Nebula Graph is an open-source distributed graph database created by Vesoft to allow users to uncover deep relationships between data sets. It has advantages over other graph databases in its architecture which uses a shared-nothing structure for high availability and separates storage and computation for high scalability. Nebula Graph has been adopted by major companies and can store and process huge amounts of data, with performance benchmarks showing it to be significantly faster than alternatives.
2. https://nebula-graph.io
Agenda
l Who We Are?
l What is a Graph Database?
l Why Open Source?
l Architecture
l Advantages
• In Architecture
• In Data Amount
• In Performance
l Adopters
l Product Roadmap
l Summary
3. https://nebula-graph.io
Founder: A Graph Database Forerunner
Founder & CEO
Infra Software
Engineer
(2010-2015)
Started and led the high-performance
distributed Graph Database project - Dragon
• First distributed Graph Database ever in the industry
• Deployed on more than 500 nodes by Jan. 2015
• Together with the write-through cache system TAO, Dragon served all
relationship queries across the entire Facebook application stacks.
Principal
Software
Engineer
(2015-2018)
Formed the team to build the high-performance Graph
Database GeaBase from the ground up
• GeaBase is the only technical product that was rooted in Ant Financial
and widely adopted in Alibaba Group
• Deployed on more than 1000 nodes and served more than one billion
queries every day. The average latency is about 40ms
• Served in almost one hundred scenario and covered business groups
from Ant Financial, to AliExpress, GaoDe Map, AliMama, TianMao,
CaiNiao, UC, etc.
Sherman Ye
4. https://nebula-graph.io
Who We Are?
Team
l 40+ full-time employees, more than 30 of them are
technical persons
l More than half of the technical employees are from Alibaba,
Hauawei, NetEase, etc.
l vesoft Inc. was founded in Oct. 2018, Nebula Graph became
open source in May 2019
l Finalized $3M angel funding: Matrix Partners China
l Raised $17M Pre-A funding: Source Code Capital, Red
Point Ventures, Matrix Partners China
History
5. https://nebula-graph.io
What is a Graph Database?
l Typical database consists of tables filled with same type of data, useful for quick retrieval of such
data.
l A graph database can uncover deep relationships between many data sets
l In a graph, a VERTEX (or NODE) defines an item, a defined EDGE connects it with other data sets
or items (one way, two way or multiple ways), and PROPERTIES can further classify a VERTEX
for more granular data mining
l So, instead of just calling up people with the name SMITH data scientists can instead intelligently reveal
how SMITH might relate to JONES or DOE, in a given time, a location, and more
l These capabilities lead to big users: Adobe, Facebook, Microsoft, Netflix, and many more
l Graph database market expected to grow $650M (2018) to $4.13B by 2026 (Verified Market
Research)
6. https://nebula-graph.io
Why Open Source?
l To make the graph technology more accessible to the world
l To build a healthy ecosystem around Nebula Graph
l To expand globally
GitHub star: 5200+ WeChat group
members: 1000+
Contributors: 50+ Forum posts: 1,100/month
8. https://nebula-graph.io
Advantages
Data Amount in Example:
l Data amount: 150TB
l Graph size: One trillion edges/connections
l An hourly update of 10 billion connections
Compared with other graph database solutions, Nebula Graph has the following advantages:
In Architecture
l Shared-nothing structure - ensures high availability
l Storage and computation separation - ensures high scalability and cloud ready
9. https://nebula-graph.io
In Performance: Meituan
Link to the topic on the forum:
https://discuss.nebula-graph.io/t/benchmarking-the-mainstream-open-source-distributed-graph-databases-at-meituan-nebula-
graph-vs-dgraph-vs-hugegraph/715
Real-Time Write
We invite you to read a real large customer’s own performance benchmarking, conducted by the
NLP team at Meituan: NebulaGraph vs. Dgraph vs. HugeGraph
11. https://nebula-graph.io
In Performance: Tencent Cloud
Data import 1-degree friends query 2-degree friends query Common friends query
Performance comparison conducted by the Tencent Cloud team:
NebulaGraph vs. Neo4j vs. HugeGraph
Link to the topic on the forum:
https://discuss.nebula-graph.io/t/performance-comparison-neo4j-vs-nebula-graph-vs-janusgraph/619
12. https://nebula-graph.io
In Performance: 360 Digitech
360 Digitech has shared their experience migrating from JanusGraph to NebulaGraph and the
huge performance gains after the migration.
Link to the topic on the forum:
https://discuss.nebula-graph.io/t/data-migration-from-janusgraph-to-nebula-graph-practice-at-360-finance/672
HBase network I/O Nebula Graph network I/O
HBase disk I/O Nebula Graph disk I/O
13. https://nebula-graph.io
In Performance: 360 Digitech (Cont’d)
Test Results from 360 Digitech
l NebulaGraph significantly outperforms in disk or network I/O
l Performance achieved using only 30% of HBase cluster machine resources
l When JanusGraph needs 2-3 seconds per query, Nebula Graph just needs 100 ms
l When JanusGraph needs 10-20 seconds per query, Nebula Graph needs 2 seconds
l Overall Nebula Graph performance is more than 20 times improvement over others
16. https://nebula-graph.io
Summary
l Nebula Graph is a VC-funded solution already adopted by some of the world’s
largest Internet companies
l Nebula Graph is proven the world’s highest performing Graph Database
l It can store and process hundreds of billions of data points with trillions of relational connections in a
shared-nothing distributed architecture
l Graph database market to quadruple in size by 2026