Nutanix is directly targeting competitor VMware with its latest product announcement: software-defined networking. The vendor announced some enhancements to its software last week that make it easier to configure network policies using software, and show administrators what’s happening inside the virtual and physical networks in a Nutanix-controlled box.
Hyperconvergence traditionally focused on collapsing storage and compute hardware together in a tightly-integrated architecture. Instead of having a separate storage area network (SAN) to manage a virtualized pool of storage, hyperconverged appliances put the storage alongside the computing resource on each node.
This effectively returned to a direct-attached storage model, in which the storage was as close to the compute server as possible. The difference is that the hypervisor controlling the nodes pools their storage together so that data can be distributed between all of them.
“The data store just gets bigger when you add a node,” explained Richard Fichera, vice-president and principal analyst serving infrastructure and operations professionals at Forrester.
Adding storage to a SAN typically takes detailed configuration by a specialist storage administrator, but things are different with software defined storage, he added. “It takes 15-30 minutes to add a node and half a dozen mouse clicks and then you suddenly doubled your storage.”
Software-defined storage (SDS) takes care of the storage pooling, but the network managing the communication between the virtual machines is still configured on a largely manual basis. Collapsing the network functionality into software is the next part of the challenge, and that’s what Nutanix started doing last week.
Prabu Rambadran, senior director of product and technical marketing for Nutanix, has a vision of breaking down all of the silos that currently exist in data centre infrastructure.
“Compute, storage, virtualization, networking, all of these things are problems that have to be solved,” he said.
The company is including APIs in its Acropolis hypervisor software that enable it to update third-party hardware such as top-of-rack switches and firewalls. The idea is to define networking policies in these devices via software, based on key events in the application lifecycle. When an application is deployed, administrators may want to set a networking policy that all its queries are routed through a firewall service, for example.
The API capability will ship in version 5.0 of its software, and a subsequent release will support the stringing together of network services that can be used to create network flows for specific applications. These workflows will be managed via the Prism software.
In the next version of its Prism hypervisor, the company will also provide a network visualization feature that lets administrators see how virtual machines are connected via the physical networking cards in the system and the virtual switches that each node uses to route communications.
The company is also planning to include support for microsegmentation in a forthcoming version of Acropolis. In a blog post on the topic, Rambadran pointed out that applications are becoming smaller and more modular. This is his hat-tip to the burgeoning microservices trend, in which applications are carved up into small parts that then interact with each other over the network.
This modular application structure creates potential security risks as virtual servers exchange traffic between each other (known in datacentre parlance as ‘east-west’ traffic). The networking within the Nutanix-controlled box will be able to monitor and apply policies to this traffic using software firewalls.
SDN is already something that VMware has explored in the NSX networking product that ships as part of its EVO Rail hyperconverged unit.
“With things like NSX you can start doing software-defined networking which will help them use standard network interconnects to make them smarter,” said John Abbott, founder and research vice president at 451 Research.
SDN in products like these will enable administrators to avoid the networking bottleneck problems that can emerge as you scale out a system with more nodes, he added.
