[
https://issues.apache.org/jira/browse/CASSANALYTICS-114?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stefan Miklosovic updated CASSANALYTICS-114:
--------------------------------------------
Description:
Users who happen to have GPUs to spare can leverage its computing power via
Nvidia's project called RAPIDS which is a plugin to Apache Spark which maps
data to VRAM and its processing is accelerated via CUDA.
It would be nice to explore this in context of Cassandra Analytics so we can
speed up the processing significantly.
My preliminary investigation led to the realisation that we would likely need
to process rows in a columnar fashion - the current processing would stay, it
is just about the way we present the data to Spark API / internals.
(1)
https://www.nvidia.com/en-us/ai-data-science/spark-ebook/getting-started-spark-3/#p7-s1?ncid=no-ncid
was:
Users who happen to have GPUs to spare can leverage its computing power via
Nvidia's project called RAPIDS which is a plugin to Apache Spark which maps
data to VRAM and its processing is accelerated via CUDA.
It would be nice to explore this in context of Cassandra Analytics so we can
speed up the processing significantly.
My preliminary investigation led to the realisation that we would likely need
to process rows in a columnar fashion - the current processing would stay, it
just about the way we present the data to Spark API / internals.
(1)
https://www.nvidia.com/en-us/ai-data-science/spark-ebook/getting-started-spark-3/#p7-s1?ncid=no-ncid
> Investigate GPU accelerated processing, e.g. via Nvidia RAPIDS for Cassandra
> Analytics
> --------------------------------------------------------------------------------------
>
> Key: CASSANALYTICS-114
> URL: https://issues.apache.org/jira/browse/CASSANALYTICS-114
> Project: Apache Cassandra Analytics
> Issue Type: New Feature
> Components: Writer
> Reporter: Stefan Miklosovic
> Assignee: Stefan Miklosovic
> Priority: Normal
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Users who happen to have GPUs to spare can leverage its computing power via
> Nvidia's project called RAPIDS which is a plugin to Apache Spark which maps
> data to VRAM and its processing is accelerated via CUDA.
> It would be nice to explore this in context of Cassandra Analytics so we can
> speed up the processing significantly.
> My preliminary investigation led to the realisation that we would likely need
> to process rows in a columnar fashion - the current processing would stay, it
> is just about the way we present the data to Spark API / internals.
> (1)
> https://www.nvidia.com/en-us/ai-data-science/spark-ebook/getting-started-spark-3/#p7-s1?ncid=no-ncid
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
