Case Study: Achieving a 200% increase in time on site with the Blazegraph High Availability Graph Database
Yahoo7! achieved a 200% increase in time spent on site using a knowledge-driven publishing approach and a High Availability (HA) graph database platform. According to Yahoo7! CTO, Craig Penfold, picking the Blazegraph graph database platform was a key technology choice. He says, “By changing the underlying datastore we are able to pull in all the content around a character dynamically and at scale.”
A Yahoo7! team of eight – including a product manager, editor, engineering, front end developers and data scientists developed and rolled out the semantic web publishing, initially on the Home and Away web site, which is a highly popular Australian soap opera. In order to represent the content within a knowledge graph, the team needed to both understand the format of the content, e.g. text, video, image, as well as the context. Relationships are then inferred between the entities into a graph structure. For example, a scripted drama would involve cast, characters, episodes and storylines. All of this context is is captured in a knowledge graph and stored in the Blazegraph platform, which provides graph database management and graph query capabilities for the portal.
With the new capabilities, native advertising is now injected into the stream of content rather than being served in an embedded site. Blazegraph helps drives this in real-time using graph queries as users browse the site. The rich context and graph query allows advertisers to customise the advertising to reflect the themes of the program and improve engagement with the advertising.
Scaling graphs to the level of a major media and content provider can be a difficult challenge. Yahoo7! choose Blazegraph™ by SYSTAP as the graph database platform. Blazegraph provides a Highly Available (HA) deployment that was a perfect match for the requirements of Yahoo7!. It both provided support for the semantic representation of their site content in RDF and enabled fast graph query using the SPARQL graph query language. The HA features are implemented using a shared-nothing architecture; in the case of a failure, the system continues to operate as long as a quorum of nodes is present. Using the HA features provided Yahoo7! a scalable path to deploy these capabilities. Each node in the HA cluster is a perfect replica meaning that query capacity can scale linearly with respect to the number of nodes in the system. While not yet at this scale, each Blazegraph instance can handle a graph of up to 50 Billion Edges.
Get Blazegraph now: https://www.blazegraph.com/try.