Graphistry 2.36.6 brings updates to both our no-code users and hands-on GPU ones for how they explore relationships across their data. In addition, it coincides with a new release of `graph-app-kit` focused on TigerGraph support and fixes.
GPU updates: RAPIDS 0.18 and Multi-GPU sharing
The latest release helps GPU users do two important things:
- RAPIDS 0.18: Quicklaunch and use a stable release of the new RAPIDS 0.18. The RAPIDS 0.18 ecosystem update includes important fixes and speedups to core cudf, dask_cudf, and blazingsql, such as on graph-centric queries like `SELECT * GROUP BY src, dst`. Likewise, it brings in new features, like new cuGraph algorithms. Graphistry uses RAPIDS underneath, and for advanced users, bundles in the full RAPIDS ecosystem, such as a Jupyter environment with all the libraries and dependencies already setup. Pick your instance size and it just works!
- Multi-GPU sharing: For bigger-than-memory and multi-GPU processing, you can now work with large files with our preprackaged Dask runtime. The latest release adds a shared folder `/dask-shared/`, so your custom apps and notebooks can get a unified view of files for those tasks. This will be added for graph-app-kit users as well.
For example, a notebook user can now download a big file and have a Dask task access it:
#! wget -O /dask-shared/1GB_of_events.csv https://github.com/1GB_of_events.csv
import dask_cudf, dask.distributed
with dask.distributed.Client('dask-scheduler:8776'):
dgdf = dask_cudf.read_csv('/dask-shared/1GB_of_events.csv')
To use a special location for that folder, mount it as `data/dask-shared` in your Graphistry release before starting. Your `docker-compose.yml` will automatically link upon start.
TigerGraph in graph-app-kit
We’re excited to welcome our partners at TigerGraph, a scalable graph database, into the graph-app-kit project for quickly launching interactive graph apps and dashboards with Streamlit. To see it in action, check out our recent webinar on no-code and low-code for GPU visual graph analysis that we held with the TigerGraph team. As soon as the current release update lands in AWS Marketplace, we’ll update the graph-app-kit quick launchers to use it as well.
Alongside the TigerGraph support in graph-app-kit, a variety of bugs and speedups were added to the project, so take a look!
Additional highlights
Graphistry has some additional updates, such as fixes to datetime handling and around the new file uploader. To learn more, dig into the v2.36.6 release notes.
One of our favorite improvements is start time: Autochecks kick in faster, so for many users, a new `docker-compose up` command will launch in 1 minute.
Happy graphing!