the perspective of practical applications, IDE disk arrays are faster than SCSI disk arrays in single tasks. In multitasking, SCSI disk arrays are much faster than IDE disk arrays.
However, IDE disk arrays have a fatal disadvantage: they cannot be hot swapped. This disadvantage makes IDE disk arrays doomed to be used only in non-essential situations. If your application cannot shut down, be sure to choose a SCSI disk array.
5. Is it a single controller or a redundant controller?
Disk arrays are generally connected to a host and a disk with one controller, and the data integrity is achieved under fault tolerance of the disk array. However, the disk array controller will also fail. In this case, the data may be lost. To solve this problem, two controllers can be connected by cables to back up each other. However, the connection of two independent controllers in the chassis means that the chassis must be opened to switch off the controller in the event of a failure, that is, it must be shut down. This is simply not possible in many applications. Therefore, we should choose hot-swappable dual-control redundancy. Architecture. Some new disk array products now use cache memory and memory mirroring to ensure that no data is lost when a failure occurs. After the controller is replaced, the pre-fault working settings are automatically restored and the workload is distributed to each other. The controller achieves load balancing. This architecture can provide high performance and high security that can not be achieved by a single controller.
6. SCSI interface or fiber channel interface?
SCSI's complete specifications, mature technologies, and high performance have always attracted small systems. However, from the current situation, the Fibre Channel has formed a market with double rings up to 200MB/s, a transmission distance of 10km, and 126 devices. Fibre Channel combines the bus and network into one and is fundamental to the storage network. It has become a general trend to replace SCSI. Therefore, in order to ensure the vitality of the system, the Fibre Channel interface should be selected. However, the cost of the Fibre Channel network is particularly high, which is about 4 to 5 times that of the SCSI interface network. From the actual situation, the Fibre Channel is still a weak point in management, and the software requirements of the client are higher than that of the school, so the choice is When should be based on the actual situation to choose.
SCSI is developed in the following order: SCSI2 (narrowband, 8-bit, 10MB/s)→SCSI3 (broadband, 16-bit, 20MB/s)→UltraWide (16-bit, 40MB/s)→Ultra2 (UltraUltraWide, 80MB/s) → Ultra3 (Ultra Ultra Wide, 160 MB/s)→ Ultra3 (Ultra Ultra Wide, 320 MB/s). When using UltraWideSCSI disk arrays in the past, the CPU requirements do not need to be too fast, because SCSI itself is not very fast. But when SCSI evolves to Ultra2, the requirement for CPU is very critical, and the general CPU (ie, 586-level CPU) must be changed to a high-speed RISC CPU.
The structure of the server has changed from the traditional I/O structure to the I2O structure. The purpose of the server is to reduce the burden on the CPU in the server and to separate the system I/O from the server CPU load. I2O is a RISC CPU responsible for I/O work. The RISC CPU has been used on the server. Of course, the RISC CPU must also be used on the disk array. In addition, most of the commonly used network operating systems are 32-bit or 64-bit. When the operating system is switched from 32-bit to 64-bit, the CPU on the disk array must be a RISC CPU to meet the requirements.
2. Is there any order requirement for the hard disk in the disk array?
In Other words, can the hard disk be inserted back into the array in the original order and the data can still be accessed normally? Many people take it for granted that there should not be order requirements at all. It is not. We have used an array that must be in the correct order to access data. Now suppose that such a situation, we are ready to clean up the hard disk array, put all the hard drives together, the results can not remember the order, in order to access the data, we only have one test, and for an array of eight hard drives In the worst case, 88 attempts will be required. Now that the disk array product has a function that does not require the hard disk sequence, in order to prevent the occurrence of the above events, an array that does not require an order should be selected.
3. Is it a hardware disk array or a software disk array?
A software disk array refers to the connection of a SCSI card to a disk. A hardware disk array refers to an array with a backplane in an array cabinet. It is very different from a software disk array. Hardware disk array is a complete disk array system connected to the system, built-in CPU, parallel operation with the host, all the I / O is completed in the disk array, reduce the burden on the host, increase the overall system performance, SCSI bus master With DMA channels to accelerate data access and transmission. The software disk array is a program that is executed on the host and forms an array by connecting a SCSI card to the disk. Its biggest drawback is that it greatly increases the burden on the host. For a large number of input and output systems, it is easy to paralyze the system. Obviously, hardware disk arrays should be chosen as much as possible.
4. Is IDE disk array or SCSI disk array?
IDE disk arrays have recently appeared on the market. They are very fast. For example, the enhanced IDE has a transfer rate of 66 MB/s under the PCI bus. The price is much lower than that of SCSI disk arrays. The SCSI Ultra3 speed is close to 160 MB/sec. s. However, from the perspective of practical applications, IDE disk arrays are faster than SCSI disk arrays in single tasks. In multitasking, SCSI disk arrays are much faster than IDE disk arrays.
However, IDE disk arrays have a fatal disadvantage: they cannot be hot swapped. This disadvantage makes IDE disk arrays doomed to be used only in non-essential situations. If your application cannot shut down, be sure to choose a SCSI disk array.
5. Is it a single controller or a redundant controller?
Disk arrays are generally connected to a host and a disk with one controller, and the data integrity is achieved under fault tolerance of the disk array. However, the disk array controller will also fail. In this case, the data may be lost. To solve this problem, two controllers can be connected by cables to back up each other. However, the connection of two independent controllers in the chassis means that the chassis must be opened to switch off the controller in the event of a failure, that is, it must be shut down. This is simply not possible in many applications. Therefore, we should choose hot-swappable dual-control redundancy. Architecture. Some new disk array products now use cache memory and memory mirroring to ensure that no data is lost when a failure occurs. After the controller is replaced, the pre-fault working settings are automatically restored and the workload is distributed to each other. The controller achieves load balancing. This architecture can provide high performance and high security that can not be achieved by a single controller.
6. SCSI interface or fiber channel interface?
SCSI's complete specifications, mature technologies, and high performance have always attracted small systems. However, from the current situation, the Fibre Channel has formed a market with double rings up to 200MB/s, a transmission distance of 10km, and 126 devices. Fibre Channel combines the bus and network into one and is fundamental to the storage network. It has become a general trend to replace SCSI. Therefore, in order to ensure the vitality of the system, the Fibre Channel interface should be selected. However, the cost of the Fibre Channel network is particularly high, which is about 4 to 5 times that of the SCSI interface network. From the actual situation, the Fibre Channel is still a weak point in management, and the software requirements of the client are higher than that of the school, so the choice is When should be based on the actual situation to choose.
Seamless Steel Pipe Co., Ltd. , http://www.cz-steelpipes.com