Deploying a scalable virtualized mobile infrastructure on OpenStack: A use case

The traffic on mobile telecom networks continues to grow rapidly, driven by demand for mobile data and video content. Globally, mobile data traffic is expected to grow tenfold from 2014 to 2019, according to the Cisco Visual Networking Index. These networks no longer just support people, they also have to support the Internet of Things (IoT) and new applications such as virtual reality. Usage is unpredictable, so the infrastructure has to be elastic and far more scalable. Mobile operators want to be able to deliver agile, robust services that successfully meet that demand. One way to do that is to transition from the traditional, purpose-built and hardware-based mobile packet cores to open, virtualized solutions based on NFV. To get it right, operators need operator/carrier-class solutions that are scalable, manageable and cost-effective.

Red Hat and Cisco have put together an NFV solution that several large mobile operators are currently trialing and that can serve as a framework for 5G mobile packet cores. The solution uses VNFs from Cisco, including VNF Manager and NFVO and Red Hat’s NFV Infrastructure (NFVI), which includes technologies based on upstream projects like OpenStack, CEPH Storage, KVM hypervisor, QEMU, libvirt, SR-IOV and OVS (Open vSwitch).

This NFV solution enables a service provider’s network with an orchestrated virtualized infrastructure that decouples network functions from dedicated and proprietary platforms so that carriers can benefit from higher efficiencies, reduced costs and increased service velocity. By using commodity, commercial off-the-shelf servers, Cisco’s VNF Manager and NFVO, and Red Hat’s NFVI, the solution can provide a liquid pool of dynamically allocated resources. These resources are generally compute (CPU and memory), storage and network (I/O). This pool of resources can be allocated to any function at any time, enabling the telco to easily respond to spikes in demand for a particular system. The reason this level of flexibility is important is because of the diverse nature of traffic content and their context. For example, machine to machine (M2M) applications require massive amounts of sessions with minimal throughput;  therefore, they may need large amount of memory but less amount of network bandwidth. Other applications such as business analytics may require deep packet inspection with policy controlled sessions and advanced contextual analytics. Such applications will require lots of CPU capacity and probably decent amount of storage.

For telcos, performance, high availability and security are non-negotiable. The mobile packet core network has some of the stringent requirements. These include support for:

  • data plane latency of 50ms;
  • gateway bandwidth of 10-100Gpbs;
  • number of concurrent sessions of 1-100Millions;
  • user plane packet of 5M packets/sec;
  • variable packet size between 64 to 1500 bytes;
  • jitter of less than 30ms;
  • user bandwidth of 10-50Mbps;
  • packet loss of less than 1 percent; and
  • high availability that includes application level stageful redundancy.

In LTE Plus we have started to see control and data plane separation in the mobile packet core. For example, the telco can virtualize an array of systems that have both control and data planes, such as such as IP router, firewall and mobile network gateway. These network elements will likely have separate VM or containers for their control planes than from their data planes. This will enable much better flexibility in terms of the types of resources that are consumed by these different types of workloads.

NFV can help telcos succeed in a fast-paced, highly competitive market. To meet their needs for performance, high availability and security, Cisco and Red Hat have implemented several features within the VNF Manager, NFVO, VIM (based on OpenStack) and NFVI (which includes CEPH Storage, KVM hypervisor, QEMU, libvirt, DPDK, SR-IOV and OVS) that specifically address all of the requirements identified above. These added features include CPU pinning, memory huge pages, non-uniform memory access (NUMA) affinity, SR-IOV and jumbo frame support. With this pre-validated design, simplified operations, innovative and carrier-grade support, the Red Hat and Cisco NFV Infrastructure solution lets telcos virtualize their network faster and more reliably.

By the way, we presented on this topic at the recent OPNFV Summit 2016, and you can watch the full presentation. Here’s the link to “Deploying Virtualized Mobile Infrastructures on Openstack.” Take a look!

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s