Wisedu Digitalization Reduces Costs and Improves Efficiency

Background

Wisedu is a provider of higher education information services based in Nanjing, Jiangsu, China. The company offers customers independent R&D software products and services for teaching, scientific research, management, and life services. These services help universities transform their talent cultivation and school affairs governance models through digitalization and intelligence.

Pain Points

Wisedu relies heavily on Redis for many of its core service systems. However, as service volume has grown in recent years, the on-premises open-source Redis has encountered several pain points.

• Primary/standby switchover in open-source Redis can result in key loss, which must be manually restored.

  Open-source Redis (primary/standby) uses asynchronous replication, which can result in significant data loss during a failover.

  Wisedu services utilize Redis to store tokens for app login verification, but a primary/secondary switchover can cause users to repeatedly verify their login, a common problem with open-source Redis.

  In addition, data integrity is crucial in the student report service, as lost data can prevent access to information about certain students. In such cases, manual data supplementation is necessary, which can be a time-consuming process.

• Services are often impacted by flow control on open-source Redis shards.

  The access model for the service usually encounters hotspots. Although the open-source Redis cluster has a high overall bandwidth, each shard's bandwidth is often low, leading to frequent shard flow control and service disruptions.

  In Wisedu services, flow control can have a significant impact on service availability, emphasizing the need for a Redis service that doesn't require flow control.

• The rising costs of upgrading and configuring a growing volume of open-source Redis data can have a negative impact on services.

  In scenarios involving massive data processing, Redis is often used to store data for extended periods. Expanding the capacity of open-source Redis involves shard addition, which is a time-consuming process that can negatively impact services.

  In addition, as data increases, the cost of using open-source Redis increases, resulting in high expenses for the company. As a result, Wisedu requires a cost-controllable KV database urgently.

GeminiDB Redis API Solution

GeminiDB Redis API is designed to address the challenges of open-source Redis and improve efficiency while reducing costs. Over the years, GeminiDB Redis API has improved many features based on the open-source ecosystem. For instance, it solves the fork problem to enhance performance, supports automatic scale-out in seconds, enables in-place PITR data restoration, and facilitates cross-region DR.

To tackle the issues faced by Wisedu when using Redis, GeminiDB Redis API offers comprehensive solutions.

• Data reliability has been greatly improved.

  If write traffic surpasses 10,000 QPS, a primary/standby switchover could result in the loss of 10,000 service data records, even with second-level AOF flushing configured. This is due to the negative impact of the flushing process on performance, making it unsuitable for online environments. Consequently, in service scenarios where data is essential, open-source Redis may not be the best option due to the risk of data loss.

  GeminiDB Redis API ensures high data reliability by storing three copies of user data by default, but only charging for one backup. Compute nodes use write-ahead logs (WALs) to implement command-level reliable storage. Even if a shard failover occurs, the integrity and security of service data can be ensured because a storage pool is available to ensure the reliability of full data.

  Wisedu services have significantly improved after migration to GeminiDB Redis API, eliminating the need for manual data supplementation.

• Exclusive containers provide enough shard bandwidth, eliminating the need to manage flow control.

  In open-source Redis deployment, multiple tenants share a single container, which requires flow control to prevent interference between them. Shard flow control often leads to the bucket effect. This issue can only be resolved by deploying Redis in separate containers.

  GeminiDB Redis API uses independent containers for each shard, eliminating the need for extra flow control on bandwidth. The bandwidth of each shard is allocated to service programs, so there is no need to worry about flow control even if service access skew occurs.

• Automatic scale-out has no service interruptions. A high data compression ratio can cut storage costs by over 30%.

  GeminiDB Redis API provides excellent scale-out capabilities, allowing for automatic and seamless scaling without any manual intervention. This ensures that services can expand smoothly without any interruptions or latency problems, and the server can handle increased demand automatically.

  Migrating 100 GB of data to GeminiDB Redis API takes up less than 50 GB of space, resulting in long-term cost savings for Wisedu. This is due to the efficient compression technology used by GeminiDB Redis API, which compresses both logical and physical data blocks, effectively reducing storage overheads.

Customer Experience

Wisedu migrated its core services to GeminiDB Redis API nearly a year ago. The services are stable, with an average service latency of less than 1 millisecond and a P99 latency of less than 2 milliseconds. Efficiency of O&M and development have both increased. Moving forward, Huawei Cloud Database will continue to collaborate with the Wisedu to explore more service-oriented capabilities in the KV database service direction, further improving customers' O&M and development experience.