Do you want to improve your OpenStack-based cloud performance by 15x? Well Charlie Ashton from Wind River has the answer here…
Guest blog by Charlie Ashton.
As communications service providers (CSPs) explore architectural choices for virtualising their network infrastructure, OpenStack has emerged as the leading option for the cloud management software that manages the lifecycle of Virtual Machines (VMs).
There are sound business reasons for this selection: OpenStack is an established open-source community project with tens of thousands of contributors worldwide, even though most of them are working on enterprise rather than telecom components, while the availability of multiple OpenStack-based solutions mitigates the risk of vendor lock-in.
During their evaluations of OpenStack-based platforms for telco clouds, CSPs typically focus heavily on four aspects of a platform:
By providing best-in-class functionality in all these areas, the Wind River Titanium Cloud portfolio of OpenStack-based virtualisation platforms has emerged as the industry’s leading solution for telco network virtualisation.
While Wind River blogged extensively about all aspects of Titanium Cloud since its launch in 2014, a recently-published white paper written by GENBAND and Wind River contributes additional interesting, real-world information to the topic of system-level performance.
To evaluate the performance difference between ‘vanilla’ (i.e. unmodified) OpenStack and an optimised open source solution, two different OpenStack Neutron networking implementations were tested with GENBAND’s Advanced Media Software (AMS) Virtual Network Function (VNF) software. The paper compares the performance of the GENBAND AMS VNF with vanilla OpenStack in one environment and Titanium Cloud in another environment.
The tests analysed call load, packet loss, CPU utilisation and the amount of CPU cycles required for each stage of processing. Metrics were collected in real time by the iMedia component of the GENBAND AMS VNF and displayed in dashboards leveraging the Representational State Transfer (REST) application programming interfaces (APIs) within the VNF Manager.
Summarising some of the key results, the vanilla OpenStack environment processed 200 calls while the Titanium Cloud environment handled 3,000 calls, an improvement of 15x achieved with no difference in hardware cost.
The difference in CPU utilisation was also significant between the two environments: 12% utilisation for the vanilla environment while 60,000 packets were being processed, compared to only 3% utilisation for Titanium Cloud while processing 400,000 packets. This frees up significant CPU resources for other, revenue-generating applications and services.
The white paper provides details of several other metrics that were evaluated and analysed, including the call processing rate, the jitter associated with CPU utilisation and the packet loss experienced by the system.
As you’ll see from reading the complete paper, the bottom line is that Titanium Cloud achieved a 15x performance improvement, processing 3,000 calls compared to 200 by vanilla OpenStack, while also delivering better call quality.
As the paper explains, these numbers are important not only because of the pure technical advantages that they enumerate but also because of the resulting business benefits delivered to CSPs. By selecting Titanium Cloud as their network virtualisation platform, they can minimise both their CAPEX and their OPEX, while at the same time maximising their top-line revenue by guaranteeing the level of service uptime that their customers expect. And all while mitigating the risk of vendor lock-in because Titanium Cloud is a 100% open solution compatible with all the relevant open standards, validated through its extensive partner ecosystem.
You can download the whitepaper below and can contact Wind River here for more.
Courtesy of Wind River.