Virtualization Technology News and Information
VMblog's Expert Interviews: Alluxio Talks Revolutionizing Big Data

Interview Alluxio

It's always a big day for a startup when it ships its first product.  Andreessen Horowitz-funded Alluxio just unveiled two new offerings that it hopes will accelerate its mission to massively disrupt the universe of big data storage similar to what Mesosphere did for Apache Mesos and Databricks did for Apache Spark.  The third layer of the BDAS stack enters the enterprise market.  All three had Andreessen as their series A lead investor and all were hatched at UC Berkeley's AMPLab, ground zero for big data innovation.  Yes Virginia, there is a Berkeley AMPLab Big Data Analytics Stack (BDAS) based on Spark as the compute layer, Mesos as the management layer and Alluxio as the storage layer.

I recently caught up with Neena Pemmaraju, vice president of products at Alluxio - a seasoned enterprise technology executive with data center computing roots back through AMD, Cisco, Force10, Intel and SeaMicro - to ask about the new products.

VMblog:  What is the customer pain you are addressing with your first products?

Neena Pemmaraju:  In the past decade, we've seen an explosion of big data frameworks and a similar sprawl of largely silo'd data storage solutions in the enterprise. It started at Google and the release of MapReduce and the industry embraced these new approaches with Hadoop, HDFS and now today there are more than 50 popular computational frameworks in production in enterprise deployments.

At the same time, there are more than 30 popular storage solutions, from S3 at Amazon and Red Hat Ceph to traditional and expensive high performance solutions from EMC, Hitachi, Huawei, IBM, NetApp and more. But how can you get both sides of the equation to work together easily? With our Alluxio Community Edition and our Alluxio Enterprise Edition products we're trying to help customers bridge that computation - storage gap. It's a mess today. It's complicated. Performance is painfully bounded by I/O constraints. Enterprises really struggle to get value out of their disparate storage systems.

We offer a new way to avoid all that pain and time of doing it yourself with complicated ETL approaches. Think of us as a unified data access plane. We unify data in your enterprise data centers, remote storage and cloud at memory speeds. 

VMblog:  Can you explain how it works?

Pemmaraju:  You simply mount your preferred mix of storage solutions to Alluxio where each storage system appears as a folder to the application. Once it's in Alluxio, all of your data is available to application in memory at memory speeds. There's nothing faster. Alluxio exposes a unified API in a global filesystem namespace and provides high performance for a wide range of applications such as machine learning, AI, predictive analytics, fraud detection, whatever the use case where you have a lot of data and performance is paramount. It's software only with a very light footprint that puts the data local with the compute.

VMblog:  How are you different from other cloud gateway solutions, hyper-converged hardware appliances or similar storage solutions that aim to virtualize the data for an application?

Pemmaraju:  We do several things uniquely compared to other approaches to solving this storage problem. We co-locate compute and data to provide memory-speed access to the data. There is no faster way to compute. Our software also effectively virtualizes all of your underlying storage systems. Any compute you want to run sees all the different data stores in memory mounted under a unified namespace. We're the aggregation point for all the different stores. Whether the data is stored in the cloud, NAS or is direct attached storage, it's all mounted to Alluxio like a file folder. 

We're software only and built in a distributed fashion to support massively scale-out architectures so we can keep up with almost any growth. Our API is completely compatible to how applications access data today, through a simple filesystem API. Applications don't require any software changes to take advantage of Alluxio. We're very, very easy for any application to use. No one else can do all of these things.

VMblog:  How is Alluxio Enterprise Edition different from Alluxio Community Edition?  It's always a challenge with an open source-based offering to differentiate sufficiently with your commercial products while continuing to encourage and advance your project and community.

Pemmaraju:  Enterprises today are embracing open source for core infrastructure around cloud and big data computing. At Alluxio, we are committed to open source. As the creators and top ten PMCs of the Alluxio open source project, we have been continuously contributing to open source to bring more value to users. Today we announced Alluxio Community Edition (ACE), which is free to download. ACE makes it easy to deploy, manage and monitor an Alluxio cluster. This will encourage more adoption which benefits the whole community.

In addition, we created Alluxio Enterprise Edition, which contains features and capabilities that make Alluxio enterprise ready. We provide 8x5 and 24x7 support to meet the needs of mission critical production environments. With our support, organizations can leverage all the benefits of Alluxio and focus on their core business.


Once again, thank you to Neena Pemmaraju, vice president of products, at Alluxio for taking time out to speak with
Published Wednesday, October 26, 2016 11:17 AM by David Marshall
Filed under: ,
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
<October 2016>