Today, I will talk about the development and transformation of the server side in my work experience. Since I have been fortunate enough to contact server maintenance work in 2012, I have accumulated a little experience in this field along the way, and I will share it today.
The first contact is VPS.
At that time, I really didn't understand why it was called this name. I checked the translation and called it a virtual server system. It was not until I learned to use a virtual machine to build a server that I realized that the original VPS is equivalent to this manually built virtual machine. I was very surprised at the time. I always thought that the virtual machine was only used to simulate multiple devices. I didn't expect that it could also be used as a server.
Advantages: Compared with the physical machine, the configuration is a little bit more laborious. Before purchasing, choose the configuration preset by the manufacturer at one time: memory, cpu, hard disk, bandwidth, operating system.
Cons: Configuration updates are difficult after purchase. Performance is lower than physical servers. In terms of network, BGP lines are rarely supported, and dual lines often provide two IPs respectively. Difficulty binding multiple IPs
touch on physical servers later
Due to the large estimated load of the GPS service online, I searched for information everywhere, and began to understand the physical server with a little knowledge. The server that was purchased at that time was a rack-mounted IBM. Very fortunate to have access to this knowledge. In order to build a bulky rack-mounted server, I also learned a lot of supporting facilities. A small warehouse of 7 square meters transformed into a computer room. There are quite a lot of things needed: independent power supply, UPS backup battery, UPS head, cabinet, KVM (including keyboard and mouse), monitor, router, switch, optical modem, dual-operator line optical fiber, optical fiber jumper, terminal box, Independent san storage, san switches, anti-static floors, constant temperature and humidity air conditioners, etc. are waiting for a bunch of equipment. After all the facilities were complete, it took a few days just to debug and install. But it was also a very fulfilling period. This computer room was later upgraded and expanded several times, and even virtualization technology was built later.
Advantages: The performance of a single machine has been greatly improved. I don't know if I don't compare it. Some performance differences between VPS virtual machines and physical machines are huge. The only thing to talk about here is the performance of the hard disk. There is no loss caused by the virtual disk link, and the database IO is greatly improved.
Disadvantages: The server is cumbersome and requires daily maintenance of the environment, the line is single (2 lines are also single), the network quality is poor (non-backbone nodes, some provinces in China cannot even access, and the inter-operator delay is obvious) and so on. In addition, the cost is high, electricity and bandwidth costs continue to be paid, early hardware equipment costs are high one-time expenses, and there are continuous maintenance costs.
Later, cloud server, here is Alibaba Cloud ECS .
At that time, it was simply a matter of buying an ECS to act as a single physical machine. Going forward, as many drawbacks of the physical computer room began to be exposed, the process of improving the cloud architecture slowly began. Now, all businesses have run on Alibaba Cloud.
Therefore, I also briefly summarized a basic cloud architecture: SLB+ECS+OSS+RDS+cache+queue
This architecture is a typical elastic architecture, which achieves elastic expansion, high availability, and low coupling:
A single machine does not hang if the business is suspended, because the ip of the slb is exposed to the outside world, as long as the backend has a normal running program.
When the pressure is high, the number of ECs can be increased without any perception from the outside world, and even servers can be configured to increase or decrease elastically.
It does not matter if the server bandwidth is small, OSS does not occupy the server bandwidth.
It doesn’t matter if there are many interactions within the system. The cache and queue can be easily decoupled, and the two project teams do not have to be normal at the same time.
This is the road of server transformation that I have been in touch with. I will continue to publish articles later, comparing the advantages and disadvantages of cloud servers, and the corresponding solutions.
The original address: https://www.opengps.cn/Blog/View.aspx?id=37 , the update and editing of the article shall prevail according to this link. Welcome to the original article of the source station!