Follow Datanami:
February 27, 2018

TigerGraph Serves Collaboration and Security with Multi-Graph

The capability to share data insights with other people or departments is instrumental to collaboration, and yet it’s also the source of real security concerns in today’s dangerous world. Graph database startup TigerGraph addressed that conflict today by introducing a new multi-graph function in the second major release of its flagship product.

TigerGraph debuted last September with a new graph database that its founder, former Teradata and Twitter engineer and distributed systems expert Yu Xu, claims can do things that other graph database can’t do.

Xu designed the database to be natively parallel and able to run on clusters exceeding 1,000 nodes, (as opposed to running on a single scale-up server). He also designed TigerGraph to be updated in real time (as opposed to needing to go offline to get updated via batch). And he designed it to process complex graph queries that can perform up to 20 “hops” against graph vertices and deliver results with sub-second response time (as opposed to being limited to one or two hops on the graph).

Bolstered with $31 million in venture funding, TigerGraph went to market with this “graph 3.0” functionality and captured business. Along the way, its attracted some fairly large customers, including payment giant Visa and Uber, which recently signed on as a customer. In China, it counts customers like Alipay, the e-payment arm of ecommerce giant Alibaba, which runs what TigerGraph claims is the largest transaction graph the in production in the world, with 100 billion vertices and 2 billion daily updates. It also counts State Grid Corporation of China, the largest power company in the world, as a customer.

TigerGraph’s customers are using the graph database to run what are the two most popular graph workloads: fraud detection and customer recommendations. However, because these two workloads are managed by different employees working in different departments within the enterprise, they have different needs and are covered by different rules governing data access and privacy protection.

Typically, these needs are so different that they would necessitate the development and maintenance of entirely separate graph databases. The product department, for example, will deal with personally identifiable information (PII), while the security team will not want to touch that PII. And as companies add additional use cases, such as customer 360 or “view of the customer” type workloads, they might be tempted to implement a third physical graph database.

That seemed like a giant waste of time and resources to the folks at TigerGraph. “It started us asking the question, how can we have that same data, that same underlying graph database, be able to be viewed by different groups and still maintain the security controls needed, but benefit them by having access to all that data and be able to share the data that’s important to each part of their business?” says Todd Blaschka, TigerGraph’s COO.

“Today companies shave to copy the database from one database to another, and then deal with how do we do updates to that data,” he continues. “It becomes an infrastructure struggle and pain.”

With the new multi-graph function in TigerGraph 2.0, the company is enabling collaboration “where each group can see and view data that’s relevant to them and their roles, but they all benefit from working off the same main graph database,” Blaschka says.

That way, when the database is updated – which is how TigerGraph talks about “real-time” – then everybody benefits from the newest and freshest data for their particular workloads. “Multi-graph allows and enables this kind of collaboration that hasn’t been available before with graph technology,” Blaschka says.

Instead of implementing multiple physical databases, TigerGraph enables an enterprise to maintain one physical database, but split it up into logical components, Xu says. “So you don’t have actually have two different systems, two sets of data that they access. You have one fixed system, you have one copy of the actual data, but then logically we separate them,” he says.

There’s no limit to the number of logical databases the TigerGraph software can support. However, a customer wouldn’t want to have too many logical databases in production owing to the need to optimize the databases for high performance, Xu says.

TigerGraph says multi-graph is an industry first in the graph database market.  Enterprises can support multiple graphs that overlap in some ways or are completely separate, each with separate user privileges and queries, all within the same database.

Other features in TigerGraph 2.0 include deployment enhancements, better package management, and new language features. It also features new security and Active Directory/LDAP integration.

Related Items:

TigerGraph Emerges with Native Parallel Graph Database