TigerGraph brings machine learning and analytics features to the cloud

The cloud’s place in the data environment is growing, and TigerGraph wants to bolster its role. Today, the company rolled out several new features so cloud users can deliver more analytics and artificial intelligence (AI) power without leaving the confines of TigerGraph’s database. The updates expand the role for its tools, as TigerGraph’s product is becoming more of a data analytics and AI platform than just a mechanism for storing graph data.

“Our biggest enterprises are still on-prem,” said Jay Yu, a vice president of product and innovation at TigerGraph. “However, we know a lot of smaller and medium-sized customers or more advanced customers, they actually prefer cloud.”

Analytics for the cloud

The announcements include three features that bring a collection of analytics and analysis to the cloud customers. These options were often available to users who directly installed the database, but often they required a substantial amount of additional configuration, often including big data pipelines like Apache Spark. The cloud versions are already integrated and available to anyone, including those using the free tier.

TigerGraph Insights is one major tool that offers a low-code and no-code option, so users can create reports and graphics with a simplified interface. The goal is to let users, including more non-technical users, create “graph stories” that boil down the details from the database into reports and dashboards that will automatically update themselves.

The database already offers customers the option to download data from the API to other data analysis platforms like Tableau, but the new feature offers options that can use some of the networking data that’s often embedded in the graph. It can deliver answers to graph questions like which node is the most connected to other nodes with no more than two hops.

“We’re not trying to replace those tools that we know people already have.“ said Yu. “But the challenge is that they cannot do the graph work. You cannot [analyze the network]. You can only do the tabular view.”

Graph databases like TigerGraph were designed specifically to both store and analyze networks of nodes or data elements. Some standalone graph database companies like Neo4J or ArangoDB are attracting attention. At the same time, major database companies like Oracle are adding the features for storing and analyzing graph data. Cloud versions of these databases and others like AWS Neptune are active competitors for TigerGraph in the marketplace for storing graphs in the cloud.

ML and AI for cloud data

Another major step forward in today’s announcement is ML Workbench, an option that lets cloud users integrate popular open-source Jupyter notebooks directly with the graph data. Users can start to train the AI model using not just the raw tabular data, but also the interconnections that make up the network in the database.

Performing machine learning (ML) directly inside the database is becoming an essential feature for databases. In the best scenarios, it can save the time for exporting the data to a separate function, a delay that can be substantial with large datasets.

“When you look at a set of data to figure out what’s the pattern behind it and you train it with graph data too, suddenly the model is much smarter.” said Yu who also pointed that it’s naturally good for tasks like fraud detection where the relationships between people or other data elements can be revealing.

This kind of analysis is often a perfect match for the cloud. The training procedures can demand heavy computation using specialized processors that may only be used occasionally. Cloud users can run analysis when it makes sense and avoid investing in extra hardware.

Companies like Oracle offer their customers the option to work directly with machine learning algorithms using their databases. Some cloud vendors like Microsoft, Amazon or Google have engineered their artificial intelligence toolkits to interface directly with many of their cloud data storage options. It’s common for data storage solutions to be sold not just on their ability to retrieve information, but also to feed it directly into a machine learning algorithm.

The same reality is also roiling the world of data analytics. The gap between the analytics tools and the data warehouse or database is growing more blurred. The data scientists who once turned to companies like Databricks or Snowflake are now finding that they can do much of the work using tools that are integrated with the database. IBM, for instance, bundles their Watson AI software with their data storage in a product that mixes analytics with machine learning and data storage.

This is the same path that TigerGraph is following with these announcements. It wants to continue to integrate more features like this directly into their product, effectively turning it more into a data analysis platform than just a simple data storage solution.

The cloud versions are also becoming an essential part of marketing, in part because the company offers a free tier that allows developers or data scientists to experiment without purchasing servers or spending time installing software on local machines.

“We see a big pickup on the cloud,” said Yu.  “We keep introducing new things quickly on the cloud because we can highlight those new features. We believe that will make it even easier to push graph to everybody, Our goal is graph for all.”

‍