New Combo: Software Defined Storage, Networking
As the the ongoing trend toward the software defined data center (SDDC) and the software-defined storage (SDS) gains momentum, the lines are clearly blurring between the traditional silos of networking, storage and servers. Given this new combining of traditionally separate sectors, this article – yes, on a site that covers data storage – will look at some top solutions in software defined networking.
That said, I’ll bring a storage perspective to SDN and show how it combines networking and storage into a Vitamix blender. Whether the resulting smoothie has a taste that delights the user, of course, remains to be seen.
Greg Schulz, an analyst at StorageIO Group, thinks the SDN term itself is misleading. He suggested a better term might be software defined convergence or software-defined unified cloud computing. For storage interoperating with other SAN and LAN technologies, it is all about adding management value with more integration and the server, storage I/O networking as well as upper layer applications. Could that one day go as far as full convergence of storage and networking?
“We are eliminating the need for customers to build two separate networks,” said Soni Jiandani, Executive Vice President of Insieme Networks (recently acquired by Cisco).
So let’s take a look at some of the vendors in SDN. In addition to Cisco – the big kahuna in the space – we cover Pluribus Network, Cumulus Networks and Big Switch.
Cisco’s Application Centric Infrastructure (ACI) architecture uses policy-based automation to speed the deployment and management of applications. Instead of separate IT teams working in serial fashion to define the requirements and resources for compute, network and storage that a new application will need, ACI aims to have them work in parallel or as one team to define requirements and automatically distribute them.
Shashi Kiran, Senior Director, Market Management for Data Center, Cloud and Open Networking at Cisco, laid out several examples of how it could facilitate better storage. As ACI provides deeper telemetry capabilities and a single view of the application health, the state of the storage subsystem is better understood. This can simplify troubleshooting and reduce potential application downtime by moving workloads to healthier components. In addition, ACI provides a way to automate storage provisioning through a single policy framework.
Security is a side benefit. ACI enables isolation of storage ports, enforcing storage port-to-port communication policies, and running multiple authentication schemes across the storage subsystem based on policy to enable secure multi-tenancy capabilities.
“Think of Cisco ACI as a network architecture that takes advantage of software-defined networking but goes beyond the limitations of SDN through combined innovations in hardware, software, and ASICs,” said Shashi Kiran, Senior Director, Market Management for Data Center, Cloud and Open Networking at Cisco. “It is the first data center and cloud solution to offer visibility and integrated management of both physical and virtual networked IT resources. ACI accelerates application deployment cycles to drive faster business processes and improve bottom line results.”
However, Schulz believes Cisco has its work cut out to sell this concept. He said the company needs to move above the software-defined buzz.
“Just as the Cisco UCS converged system has become a key part of many storage environments, as have also their Nexus solutions, ACI builds on those while elevating the conversation up a few layers to start addressing the applications and other software topics that defined how data infrastructure hardware and software get deployed,” said Schulz.
Big Switch offers two SDN solutions - a Monitoring Fabric and a (currently beta) Cloud Fabric. The Monitoring Fabric is a multi-tenant network fabric that connects network taps or span ports in the production network to performance monitoring, security and network troubleshooting tools. The company uses its SDN software stack in tandem with hardware switches from other manufacturers. Cloud Fabric is a data center switching fabric for OpenStack environments. It is a controller that connects to both physical switches as well as hypervisor vSwitches.
“This approach of using bare metal switches and SDN designs is solving the cost plus automation problem faced by public and private cloud operators looking for Amazon EC2 economics,” said Kyle Forster, Co-Founder, Big Switch.
He explained that the tie-in to SDDC and SDS is via Cloud Fabric, which automates some of the tricky areas of setting up networks for cloud storage. Cloud storage traffic, particularly in OpenStack, has very specific traffic patterns across the network as VM images and tenant storage can compete with tenant and provider side network traffic for resources.
Cumulus Linux is said to be a Linux operating system for the physical layer that allows businesses to scale data centers and better automate networks. It is an operating system for networking hardware that is hardware agnostic and supports traditional and modern network architectures. Users can leverage existing Linux kernel data structures for integration with open source/commercial Linux applications. Since it is an OS for the physical layer, other SDN technologies can run on top of Cumulus Linux. The pricing structure is an annual subscription based on per switch instance (not per port).
“Cumulus Networks decouples the Operating System from the networking gear, which is a key piece in the broader shift to the software-defined data center and creates flexibility,” said Reza Malekzadeh, Vice President of Business, Cumulus Networks. “The company is backed by venture capital firms Andreessen Horowitz, Battery Ventures and four of the original VMware founders, including Diane Greene, President, CEO and co-founder of VMware until 2008.”
Pluribus Network has created a Distributed Network Hypervisor (Netvisor) that runs on its F64 silicon-based server switches and white box switches. This allows the server switch to replace the top-of the-rack switch and operate at sub micro-second latencies. The F64 Server-Switch integrates compute power, up to a TB of RAM and also optionally integrates 6 TB of PCIe-based flash storage or SSD-based storage.
“The Pluribus Netvisor and Server-Switch Fabric is managed as a single switch, which reduces overall OPEX by simplifying management,” said Sunay Tripathi, co-founder and CTO, Pluribus Network. “Pluribus uses off-the-shelf switch and server chips from Intel and Broadcom.”
Photo courtesy of Shutterstock.