Cloud Peak—Virtual Cloud Infrastructure Validation

Problem: production VNFs perform poorly and require costly over-provisioning

Network Functions Virtualization (NFV) is a key building block for 5G mobile networks. In an NFV architecture, various Virtual Network Functions (VNF) and Cloud-Native Network Functions (CNF) run on a common NFV Infrastructure (NFVI) and are controlled by a common management and orchestration (MANO) layer. This has made new use cases such as network slicing and mobile edge computing  (MEC) possible.

The flexibility that enables these new 5G use-cases creates major challenges for carriers as they seek to deliver high-quality services to their customers. NFVI must be dimensioned and configured correctly, and thoroughly validated to ensure the performance of the VNFs running on top of it. Complicating matters, the NFVI will be running diverse simultaneous workloads that interact with each other, leading to massive and costly over-provisioning.

How can carriers validate such a complex system—composed of multiple compute, network, and storage resources—and its interaction with the rest of the NFV architecture elements?

Solution: NFVI validation through VNF workload generation

Ixia’s Cloud Peak is a web application designed to rigorously benchmark the performance of virtualized network infrastructures. By deploying real Virtual Machine or Docker Container workloads on top of the NFVI system under test (SUT), the application provides key insights into the capability of the NFVI to sustain the required VNF and CNF workloads.

Using an innovative methodology that decomposes system resources into the elementary compute, network, and storage categories, the application can isolate configuration issues and performance bottlenecks. The test results are compared to preconfigured performance levels to generate an easy-tounderstand Pass / Fail result, displayed through an intuitive real-time user interface (UI). Whether testing a single small server or validating a whole infrastructure with hundreds of nodes, the product can scale to match the capacity of the SUT.

• Easy-to-use solution for virtual infrastructure validation.

• Benchmark private, hybrid, telco, edge, and public clouds.

• Compare benchmark results across different infrastructures.

• Assess the impact of config changes via repeated testing.

• Find peak system capacity and oversubscription behavior.

• Run workloads in virtual machines or Docker containers.

• Validate the NFVI SUT with VNF / CNF workloads.

• Isolate compute, network, and storage resource categories.

• Start small with testing a single compute node.

• Go large with testing multiple server racks.

• Automate test execution and schedule periodic testing.

Key features

• Complete virtual infrastructure performance benchmarking for private, hybrid, telco, edge, and public clouds.

• Workloads available as Virtual Machine or Docker Container images with similar benchmark capabilities.

• Automated workload deployment via specific API interfaces on Amazon, OpenStack, VMware, K8S clouds.

• Manual workload deployment via predefined OVA / QCOW2 / AMI images for other types of clouds.

• Predefined test methodologies with unique parameters tuned for virtual infrastructure testing.

• Industry-proven workload emulation based on the open source OPNFV Yardstick portfolio.

• Custom-built workload emulations for complete virtual infrastructure benchmarking.

• Decomposes and individually validates the compute, network, storage, and VIM performance dimensions.

• Measures the VIM performance with custom VM Instantiation and VM Termination test methodology.

• Measures the scheduler capability to isolate the good workloads and the bad noisy neighbors.

• Validates the SUT from application perspective through powerful workload emulation.

• Validates the SUT from network infrastructure perspective through high performance traffic generation.

• Scales from small test beds with a single compute node to large environments with many compute racks.

• Groups the tests into Test Sessions and Test Playlists for simplifying the test configuration management.

• Executes test scenarios sequentially (single-dimensional testing) for initial system characterization.

• Executes test scenarios in parallel (multi-dimensional testing) to assess how they impact each other.

• Controls the test tool via intuitive web-based UI that offers dynamic dashboards and real-time statistics.

• Controls the test tool via REST API to automate the test execution and run repeatable regressions.

• Deploys on standard of-the-shelf virtualization platforms across private / hybrid / telco / edge / public clouds.

• Provides subscription licensing for low startup cost and flexibility of pay-as-you-grow model.