In a recent blog I wrote about why OpenStack in and of itself is not a service orchestrator but a cloud management system (CMS) that abstracts out the complexity of a heterogeneous data center making it easier for administrators to run applications on fairly large scale compute resources. OpenStack and other CMSs such as VMware and CloudStack fall into the virtualized infrastructure manager (VIM) category of the ETSI Network Function Virtualization (NFV) framework. This framework is one of the methods for providing structure around a multi-component and multi-vendor approach to bringing modern day IT solutions to the carrier world.
However, OpenStack and other traditional CMSs lack Enhanced Platform Awareness (EPA), therefore, they can blindly deploy virtual machines and virtual network functions (VNFs) with regardless of the performance requirements or hardware capabilities of the server resulting in:
1. Inefficient use of compute resources
2. Not being able to achieve the carrier grade performance requirements
Carrier grade NFV, and the services that are deployed based on it, requires adherence to strict SLAs to ensure high-performance and availability. Why? Because NFV-enabled or enhanced services must provide better performance than the previous mode of operation to ensure adoption. The trick is that in order for VNFs to deliver on these carrier grade requirements they need to be properly deployed in the underlying infrastructure to leverage the latest hardware features such as hardware acceleration.
NFV Management and Orchestration (MANO) and the associated information models, describing both the infrastructure and VNF requirements, are key to achieving this goal effectively and in a cost efficient manner for the service provider. Together, in cooperation, the NFV orchestrator and the CMS work together to ensure high-performance, massively scalable NFV deployments.
To prove this use case, Telefonica, Cyan, and Red Hat worked on a concept called “Deterministic NFV”, which was announced back in May 2014. The goal: to develop an architecture that allows the deterministic placement of VNFs using an intelligent NFV orchestrator that can digest standard information models describing the underlying hardware infrastructure and VNF capabilities to deliver carrier grade services. The resulting Enhanced Platform Awareness capability allows Cyan’s Blue Planet NFV orchestration platform to intelligently deploy VNF workloads onto the underlying infrastructure, enabling the optimal performance and SLAs. This also unleashes the favourable total cost of ownership that NFV promises due to this more efficient use of the underlying infrastructure.
These infrastructure capabilities include features such as open source software libraries like DPDK (Data Plane Development Kit), VMDq, SR-IOV, PCIe pass through, NUMA awareness, CPU pining and others that have enabled standard high volume servers to deal efficiently with edge functions such as BNG, PE router and EPC workloads. The key to achieving the carrier grade performance is to correctly model the key attributes required by the VNFs and exposing this information as the deployment decision criteria in the NFV delivery stack, i,e., the NFV Orchestrator and the VIM. The availability of such NFV-ready orchestration components together with appropriate standardized descriptors, such as TOSCA, will be key to enable large-scale, high-performance NFV deployments.
This year, at Mobile World Congress 2015, Cyan, Telefonica, Intel, Brocade, and Red Hat have collaborated to show a complete ETSI-NFV end-to-end service deployment solution stack at Mobile World Congress 2015 that includes this intelligent placement. The figure below illustrates how all the components come together using the ETSI NFV framework.
As illustrated in the figure above, in this collaboration:
- Cyan has provided the multi-vendor Blue Planet NFV Orchestrator
- Intel has provided COTS servers, NICs as well as the latest DPDK
- Brocade has provided an OpenFlow switch and the Vyatta vRouter
- Telefónica has provided their VIM as well as their TIDGEN traffic generator
- Red Hat has provided RHEL7.0 and the KVM hypervisor
The demonstration provides two separate deployment environments for comparing an NFV-optimized and an un-optimized cloud-based deployment:
1. An NFV-ready compute, networking and storage infrastructure, with a Telefónica developed NFV ready VIM implementing the requisite Enhanced Platform Awareness (EPA) and a Cyan NFV-Orchestrator supporting advanced VNF deployment using enhanced NFV information models.
2. A standard cloud infrastructure - vanilla cloud computing, with the same Telefónica VIM connected to the same Cyan NFV-Orchestrator but in this case not using the enhanced information model.
The results show the phenomenal benefits achievable utilizing NFV optimized components, appropriate information models and an intelligent NFV orchestrator. The service deployed in this demo shows up to a 100x improvement in throughput for a typical routing scenario with respect to the same scenario in a typical cloud deployment. Join us at the Intel booth #3D30 at Mobile World Congress, March 2-5 for a demonstration of this ground breaking technology.