Alluxio In-Memory Acceleration Layer Announced

Today, Alluxio announced a plan to collaborate with Intel to offer an in-memory acceleration layer. Alluxio develops open-source data orchestration software for the cloud and has done so since 2014. Version 2 of their eponymous Alluxio software was released just last year.

Today, Alluxio announced a plan to collaborate with Intel to offer an in-memory acceleration layer. Alluxio develops open-source data orchestration software for the cloud and has done so since 2014. Version 2 of their eponymous Alluxio software was released just last year.

Following up on a whitepaper published by Intel last year, Alluxio plans to offer an in-memory acceleration layer for big data environments. Alluxio’s internal tests show that their acceleration layer takes less than half as long to complete decision support queries than just a standard disaggregated S3 object storage. Disaggregated here means that the data storage and compute clusters have been separated. Increasingly, companies have been turning to disaggregated data storage solutions to allow them to lower costs. Disaggregation lowers costs by using less expensive data storage options while still keeping faster and more expensive compute clusters. Separating your data and processing units like that has had obvious impacts on performance, and Alluxio’s aims to reclaim some of that lost performance with their acceleration layer. Interestingly, Alluxio claims that their extra layer also showed a 1.3x speedup over a co-located compute and storage architecture during their testing.

Alluxio also announced longer-term plans to continue working with Intel. They said that they would continue collaborating with Intel to provide optimization options specifically for Intel analytics. So far, Alluxio has mentioned possibly working on optimizations for Intel Deep Learning Boost, the AI acceleration technology built–into Intel Xeon Scalable processors.

Alluxio Main Site

Discuss on Reddit

Engage with StorageReview

Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | Facebook | RSS Feed