iQiyi Server Crash due to Server Overcrowding
On the evening of February 16th 2020, iQiyi, one of China’s largest online video site, experienced a large-scale crash. The Windows client and the iOS / Android mobile app could not play most of the videos, and the loading status or an “error has occurred” was always displayed. At first, many netizens thought it was a problem with their mobile phone or internet speed, but then found out that everyone was having this problem on Weibo, China’s largest social media and micro-blogging website (similar to Twitter), where this issue beame listed on the site’s top 10 hot searches.
iQiyi responded to the issue in the same evening,
“Some users reported that the iQiyi app was unable to provide services properly. We’ve immediately launched an emergency investigation and are currently working to resolve the technical problems that caused the abnormal playbacks”.
This is not the first that was encountered in the internet age. iQiyi has crashed more than once and was not the only system or server that crashed. The most widely used mobile payment app in China, “Alipay”, and shopping platform “Taobao”, even Google and YouTube has collapsed server before. Each time, the collapse of the platforms brings huge losses and inconvenience.
Why is there a network crash?
iQiyi provides a central server and assigns rated space to each video ID by default. It’s mainly the resources in the video library that occupy the space in the server platform. Videos from external links do not take up server space resources. This means that iQiyi uses a centralized server to provide data storage services.
Due to the impact of the corona virus epidemic, everyone is spending time at home, and under extremely high traffic, the server and bandwidth are seriously crowded, so there is a “collapse” situation.
On the other hand, in order to prevent congestion during peak hours, many centralized networks have no choice but to provide servers and bandwidth without upper limits, but they are not used during idle time, which will cause a waste of resources to a certain extent.
Data concentration is a very serious problem. For centralized servers, once a power outage or other failure occurs, a large number of related services will often be paralyzed. In addition, the centralized cloud server also runs the risk of causing users to completely lose data and information.
From the current status of the industry, it is very difficult to continue to improve the reliability of centralized storage, the cost will rise sharply. For example, the “two places and three centers” for disaster recovery, the cost will be tens of millions. The reliability of centralized storage has reached the bottle neck. If it needs to further improve its reliability, it needs to be decentralized.
InterPlanetary File System (IPFS)
The InterPlanetary File System (IPFS) is a protocol and peer-to-peer network for storing and sharing data in a distributed file system. As opposed to a centralized server, IPFS is built around a decentralized system of user-operators who can hold a portion of the overall data, creating a system of file storage and sharing that is resilient and reliable. Using the distributed hash table (DHT), peers in the network can find and request the file content served by other users with the content address from any node that has the content.
Distributed storage can use the underlying protocol to allow files stored on the IPFS system to be quickly retrieved from anywhere in the world, without being affected by firewalls, allowing us to openly access data faster and more securely. IPFS distributed storage is a general-purpose infrastructure with almost no storage restrictions. Large files will be cut into small pieces, which can be obtained from multiple servers at the same time when downloading. Its advantages are decentralized and the data can’t be tampered with and deleted, and the ownership of the data belongs only to the person who stored the data; the number of data backups and the storage period are unlimited.
Through such information transmission, the problem that all devices obtain information from the central network (server) and cause the server access to exceed the threshold is resolved.
CyberVein’s Blockchain + DAG based database
CyberVein’s database utilizes both blockchain and DAG technology to achieve a storage that has high efficiency, no storage limit and security. Much like the IPFS, DAG has nodes in its network, where each users represent a node. The data or files that is stored on the DAG network is sharded, where they are split into smaller pieces and kept on each node. The users that joined the network and stored information only needs to maintain the transaction units that are related to him/her, as well as the father-units that comes before, and a single unit does not need to have all the data that’s in the network, and can be verified in real time by other users with the Proof of Contribution (PoC) mechanism. In theory, there is no throughput limit to the transaction volume, and the decentralized network’s scalability can be raise securely, and through timestamps, tamper immunity and tracking, it can greatly reduce the costs. Blockchain and DAG worked together to solve the inefficiencies that exist in the blockchain technology.
The data and information stored in the network, or database, is encrypted and no one apart from the data owner can access them. If the data is needed by other parties, the CyberVein database also uses federated learning to achieve data “sharing” without compromising the data and the privacy of the owner. However, the DAG network theoretically has no limit, but it is dependent on the number of units (users) that is within the network. If by any chance that the network becomes overloaded, part, or all, of the data can be transferred to the blockchain, serving as a back-up space for the data. The data can be retrieved easily when it is needed for use again.
The CyberVein database is highly efficient and secure, with DAG solving the low throughput, efficient and scalability issue of the traditional blockchain database. Coupling with federated database management system and cryptographic modeling platform for federated learning, it has become an all-rounder database. It is already being used in parts of Asia and North America. Compared to IPFS, it is much more efficient and secure, providing a better user experience, making it a worthwhile alternative.