[Turn] talk cloud concept of availability and geographical area

Original link: talk about cloud computing in the geographical area and the concept of availability

Amazon AWS public cloud computing pioneer, a number of important cloud computing in product design and basic concepts can be said Amazon introduced. This one has two very important concepts: local ( Region) and free zone ( AZ：Availabe Zone). Many students first contact with the cloud, just look at the literal meaning of the two names, although it is possible to guess roughly means, but after in-depth learning about cloud computing for some time, to a profound experience of these two concepts for the cloud significant impact calculations. These include domestic cloud computing service providers, but also a very long time before being introduced into the area of design available in the product.

Ideally, we certainly hope to completely eliminate the impact of cloud computing area, as we do not care when electricity as power plants where it is. But the reality is clearly not so beautiful, the network between different regions of the room also can not as transparent as the power grid. So at the very bottom of cloud computing products, we first need to consider the impact of different regions. Between different regions, generally only communicate through the public network between the internal network is illogical. Of course, cloud computing service providers, as management needs, or will be generally limited bandwidth communication through different regions of room for internal cloud resources management, as well as some special scenes product, such as cross-regional mirror replication. But because of the limited internal bandwidth is generally not completely open to users.

So, different parts of the region is in the physical sense of the room, the different places, generally speaking distance, the higher the cost between the engine room with direct fiber. And user demand will be relatively large deployment of local area computer room, such as geographical Ali cloud cloud server in the territory of Hangzhou, Shanghai, Beijing, Shenzhen, Qingdao, overseas already on-line, including Hong Kong, Silicon Valley and Singapore. Ali cloud actually began as a no geographical Shanghai, Shanghai, Hangzhou, because the close proximity of the cost of deploying fiber directly connected relatively controllable, before the internal Ali many applications are deployed in Hangzhou and Shanghai, basically as a area to use, and later probably because of the large demand and separated.

Therefore, the region is well understood, the room is far apart physically, because the bandwidth between the cross-regional network room can not meet the demand, so the network between different regions of the machine is not true. Of course, with the development of the physical layer infrastructure backbone networks, cross-regional network connectivity in the future it is not entirely impossible. In this process, a public cloud computing service providers may, in some scenarios some limited open internal network bandwidth to do according to customers' demands, for example, said earlier, Ali cloud mirror replication across geographies, as well as the recent launch of OSS replication across geographies. In general, data and storage products in the field will first cross-regional support functions, after all data disaster recovery is stronger demand.

So, within the same region is divided into a plurality of zones available, why should so complicated? The reason is simple, IT systems from ancient times have disaster recovery needs of the city, after that the use of cloud computing, how to achieve inter-city disaster recovery room it? If you buy a cloud server which can not be distinguished in the room, then you can not be in the business application level design city disaster recovery. So cloud computing service providers put forward the concept of different usable area of the same territory, the simple point is understood, it is believed that the city different room, the cloud computing service Chamber of Commerce from the bottom of the computer room power / network and other aspects carefully designed to guarantee a usable area of failure time will not affect the available area to another, of course, you say that the situation was completely submerged Hangzhou Qiantang River tide, the available area that can not save you, you want to consider the level of business applications designed to disaster Recovery by different regions.

So, in simple terms, it can be understood as the area of the room in different cities, the available area understood as different rooms of the same city. Of course, in fact, different areas can also be available in the same room, the concept of the available area is strictly in accordance with mutually independent power and networking equipment designed. Between the regions with a different available in the art, it is a communication network, but there will be differences in the response time of the network. Here is what I do with Ali cloud Hangzhou area a ping test to observe network conditions between regions with different availability zones.

Host 1 is available in Hangzhou region B, the host 2 D Hangzhou available area, an available area in the database Hangzhou B. The same available area, a host of ping results from a database as follows:

ping rdsxxx.mysql.rds.aliyuncs.com
PING rdsxxx.mysql.rds.aliyuncs.com (100.98.xx.xxx) 56(84) bytes of data.
64 bytes from 100.98.xx.xxx: icmp_seq=1 ttl=56 time=0.260 ms
64 bytes from 100.98.xx.xxx: icmp_seq=2 ttl=56 time=0.472 ms
64 bytes from 100.98.xx.xxx: icmp_seq=3 ttl=56 time=0.396 ms

The same across different geographical areas available, the results from the host database 2 ping 1 is as follows:

ping rdsxxx.mysql.rds.aliyuncs.com
PING rdsxxx.mysql.rds.aliyuncs.com (100.98.xx.xxx) 56(84) bytes of data.
64 bytes from 100.98.xx.xxx: icmp_seq=1 ttl=54 time=1.63 ms
64 bytes from 100.98.xx.xxx: icmp_seq=2 ttl=54 time=1.73 ms
64 bytes from 100.98.xx.xxx: icmp_seq=3 ttl=54 time=1.74 ms

Obviously, the network between different communication areas are available, but the response time is slower than 1ms available within the same zone. Therefore, in practical applications, you need to be considered if the city or city-active disaster recovery, the need to try and distributed database applications deployed in different regions is available. If highly sensitive to response time, it is recommended to deploy available in the same area. At the time of purchase cloud servers and databases, pay attention to choose.

In addition, Ali cloud big data platform is currently deployed mainly in Hangzhou area, if the data business applications need access to Ali cloud platform for big data processing, the cloud server and database priority to buy the best in the Hangzhou area, after Ali cloud will be more multi-regional deploy big data platform. But sad reminder that, on the whole tight resource room Hangzhou, the Hangzhou area of cloud server may from time to time off for a short time.

to sum up

Here write your own understanding of "regional" and "availability zones":

These two concepts are divided on a host of regional management of. "Geographical" scope of this division is relatively large, like a range of city circle. Then the "availability zones" in this range can be understood as a metropolitan area in the neighborhood. The house over a cell, a cell can go to another stay. The same area where access between the available area is easy and fast, so that, when a service deployment, often need the same "geographical" different "availability zones" are deployed about, so that a hung up, another may propped up, and will not serve the whole hung up.

[Turn] talk cloud concept of availability and geographical area

to sum up

additional materials

Guess you like