As a cloud management framework OpenStack thus far been limited to the province of telecommunications carriers and providers of Web-scale services that have plenty of engineering talent to throw at managing one of the most ambitious open source projects there is. In contrast, adoption of OpenStack in enterprise IT environments has been much more limited.
But that may change as more advanced networking technologies that are optimized for processor-intensive virtualization come to market. Some of the technologies we have covered here include single root input/output virtualization (SR-IOV) and Data Plane Development Kit (DPDK). Another technology includes using field programmable gate arrays (FPGA) in Network Interface Cards, to make them smarter about how to offload virtualized loads.
Speeding Up OpenStack Hardware
New technologies for speeding up the performance of OpenStack and other cloud environments are coming to market all the time. For example, Netronome has added support for Express Virtio (XVIO) to its Agilio Server Networking Platform in a way that makes it possible to get performance similar to SR-IOV network interface without having to sacrifice the ability to migrate virtual machines.
Sujal Das, chief strategy and marketing officer at Netronome, says XVIO is a better alternative to rival approaches such as the Data Plane Development Kit (DPDK) originally developed by Intel, because XVIO requires no changes to be made to applications and VMs. In addition, Das notes that DPDK cannot support popular and widely accepted Kernel-based networking innovations being used in OpenStack environments.
Finally, XVIO, adds Das, makes for much more efficient use of multicore processors, which over time drives better network performance at a much less total cost.
In general, Das notes that not only will most cloud and hyperscale IT service providers benefit from being able to use XVIO to offload network services more efficiently from servers, managed service providers looking to remotely manage OpenStack environments running on premise or hosted in a cloud will benefit as well. In fact, that capability might be critical in terms of expanding the footprint of OpenStack in the enterprise because of the work being done by those organizations to make OpenStack simpler to deploy and manage at scale.
“We think there will be a lot more deployment of OpenStack in the enterprise in late 2017,” says Das. “Intel refers to this as the next 10,000 clouds.”
Netronome today is focused squarely on OpenStack. But Das says that because VMware also relies on open source networking software he expects that Netronome will be able to apply its approach to XVIO to VMware environments.
An OpenStack Managed Experience
In the meantime, Boris Renski, chief marketing officer for Mirantis, a provider of an OpenStack distribution, says that in terms of driving OpenStack adoption into the enterprise managed services are crucial because most internal IT organizations don’t have the skills required to either build or manage an OpenStack cloud environment.
Most recently, Mirantis partnered with NTT to deliver both the managed OpenStack experience as well as the IT infrastructure it runs on. Renski says Mirantis is open to working with other managed service providers to take responsibility for delivering those services to Mirantis customers. To date, however, Renski says Mirantis has not found any partners capable of delivering the service using its OpenStack distribution.
“Frankly, nobody’s asked us yet,” says Renski.
Whoever winds up delivering OpenStack as a managed service is obviously going to be keen to address any network performance issues. Assuming those issues do get resolved the irony might very well be that networking within OpenStack environment could soon be substantially faster than in any other virtualization platform.