In the first post of this series, I discussed how the software-only vs. appliance-based nature of hyperconvergence clearly shows that the various solutions are not created equal.
In this post, I want to talk about the underlying file systems used by the different hyperconvergence systems. This is another important differentiator because it determines how scalable and enterprise-ready a hyperconvergence solution will be.
Most open source file systems to date have focused on single node concepts with optimizations built around that concept. These file systems were never designed with the modern distributed data center environment in mind. Yet a number of hyperconvergence solutions available today are based open source file systems. The developers of these solutions have had to work hard to try and overcome the inherent limits of the file systems. But in spite of their efforts, their hyperconvergence solutions typically show certain weakness related to the underlying file system.
Realizing the limitations of open source file systems, Springpath founders, Mallik Mahalingam, an original author of VXLAN (virtualized network distribution) and Krishna Yadappanavar, an architect behind VMFS (VMware VM file system), applied their expertise toward building a file system ideal for the modern data center.
Their idea was to break the data management problem down to objects and then build a fundamentally scalable system that is extremely efficient in its clustering operations. This allowed them to solve the scale problem in a way that is truly enterprise-ready and not just focused on a particular use case or two.
This work is what the Springpath Data Platform is based on – a distributed ﬁle system built from the ground up to deliver high performance, reduce operational complexity, and simplify data management. (We refer to this functionality as the patent-pending HALO architecture.)
Because Springpath is based on this custom file system, it is one of the most linearly scalable hyperconvergence solutions available, with full line-rate performance (with all features on) and no tradeoffs required in terms of performance, resiliency, scale or economics.
The HALO architecture fundamentally sets Springpath apart from other hyperconvergence offerings, and is another reason we are confident in saying that not all hyperconvergence solutions are created equal.
In the next post, I’ll discuss differences in adaptive scaling among the various hyperconvergence solutions.