How much will AliCloud pay for a downtime?
In the early morning of March 3, Alibaba Cloud experienced downtime.
Affected by downtime, many Internet companies in North China have lost their APP and websites, and a large wave of programmers, operators and O&M have to get up from their beds to work.
More importantly, this is not the first time Alibaba Cloud has failed. Hong Kong VPS
The netizen "Shanghai Lanmeng Network Xia Licheng" joked that "Alibaba Cloud goes down every year, especially early this year".
After the panic of downtime, people need to think about why downtime failures occur frequently and how to compensate afterwards?
Horror for three hours
As for Alibaba Cloud's downtime, Shen Jian, 58 senior architect, said that the accident lasted about 3 hours and observed for 2 hours afterwards.
The most direct impact of downtime is that enterprise websites or apps that purchase Alibaba Cloud services cannot be used normally.
If "cannot be used" is still an abstract term, the affected enterprises can provide a more concrete understanding.
Kongfuzi's old book website announced on the 3rd that due to large-scale Alibaba Cloud failures, Kongfuzi is temporarily unavailable. Implicitly, during the downtime period, users will not be able to purchase goods in KongNet.
Then compare with the statement issued on the same day by Jihebi Fen (an application platform for live broadcast of football games), which said that Alibaba Cloud's downtime has caused some modules of Jihebi to get stuck, that is, the user experience has declined.
By analogy, the larger the fault area of Alibaba Cloud, the more affected enterprises and users will be.
About one hour after the outage, Alibaba Cloud officially replied that the ECS servers in Part C of the North China Region 2 zone had IO HANG, which was recovered gradually after urgent troubleshooting.
China News Service's query on Alibaba Cloud's official website shows that Alibaba Cloud services can be geographically divided into three parts: Asia Pacific, Europe and the Americas, the Middle East and India. Specifically, Asia Pacific includes 13 regions, including North China, East China, South China and Hong Kong.
"Part C of North China 2 Region Availability Zone" is one of the regions in North China.
Generally, in order to reduce network delay and improve customer access speed, enterprises will choose to purchase regions close to customers.
Therefore, after this outage, "North China is in chaos".
As more and more enterprises and applications move data to the cloud, every small downtime on the server may cause a disaster.
AliCloud downtime
As the largest cloud service provider in China, this is not the first time Alibaba Cloud has shut down.
In June 2018, Alibaba Cloud experienced large-scale access exceptions, and products such as image services could not be used normally, and the official website account could not be logged in. Officially, the failure was due to an operational error in operation and maintenance. Afterwards, Alibaba Cloud said that it would respect every line of code and every consignment.
In October 2016, Alibaba Cloud East China 1 zone B also experienced an ECS server IO HANG accident.
Further forward, in September 2015, the upgrade of Alibaba Cloud Yundun's Anqi product triggered a bug, which caused some normal files in the user's ECS to be mistakenly isolated. The reason is that the programmer wrote a wrong line of code. Also in that year, Alibaba Cloud launched the "Hundred Times Time Compensation Plan".
In addition, according to media statistics, Alibaba Cloud experienced failures of varying degrees in 2012, 2013 and 2014.
According to a recent report by market research organization IDC, Alibaba Cloud ranks first in China in terms of market share, accounting for 43%, which is equivalent to the sum of the second to ninth places. Next in the list are Tencent Cloud, China Telecom, AWS, Jinshan Cloud, Ucloud, Microsoft, Baidu Cloud and Huawei Cloud.
With such a large scale, every outage of Alibaba Cloud will have a significant impact on customers.
Contrary to its negative impact on customers, Alibaba Cloud has become a global leader in cloud services by virtue of China's large market.
Alibaba's financial report released on January 30 showed that Alibaba Cloud's revenue was 21.36 billion yuan, an increase of about 20 times in four years, making it the largest cloud service company in Asia. Last year, the figure was 11.17 billion yuan.
How to compensate for downtime?
After the outage, Alibaba Cloud said that it would handle the compensation as soon as possible according to the SLA agreement.
"SLA agreement" refers to the Service Level Agreement (SLA). According to Alibaba Cloud official website data, for a single ECS instance, if the service availability is lower than 99.95%, users can get compensation ranging from 10%, 25% and 100% of the monthly service fee.
In addition, the compensation standards of Huawei Cloud and Tencent Cloud are similar.
A cloud computing enterprise engineer told China News Service that the country is a through train, and the compensation for cloud service failure is basically "delivery time". Before that, Alibaba Cloud has implemented "100 times time compensation".
"But this compensation sometimes has a huge gap with the loss of the enterprise." For example, if JD Taobao cannot log in for 5 minutes, how much will it lose.
In response to this outage, some netizens also proposed that in addition to compensation for the use time and vouchers, they should also compensate for "overtime pay". Many transport and peacekeeping programmers got up from their beds to work overtime.
For enterprises, the most important thing is how to avoid failure.
Some analysts believe that although cloud service providers promise 99.99% security and reliability, everyone is likely to be 0.01% of the unlucky. Therefore, there are usually two ways to avoid failures: one is to backup data and update it regularly; One is to use more than one cloud service provider instead of putting eggs in the same basket.
But this will undoubtedly increase the cost of enterprises. How to make cloud service providers more reliable is still a problem to be solved.
