|
This Presentation date is March 8, 2017 at 13:00-15:00 local time. |
 Shahar Frank introduces Elastifile at Storage Field Day 12 |
Elastifile Company, Product, and Design Overview with Shahar Frank
Watch on Vimeo
Watch on YouTube
Shahar Frank, CTO & Founder, reviews the company’s history and product vision. They are implementing a new data platform for the modern enterprise. Their software-only focus is to support hundreds of thousands of nodes, provide scalable management containers, and focus on data, not on legacy storage.
Personnel: Shahar Frank
Elastifile Hyperconverged Scaling in the Cloud Demo with Shahar Frank
Watch on Vimeo
Watch on YouTube
Shahar Frank, CTO & Founder, demonstrates Elastifile’s hyperconverged scale out solution in Google Cloud. They provide file system services to applications within a simulated load within each node. This can be deployed as a container within a VM, as a VM itself, or on bare metal as an exposed file system, which can be mounted to an individual workload.
Personnel: Shahar Frank
Elastifile Architecture – Overview, Data Path with Ezra Hoch
Watch on Vimeo
Watch on YouTube
Ezra Hoch, Chief Architect, gives an overview of the Elastifile file system data path. They have built a dynamic data path as part of their solution that can choose where specifically to write data, with the ability to cancel and re-choose write location in the event of errors or performance issues. Notably, this is done without any overall data caching, which with NVMe doesn’t provide any performance advantage. Instead, they leverage metadata caching.
Personnel: Ezra Hoch
Elastifile Architecture – Metadata with Ezra Hoch
Watch on Vimeo
Watch on YouTube
Ezra Hoch, Chief Architect, reviews how Elastifile handles metadata within their file system solution. Instead of making metadata a bottleneck of overall performance, especially for numerous smaller file sizes, Elastifile uses a consistency layering approach to get reliable performance at scale. At the lowest level this is governed by a configuration database, which is built on by an in-memory ownership repository, and the actual metadata handling. Any errors in the top most layer has failover to the one below it, which allows for consistent operations and at the base level, is resilient to substantial node and power failures.
Personnel: Ezra Hoch
Elastifile Architecture: Clustering with Ezra Hoch
Watch on Vimeo
Watch on YouTube
Ezra Hoch, Chief Architect, reviews the clustering configuration, which allows for their high availability and fault tolerant configuration database within the file system. The system uses a clustering officer and clustering agent on each node to provide for a consensus driven methodology. Each node monitors and reports on any inconsistencies within the others. He reviews how this system can intelligently handle node failure.
Personnel: Ezra Hoch
Elastifile Node Failure Handling Demo and Discussion with Shahar Frank
Watch on Vimeo
Watch on YouTube
Shahar Frank, CTO & Founder, demonstrates how the Elastifile Architecture handles node failure. Shahar further discusses with the delegates how the solution effects overall performance, and some of the implications of their business model.
Personnel: Shahar Frank