Graphistry 2.26.1: GPU graph investigations with RAPIDS/BlazingSQL, Neo4j, custom pivots, and more!

Posted by Graphistry Staff on November 23, 2019

Release 2.26.1 features substantial GPU computing upgrades and significantly improves the investigation experience for Neo4j and workflow automation.  It’s so big that we’ll be posting multiple tutorials over the next few weeks on how to use some of the bigger features. For now, read one for an overview!


Read on to learn more about this release, and see full release notes at our new release notes page . Docker downloads are now available on the enterprise portal, and the AWS and Azure Marketplace releases should approve the update within 2 weeks. For upgrade assistance, please contact Graphistry staff and we’ll be happy to help.



1. Open GPU Computing: RAPIDS 0.10 and BlazingSQL 0.4.6

The underlying Graphistry engine and out-of-the-box Jupyter notebook environment has been updated around RAPIDS and BlazingSQL. If you’re familiar with SQL or dataframes (Python Pandas, R, Matlab), you write code as usual except now run it on GPUs, plot with Graphistry, and then visually analyze with further automatic GPU computing.

These upgrades are substantial:

  • RAPIDS: GPU umap/t-sne, multi-GPU (Dask), cuGraph algorithms, and many bugfixes
  • BlazingSQL: More SQL constructs, simpler API, and many bugfixes

Stay tuned for tutorials on combining BlazingSQL and cudf/cuGraph/cuML/t-sne/umap with Graphistry. Our results with early users have been spectacular!

Graphistry has RAPIDS built into Jupyter

Figure: Graphistry’s Jupyter notebook environment is preconfigured for Nvidia RAPIDS & BlazingSQL  for easy 20X-200X GPU speedups on normal pydata wrangling 



Neo4j Graph Database

2. Neo4j Pivots: Investigate, File, Share, & Automate

Neo4j made graph databases easy with the Cypher query language, and Graphistry extends this to analyst teams with visual pivots and investigation automation. The upgrade improves our Neo4j support so you can get quite far in visualization and automation without manually writing any queries:

  • New base pivots: visualize schema, search text indexes, search nodes, expand on nodes
  • Create code-free derived pivots for common tasks: See below
  • Schema inference: pivot parameters autocomplete around your database schema
  • Ontology customization: Improved support, such as for assigning node titles based on their label type
  • Time support: Ability to bind one or more properties for use during global+local search

Demo: Graphistry Neo4j autocomplete for no-code Cypher querying

3. Custom Pivots and Query Macros

We are starting to expose’s Graphistry plugin system. The custom pivot layer drastically simplifies pivoting for all users. During the heat of the moment, you don’t want to get derailed by coding and trying to remember a database’s schema. With custom pivots, you can  much more directly follow your investigation idea. Even better, team members can setup custom pivots without heavy coding, just configuring.

Derived pivots give a simple way for users to create new pivots by remixing existing pivots. As soon as you realize one of the powerful base pivots works for a common task, you can bake in smart defaults and hide distracting parameters. For example, after you do a tricky account lookup pivot for one investigation, you can save it as a derived pivot, and reuse it in future ones. Derived pivots definitions are easy to setup too as they’re just some JSON!

Query macros help you create much more powerful and accessible custom pivots. If you know the basic query you want a pivot to do, but want to parameterize it by a few fields, you can create new pivot input fields, and fill them into a hidden query as macros. For example, instead of exposing an account search as a database query, you can now just expose a search box, and treat it as a macro variable within a query hidden underneath. Macros work with the derived pivot JSON system, so you can setup power pivots without dropping into heavier coding layers.

Demo: Configuring a simple search box to drive a powerful query underneath


4. Tweaks & fixes

As always, tons of tweaks and fixes. Our favorites:

  • Time support fixes, such as handling nulls and mixed timezones
  • Performance dials for number of GPU-using processes, giving more control over CPU/GPU process and memory use
  • The RAPIDS upgrade means fixed RTX support, including for the new cheaper G4 (T4) AWS GPUs
  • Air-gapped tarball has shrunk 2-3X to just 3GB, so Graphistry now fits on one DVD!
  • Tons of dependency upgrades throughout the stack: Everything should be just that much more snappier and more secure

As part of releasing new features at a fast pace, we’re also making sure they go through the appropriate testing, and based on everyone’s feedback, tweaking and fixing them. Check out the release notes for further details, and we encourage you to reach out to our team on ways to make your experiences even better.

Curious to multiply your Azure or AWS analytics experience with GPU visual graph analytics and investigation automation? Try Graphistry in your Cloud Marketplace!

Launch in your AWS