NexentaEdge – Taking Nexenta’s ZFS DNA to Cloud Scale
Thomas Cornely, Chief Product Officer, Nexenta
VMworld 2014 in San Francisco promises to be an incredible event for Nexenta. In addition to our OpenSDx Summit on 8/28, as a VMworld 2014 Platinum Sponsor we’re gearing up for a slew of new product announcements and demos. We’re particularly excited about the opportunity to showcase NexentaEdge, the latest addition to our Software-Defined Storage portfolio, specifically targeting the petabyte scale, shared nothing, scale-out architectures required for next generation cloud deployments. NexentaEdge is a software only solution deployed on industry standard servers running Ubuntu, CentOS or RHEL Linux. Version 1.0 will support iSCSI Block services with OpenStack Cinder integration, Swift and S3 Object APIs as well as a Horizon management plugin. File services are part of our design and will be delivered in follow-on versions.
If you’re familiar with Nexenta, you know all about NexentaStor, our unified block and file Software-Defined Storage solution built on the ZFS file system. What made ZFS great were some key design choices that took advantage of emerging trends in the hardware landscape. Things like Copy On Write, delivering guaranteed file system consistency at very large scale, as well as high performance unlimited snapshots and clones. Things like block level checksums, trading increasingly cheap CPU cycles for way more valuable end-to-end data integrity. This strong technology heritage, paired with Nexenta’s continued investment on performance, scale and manageability has led service providers across the globe and a growing number of enterprise customers to rely on NexentaStor as the storage backend for their legacy cloud and enterprise applications.
If you’re in technology, you know that the only constant is change. Our service provider customers are busy scaling their infrastructure, moving to next generation open source cloud infrastructure like OpenStack and CloudStack and looking for even more scalable and cost efficient storage backends. For that, we’ve built NexentaEdge. And rather than quickly combine existing open source pieces and hack on top of them, we deliberately took some time to design a solution from the ground-up, and we made sure our design reflected the lessons we learned from our years of working with ZFS. We also made sure our design looked forward and was ready for what we see as new emerging trends in the storage landscape. The net result is a truly next generation scale out solution that incorporates what we like to call our ZFS DNA.
For example, one core design aspect of NexentaEdge is something we call Cloud Copy On Write. While the system is fully distributed truly, without any single point of failure, data in the cluster is never updated in place. Very much like ZFS, but applied in a distributed context, this gives us a great foundation for advanced data management and the ability to gracefully handle a variety of failure scenarios that affect large scale out clusters. Another example is end-to-end data integrity. All chunks of objects in NexentaEdge are stored with cryptographic hash checksums that deliver ultimate data integrity. The system is also built with automatic self-healing in case corruption is detected in a chunk.
Another critical aspect of the design is the recognition that building large scale out clusters is as much a storage challenge as a networking one. So we paid particular attention to how we consume precious network bandwidth and how we automatically route data around hot spots and busy nodes. These functions are implemented as part of what we call FlexHash (for dynamic flexible hashing that automatically selects the best targets for data storage, or data access based on actual load states) and Replicast (our own network protocol that minimizes data transfers and enables lower latency data access).
Last but not least, a great design proves itself by how advanced features naturally flow out of it. One such feature is cluster wide inline deduplication. NexentaEdge clusters get inline deduplication of all data at the object chunk or block level. These are variable sizes and can be as small as 4KB as we will demonstrate at the Nexenta booth at VMworld 2014 we will create 100’s of virtual machines on iSCSI LUNs while consuming slightly more than one copy worth in the cluster.
NexentaEdge is here. And we think it will be big. Over the coming weeks, we’ll dig a bit deeper into the technology and share details on Cloud Copy On Write, Flexhash and Replicast. See you at the Nexenta booth at VMworld 2014!