Save 58% IT cost, call function calculation more than 3 billion times, serverless practice of graphite documents

Author: Jin Zhongxi

For Graphite Documents, "user experience, cost consumption, and online speed" are important factors that determine whether it can show up in the vast business world. It requires a set of elastic, highly available, distributed services to handle document editing conflicts in real time: not only to ensure acceptable costs, but also to ensure smooth processing of peak loads during sudden load surges, and to achieve millisecond-level document synchronization responses.

On July 20, 2021, a heavy rainstorm broke the tranquility of Henan and touched the hearts of people across the country.

That night, a large amount of information about the heavy rain in Henan began to spread rapidly on various social media, "list of rescuers", "rescue site information", "accommodation and transportation materials information"... One by one online documents carried life-saving information. Graphite Documents provided free premium products and services to government departments, medical institutions, non-profit organizations, and social volunteers on the front line of the flood fight.

Within two weeks, more than 1,600 disaster coordination documents were created across the country, providing online coordination services for millions of devices, making flood-fighting on-site rescue, material transportation, medical care, and volunteer organizations more efficient. At the same time, it also makes the IT system behind the graphite document face huge challenges, especially: how to ensure the smooth operation of the system.

Only when the system runs smoothly can more people participate in document collaboration, organize, edit, and summarize all kinds of life-saving documents at the first time, so that life-saving information can be transmitted in a timely manner.

Graphite document has withstood the test of this extreme traffic. Behind it, an intelligent operation and maintenance system was built on Alibaba Cloud five months ago, which opened up the whole system of observable data, and realized unified management, intelligent alarm and Pre-check in advance.

As soon as the disaster happened, the engineers of the Graphite Documentation team received an alarm call, and the voice automatically broadcast the K8S (Kubernetes, container management and control system) cluster resource utilization rate on the Graphite online soaring, and the PV (Page View, page views) increased by more than 200% year-on-year; The traffic of some business interfaces has increased sharply, and multiple resources are about to reach the pressure level. It turns out that the number of reads and writes of multiple documents and the number of visitors have increased abnormally.

Due to the surge in the number of short-term visitors and the inability to accurately determine the subsequent upward trend, engineers quickly coordinated several times the previous Alibaba Cloud computing resources to expand the infrastructure on a large scale, ensuring that these disaster relief documents can be coordinated by thousands of people at the same time and millions of people. It can also maintain a stable and smooth experience even when the number of people browses, so as to ensure the smooth progress of the disaster relief work.

In the end, Graphite Documentation has done an excellent job of serving as an "information carrier", and the documentation is always available stably.

Being able to quickly deal with the flow pressure in extreme situations also depends on the technical accumulation of Graphite Documents. Since its establishment in 2014, Graphite Documents has served more than 200,000 enterprises, and it faces two major challenges at the IT level.

One is multi-user real-time collaboration. In order to support multi-user and multi-terminal real-time online writing, the Graphite documentation technical team has carried out a very in-depth exploration and modification of the OperationalTransformation algorithm, and solved the conflict problem caused by the simultaneous modification of a piece of text by multiple clients.

But this will bring the second challenge of , that is, the huge pressure that multi-user real-time editing brings to the server. only takes a few milliseconds for a user to type a word on the keyboard, but if a large number of users write documents on graphite documents at the same time, it is easy to have uneven distribution of data in a small period of time. An intuitive manifestation is " write serial".

For Graphite Documents, "user experience, cost consumption, and online speed" are important factors that determine whether it can show up in the vast sea of business. It requires a set of elastic, highly available, distributed services to handle document editing conflicts in real time: not only to ensure acceptable costs, but also to ensure smooth processing of peak loads during sudden load surges, and to achieve millisecond-level document synchronization responses.

To this end, Graphite Documents cooperates with Alibaba Cloud in depth, and uses the public cloud serverless architecture to deal with traffic challenges.

Function Compute is a serverless service provided by Alibaba Cloud. It dynamically allocates the execution environment according to the request volume, schedules computing resources in milliseconds, ensures stable latency when the load is high, and has high resource utilization when the load is low. Pay for the computing resources your code uses when it runs.

Graphite Documents uses function computing to build a real-time document editing service. The intelligent scheduling system of function computing automatically allocates the execution environment to handle the peak load caused by the simultaneous writing of documents by multiple users. The dynamic expansion and contraction capability of function computing ensures the reliable operation of the application.

With the help of the millisecond-level resource scaling capability of function computing, Graphite Docs solves the problem of computing resource expansion due to sudden load increases during peak periods. Compared with self-built computer room maintenance servers, it saves 58% of IT costs. Currently, Graphite Docs calls functions every month. Calculated more than 3 billion times.

cloud-native era, gradually improving IT infrastructure enables enterprises to carry out digital innovation more efficiently.

Since there is no need to consider the load balancing problem of CPU-intensive computing, not only the stability of development has been improved, but the pace of iteration and launch of each project has also accelerated. As the first enterprise office service software in China that supports cloud collaboration, Graphite Docs achieves a millisecond-level synchronous response speed to realize online real-time collaboration among multiple members and multiple terminals. Powerful and stable technology supports the platform's efficient and stable service experience, and also It is stimulating greater imagination.

With the ultimate flexibility of Alibaba Cloud function computing, Graphite Documents released the "Remote Office Edition", which brings a series of remote office templates to enterprises. These templates are easy to use and enable multiple people to work online in real time. Going further, on December 1st, under the theme of "Reshaping Online Office", Shishi Documents officially launched Shishi Office, a domestic enterprise-level cloud Office for enterprises and organizations, taking a key step towards a broader market.

The Graphite team has been innovating continuously around the needs of enterprise document collaboration, which is an important exploration for the future office collaboration model.