If reprinting, please indicate the source in an obvious place: http://www.cnblogs.com/kelamoyujuzhen/p/8981057.html
I mentioned the two problems of storage before, capacity + speed 003 - RAID
With RAID, the problem of capacity and speed is temporarily alleviated. However, RAID cards generally rely on SCSI cards. One channel on a SCSI card can connect 15 disks. take it out
The disks are placed in a separate box, and the controller is inside the host. The host and JBOD are connected by SCSI cable. There are several problems with doing this:
1. The disks in the disk cabinet are far away from the controller of the host if there is a little bit of trouble.
2. The longest SCSI cable is 25 meters, which is destined to be unsuitable for large-scale implementation.
3. A channel on the SCSI controller can connect 15 disks. If it is not enough, you can only add a SCSI card to the host, and then use this SCSI card in an external enclosure.
The above approach to expanding storage scale is only DAS, not SAN. DAS emphasizes that the storage system is only used by one host. A SAN is essentially a network. In order to solve the first problem, liberate the host and take the SCSI controller from the host to the enclosure. However, at this time, it cannot be called a disk cabinet, but is called a disk array or disk array.
The disk controller is removed from the host and placed in the disk array, so that the array can handle the underlying problems of the disk by itself, greatly liberating the host. The interfaces of the disk array controller are divided into front-end and back-end. The front-end and back-end interfaces can be different.
Backend: can be IDE, SCSI, SATA, FC, TCP/IP
Backend: can be SCSI, FC, TCP/IP
The disk array controller is a virtualization engine, and it only needs to provide LUNs to the host.
The disk array controller can also be connected to multiple host channel interfaces, so that multiple hosts can be connected. At this time, the network has already sprouted a bit, and it can be regarded as a SAN. And the LUN seen on each host channel interface can also be adjusted on the disk controller, with great flexibility. The role of the disk controller in the disk array is equivalent to the CPU inside the host. It is a hot spot and is easy to SPOF. For this purpose, dual controllers can be introduced. According to how the dual controllers work, they can be divided into Active-Standby (ie HA) and Dual-Active. Split Brain is prone to occur for the latter case. The so-called Split Brain means that the two controllers are still OK, but the link between the controllers is kneeling, and the controllers do not know whether the other party is alive or dead, so they all take over the back-end disks. At this time, the disk array is like a madman, commonly known as split brain. Solving the split-brain problem involves arbitration, which will be discussed in detail later.
In the above figure, JBOD is connected to the disk array through the expansion channel interface. There is one JBOD in the picture. If the controller allows, you can connect several more JBODs. Why do you do this? The main reason is that the disk array controller is too expensive, and two are used. If you can't let him be idle, you need to connect more disks and let him take care of them.
The composition of the storage device
Host server type
The operating system used by the host server
A filesystem is also required on top of the OS
With the OS and file system, the application software can be installed on the host server according to the actual application, see the figure below
Disk array type
Install "front-end-back-end" mode. Currently there are SCSI-SCSI, SCSC-FC, FC-FC, SATA-FC. A bunch of nonsense in front is talking about SCSI-SCSI.
The biggest problem of SCSI: it is a parallel bus, which is destined to not be suitable for long-distance transmission. The maximum length of SCSI cable is 2 meters. For this reason, people have made various attempts to transform the front-end and back-end, essentially extending the distribution of the entire storage architecture in terms of physical distance. 010 - Evolution of storage system architecture