Graph Compute is a high-performance distributed graph computing service developed by Alibaba Cloud. It provides an end-to-end graph computing service that supports trillions of data records. Graph Compute allows you to efficiently store, query, and compute complex graph relationship data by using graph algorithms and models. Graph Compute can be used in a wide range of scenarios such as recommended advertisements for searches, real-time risk control, knowledge graph, and social networks.
Why choose Graph Compute?
Simplified expression of complex relationships
You can use a PKey-SKey-Value (KKV) table to efficiently express the relationships in a graph model. Commonly used two-level deep queries are simplified into one-level deep queries. This makes the query syntax simpler.
Better than open source options
Graph Compute supports open source Gremlin syntax and optimizes the performance of key operators. This improves the efficiency of graph computing.
Flexible extension to support large amounts of data
Graph Compute instances are deployed in a distributed cluster architecture. This allows you to flexibly perform O&M operations and scale the capacity of graph computing and data storage within minutes.
Deep integration with the big data ecosystem
Graph Compute is integrated with the big data and AI ecosystem of Alibaba Cloud, and provides scenario-specific solutions for graph computing when used together with MaxCompute and Realtime Compute for Apache Flink.
Benefits
High performance
Graph Compute provides low latency queries and supports fast data import. The latency for a query of hundreds of billions of data records is within 2 milliseconds, and the speed for importing data can be up to five million documents per second.
Low cost
Graph Compute provides the extended features of inverted query and vector processing. This reduces the consumption of engine resources by half in the same scenario compared with other solutions.
High availability
Data is backed up in multiple versions in the offline system, and you can restore data within minutes. This provides Graph Compute strong capabilities in disaster recovery and data rollback.
Millions of TPS
Graph Compute uses an asynchronous update architecture. A single node supports millions of transactions per second (TPS) for updates. This ensures high data timeliness.
Fully managed
Graph Compute provides a graphical console that helps improve the efficiency of data development. Your data is fully managed without O&M investment.
Advantages of Graph Compute over other open source graph computing services
1. Better query performance
iGraph is the kernel engine of Graph Compute. Based on partition-based concurrent queries, iGraph uses the asynchronous coroutine framework developed by Alibaba Cloud for concurrent recalls. This transforms synchronous serial disk access into asynchronous parallelism, which greatly improves query performance. Also, some computing tasks are performed on storage nodes to ensure the performance of complex computing.
io_uring is used in disk-based queries. This way, queries are stably executed in high IOPS scenarios.
iGraph uses the multi-level cache technology developed by Alibaba Cloud. This makes iGraph more suitable for hot data queries.
2. Advanced data import capacity
Graph Compute provides a distributed index-building service developed by Alibaba Cloud, and allows you to manage full offline index data, incremental index data, and real-time index data in a centralized manner. This helps resolve the excessive expansion of the log-structured merge-tree (LSM tree) for online indexes.
Graph Compute ensures the eventual consistency of data that is written in real time. Compared with the strong consistency supported by open source services for data writes, Graph Compute has an advantage over the amount of data. Graph Compute improves the performance by 10 to 20 times.
3. Lower cost
Priority is given to low cost and high performance in terms of the architecture design of Graph Compute. This ensures the high usage of resources.
The automated and comprehensive O&M system of Graph Compute greatly reduces O&M costs.
Graph Compute provides native support for multiple types of indexes, such as key-value (KV) indexes, KKV indexes, inverted indexes, and vector indexes. You do not need to apply for additional business resources.
4. Higher stability
iGraph has been optimized and refined from use in years of extreme scenarios such as big sales and severe exceptions. iGraph works with a highly available and automated O&M system that supports automatic load balancing, automatic scaling, and automatic downgrade to dynamically balance traffic and recover from exceptions at the earliest opportunity.