Jan 16, 2020

Hello, Containers... Goodbye, HDDs.

Hello, Containers... Goodbye, HDDs.

Posted by

Jeff Denworth

Today, VAST announced the availability of a CSI driver for container environments. As things go with containers and the container orchestration platforms that are en vogue today, we are primarily shipping this to customers running Kubernetes (K8S) - but we're also excited to see this used beyond K8S, with platforms such as OpenShift and more. To customers running K8S, this hasn't been a mind-blowing feature so much as it is table-stakes for organizations who want to programmatically deploy, provision and manage persistent container volumes. That said, we wouldn't be doing our jobs unless we put our spin on how to make container storage even more easy, even more affordable and even more performance.

Paradoxically, many of today's modern (containerized) applications are being powered by mechanical media that is anything but modern. Customers are refactoring codes to be cloud native, and then jumping into the wayback machine and powering them with a storage media technology that was invented in 1992. 1992... I have members of my team that weren't even born yet! ps: not Howard or I if you're wondering :). Anyway, VAST has innovated so many different parts of storage architecture and rethought how flash infrastructure can be affordable for the masses such that customers no longer need to compromise on performance to buy capacity - we're eliminating this compromise to be an extinction level event for the HDD.

Chris Mellor of the Register and Blocks and Files may have said it best:

VAST Data wants to direct an extinction event for hard drives. This is a bold ambition but VAST Data is a remarkable startup, certainly in the scope of its claims and use of new technology.

Beyond making flash simple, affordable and scalable - we're adding even more sizzle to this (container) steak:

  • Our NFS-over-RDMA support is critical for customers who need better-than-TCP levels of performance from NFS... we find this in places where people have historically opted for SAN storage or where parallel file systems have been the only way to really open up the pipe to a containerized server. HPC and AI are the obvious candidates for this acceleration, where we eliminate the kernel-side complexity of parallel file systems, but we also see this capability being used in enterprise environments to accelerate big data, content delivery, databases, search and more. At nearly 9GB/s of single-mount performance per each 100Gb connection, NFSoRDMA becomes the easy button that makes it possible for containerized workloads to no longer worry about stream rates.

  • Server pools also help... since we typically never find single-application container environments, but rather a grid or farm of k8s pods that are distributed over commodity computing infrastructure. Moving to shared storage has been a bit of a nail-biter for customers who previously used DAS where each application enjoyed its own dedicated storage devices. By segmenting VAST file/object servers into pools and dynamically allocating pools of server HW to each k8s client, VAST's disaggregated, shared-everything (DASE) architecture eliminates any internal east-west traffic and makes it possible to provide exclusive ingress and egress performance to demanding k8s applications.

The VAST Data Platform for Containers summarized: - as fast as a NVMe all-flash SAN - dedicated performance that rivals DAS - A simple enterprise NAS appliance delivering multi-protocol NFS and S3 - as affordable as object storage

If your application is being refactored in 2020 for cloud readiness, we would humbly suggest you don't build it on 1992 technology. #nomoretiers

More from this topic

Learn what VAST can do for you
Sign up for our newsletter and learn more about VAST or request a demo and see for yourself.

By proceeding you agree to the VAST Data Privacy Policy, and you consent to receive marketing communications. *Required field.