NFV Accelerator

Network Functions Virtualization (NFV) is transforming the telecom industry by changing the way networks are built and operated. End-to-end Network Services (NS) like Evolved Packet Core (EPC) and IP Multimedia Subsystem (IMS), or elementary functions like edge routers and firewalls, are being delivered as software modules and instantiated in virtualized and cloud environments backed by high-volume commodity servers. This delivery model replaces the traditional approach of using fixedfunction, proprietary and expensive hardware.

A determining factor for NFV adoption is performance. Virtualized Network Functions (VNFs) running on standard multicore servers should perform at least as efficient as their legacy equivalents.

In a carrier network this requirement becomes even stronger to meet critical targets such as low latency, high throughput and deterministic behavior (low jitter), that ultimately guarantee exceptional quality of experience (QoE) for the end customer.

Intracom Telecom's NFV Accelerator is a platform that boosts the performance and energy efficiency of virtualized Network functions running on commercial of the shelf x86 servers to carrier grade levels.

  • Intelligent Network Service placement.
  • Non-disruptive Network Service collocation using programmable isolation technologies.
  • Adaptive, SLA-guided Network Service optimizations.
  • Energy-friendly deployment choices.
  • Intuitive and easy-to-use Graphical User Interface with comprehensive graphs.
  • Automatic creation of editable optimization profiles.
  • Leverage best practices from High Performance Computing/ Systems disciplines.
  • Take advantage of data analytics and machine learning techniques.
  • Exploit combined automatic discovery of hardware, software and NFV workload characteristics.

Benefits

  • Evaluate the effectiveness of Network Services based on their KPIs.
  • Increase the performance of Network Services.
  • Reduce the power consumption of NFV data centers.
  • Minimize CAPEX and OPEX costs.

Intracom Telecom's NFV Accelerator optimizes the execution of virtualized Network Services on managed Data Centers, improving performance, energy efficiency and infrastructure utilization as follows:

  • Intelligent NS placement: the NFV Accelerator makes intelligent decisions about how to optimally place VNFs on the target hardware platform.
  • Non-disruptive NS co-location using programmable isolation technologies: the NFV Accelerator utilizes state-of-the-art techniques enabling the safe physical co-location of critical VNFs.
  • Adaptive, SLA-guided NS optimization: rather than deploying a NS in a "fire-and-forget" fashion, the NFV Accelerator continuously monitors the NS' KPIs to discover room for further performance or energy improvements, or to counteract variations in traffic load and patterns.
  • Energy-friendly deployment choices: for the NFV Accelerator energy efficiency is a first-class citizen. Rather than focusing solely on performance, it considers power optimization as an additional target, leveraging mechanisms like CPU frequency scaling and server consolidation achieving a goldilocks optimization that simultaneously improves performance and energy efficiency.

Platform Architecture

The NFV Accelerator platform interacts with the following components of the NFV stack to monitor important metrics and communicate its decisions:

  • NFV orchestrator.
  • Network Services and their VNFs.
  • Virtual Infrastructure Manager (OpenStack).
  • Compute Hosts.

The platform can manage and monitor VNFs implemented as:

  • Virtual Machines (VMs)
  • Docker containers
  • Linux native applications

The latter category includes user-level Virtual Switches which the accelerator handles as a special, high-priority class of VNFs that also need careful resource allocation.

The operation of the accelerator is summarized in high-level steps, performed by the following modules:

  1. Monitoring & Capabilities module, responsible for discovering static properties of the infrastructure, like topology of resources, hardware capabilities, and NS description, as well as gathering real-time statistics of physical and virtual resources utilization, power consumption, VNF low-level metrics (e.g. packet rates/loss) and NS KPIs .
  2. Learning & Decisions module which is the brain of the platform. It processes information coming from the Monitoring Module and automatically constructs an optimization profile to meet the user-provided objectives (NS performance, energy efficiency, co-location etc.)
  3. Acting module that applies the optimization profiles mandated by the Learning & Decisions module down to the compute hosts.

Workflow

From the administrator's perspective, the NFV Accelerator Platform provides an intuitive and easy-to-use Graphical User Interface (GUI) to drive optimizations.

A typical workflow follows the steps below:

  1. Network Service Definition:
    • the administrator views the whole Data Center and drills down to the individual compute hosts running the VNFs that cooperate as part of a NS.
    • she defines a new NS, selecting the VNFs that comprise it.
  2. Optimization Profiles Generation
    • the NFV Accelerator intelligently tests different resource allocations to the NS' VNFs, and records the efficiency of each in terms of performance and power
    • it selects three of the best configurations of the previous step to make up the following predefined optimization profiles:
      1. Max Performance
      2. Balanced
      3. Max Economy
  3. NS Monitoring and Management
    • at any time without shutting anything down the administrator can switch the operating mode for a NS from default ("Operating System") to one of the previous three generated profiles. The resource allocations corresponding to the profile are enforced down to the compute host and VNFs (Figures 2 and 3).
  4. Feedback-driven Optimization
    • The NFV Accelerator continuously monitors the NS KPIs against its SLAs (Figure 4), and tests alternative configurations in the background in order to discover further improvements or address variations in traffic load.