Browse categories
Find the answer you’re looking for
Overview
Standard Linux (with Linux bridge or Open vSwitch) is a common method of providing L2 functions to the virtual environment. Its drawbacks all have one point in common: the Linux kernel, which is not optimized to receive and send packets at very high rates. In fact, the transmit and receive processes compete for CPU cycles along with all the other Linux processes and applications. Let’s see how standard Linux stands up to the six points:
- Hardware independence – Good
- Infrastructure Throughput – Poor (used as a baseline)
- Networking features – Good
- Environment Compatibility (Platform independence) – Good
- Application Throughput (VM2VM) – Fair
- Live migration – Good
OVS DPDK increases throughput thanks to DPDK. Currently, only L2 features are supported and VM2VM throughput is still bottlenecked.
- Hardware independence – Good
- Infrastructure Throughput – Good
- Networking features – Poor, L2 features only
- Environment Compatibility (Platform independence) – Good (once fully supported by Linux vendors and OpenStack)
- Application Throughput (VM2VM) –Poor
- Live migration – poor
SR-IOV or PCI pass-through dedicates specific NIC ports to VMs to achieve bare metal-like performance. L2/L3 processing in the virtual switch is lost (required by OpenStack scenarios for example), and high throughput service chaining use cases are limited by the bandwidth of the PCI bus.
- Hardware independence – Poor (although most NICs are supported by standard VMs)
- Infrastructure Throughput – Good
- Networking features – Poor
- Environment Compatibility (Platform independence) – Good
- Application Throughput (VM2VM) – Poor (because hardware barrier crossing is involved)
- Live migration – Impossible
No modification is necessary to the VNF. Virtual Accelerator supports Virtio and is transparently integrated with OpenStack and Linux management tools.
Throughput is the main reason for using acceleration, but it is not the only reason. 6WIND Virtual Accelerator addresses six points:
- Hardware independence – Good, VMs are not tied to the physical NIC.
- Infrastructure Throughput – Good, dramatic increase in packet processing.
- Networking Feature Set – Good, L2 and L3 switching and networking, multi-tenancy, security, filtering and NAT.
- Environment compatibility (Platform independence) – Good, the price of performance is not tied to a single vendor solution, Linux distribution, management/orchestration, etc.
- Application Throughput (VM2VM Communications) – Good, east/west traffic throughput performance increases thanks to offload features.
- Live migration – Good, virtual environment integrity is not compromised.
- Maximize VM density by optimizing server resources without sacrificing performance.
- Enable migration to high performance, hardware agnostic virtual environments (NFVI and Data Centers).
- Bare metal performance with hardware independence.
- Reduce the number of processor resources, or cores, used by infrastructure.
- Telecom/Service Provider NFV Infrastructure: allows migration to a virtual environment and allows service chaining without incurring a performance penalty.
- Cloud Provider realizing ‘economies of scale’: in an industry where efficiency is measured in terms of VMs per sq. ft. per watt, 6WIND Virtual Accelerator increases forwarding performance and VM density by reducing processor resources dedicated to the hypervisor domain.
- OEMs: cost-effective and dramatic Virtual Network Function/VM throughput improvement.
Yes, 6WIND Virtual Accelerator is based on the proven 6WINDGate technology that has been deployed for years in Tier 1 networks worldwide.
6WIND Virtual Accelerator is an add-on software package that accelerates networking performance in virtual environments so that application performance is the same as on bare metal and is fully portable.
6WIND Virtual Accelerator removes performance bottlenecks in the virtual infrastructure and includes support for Open vSwitch, Linux bridge, multi-tenancy, security, filtering and guest processing offloads. 6WIND Virtual Accelerator is transparently integrated with OpenStack, SDN Controllers and Linux.
How to buy
Please contact 6WIND for ordering information here.
Only the IPsec option for the 6WIND Virtual Accelerator implements cryptography. 6WIND has all the required authorizations to export cryptography to all countries that are not under embargo. We have a dedicated document that details our export license.
6WIND Virtual Accelerator software is a mix of software under GPL or other open source licenses and 6WIND license. Software under open source license included in the Virtual Accelerator is listed in the Publicly Available Software document available with every software release.
6WIND Virtual Accelerator is licensed per server and priced per core running it. VM or container count has no effect on pricing.
Features
A list of supported features is available on the 6WIND Virtual Accelerator data sheet. Features include virtual switching, virtual networking and multi-tenancy, overlays, L3 and IPsec (options), multi-vendor physical NIC drivers and Virtio host drivers.
Management
Yes, it is transparent to SDN Controllers. Examples include OpenDaylight, Nuage Networks and Midokura.
Yes. 6WIND has developed plug-ins to have 6WIND Virtual Accelerator installed with Mirantis Fuel and Canonical Juju Charm.
Yes. 6WIND Virtual Accelerator is compatible with the latest versions of OpenStack.
6WIND Virtual Accelerator comes as a package for the Linux distribution and is installed through standard package manager tools (yum, APT). It then runs as a service that can be started, stopped and restarted. 6WIND Virtual Accelerator is transparently integrated with Linux, so that management and orchestration tools that support Linux, such as OpenStack, behave as usual.
Performance & Scalability
In a simple use case involving VMs, 6WIND Virtual Accelerator can provide roughly 20 Gbps of forwarding throughput per core. This translates to the capacity of a single 10 Gigabit Ethernet Full Duplex link. The performance grows linearly and is a function of the number of cores. With two cores 6WIND Virtual Accelerator supports 2x10GE FD links. To support one 40 Gigabit Ethernet Full duplex link, four cores are required. Note that this is dependent on the use case and network protocols involved. Detailed benchmarks are available on request.
Compared to standard Linux, and depending on the use case, Open vSwitch performance is increased by up to 8x, and CPU core usage on the hypervisor is reduced by a factor of 10.
6WIND Virtual Accelerator fast path technology was written with high performance packet processing on multicore architectures in mind. The algorithms are lock-less and minimize the number of cache misses. Fast path technology dedicates CPU cores to all packet forwarding and ‘offloads’ the forwarding task from the Linux kernel.
Processor Support
x86 Intel processors are supported and 6WIND will continue to support Intel’s latest processors (see datasheet for more information).
Software Platform Support
No.
Red Hat RHEL, Canonical Ubuntu and CentOS are supported. Please see data sheet for latest information.
Boards And System Support
Cisco, Intel and Mellanox. Please see data sheet for most up to date information.
6WIND Virtual Accelerator runs on standard Intel x86 servers and has been tested with Dell, HP and Advantech servers.
Documentation
6WIND Virtual Accelerator comes with a complete set of documents, including a Quick Start Guide, User Guide, Publicly Available Software, and product documentation.
Evaluation
Yes. The 6WIND Virtual Accelerator software can be evaluated. Please contact 6WIND to request an evaluation here.