Astonishingly, a little over a year has passed since F5 launched the BIG-IP VE for SmartNICs solution, which as a reminder, is the product of an integration between the BIG-IP Virtual Edition and an Intel SmartNIC that elevates performance and efficiency while lowering costs. During this time, both the internet and application landscape have continued to rapidly evolve, with the data points below collectively hinting at not only the relevance of this solution, but also the significance of the new use cases we’re announcing here today (more on those shortly – stick with me):
A genius it does not take to establish that such a sizable rise in internet users will result in greater demand for applications and services. Supporting this expanding user-base will not only require workloads to be scaled up and out, but the encompassing networking infrastructure will likely require levelling-up as well to cope with heightened traffic volumes. And with the majority of users now using a larger medley of both mobile and fixed-location devices, guaranteeing a high-quality experience regardless of client-type is more vital now than ever before, as data point two confirms:
Hyper-fast web and mobile experiences are now considered table stakes for today’s applications. Since users will not accept anything less, organizations are under increasing pressure to architect their application environments for speed and reliability to mitigate against losing customers and revenue. Beyond operating the highest performance servers and lowest latency routers and cabling, many are now electing to move their services and content closer to end users to improve experiences as our final data point indicates:
In summary, to exceed user expectations IT teams will require networking solutions with higher performance, greater reliability and lower latency, that also offer the flexibility to be deployed anywhere from centralized data centers to distributed edge locations. In essence, organizations require the likes of F5’s BIG-IP VE for SmartNICs solution.
For those unfamiliar with the BIG-IP VE for SmartNICs solution, a high-performance Field Programmable Gate Array (FPGA) enabled SmartNIC is used to offload specific compute intensive functions from a BIG-IP Virtual Edition. As BIG-IP VE’s are usually deployed atop common-off-the-shelf (COTS) servers that offer performance targeted more towards general purpose workloads, the BIG-IP VE may in some instances not have sufficient compute power to effectively carry-out resource-intensive tasks such as DDoS mitigation. SmartNIC usage, however, not only significantly boosts system performance but also alleviates strain on BIG-IP VE’s compute utilization.
Building out from the original DDoS use case released with the solution’s launch last year, F5 is pleased to announce that both Carrier-Grade Network Address Translation (CGNAT) and Layer 4 Traffic Acceleration (Layer 4 load balancing for example) can now be augmented via this SmartNIC integration. Figure 1 below explains at a high level just how the SmartNIC integrates with BIG-IP VE to deliver these use cases:
In fact, when performing CGNAT and L4 traffic processing functions within the SmartNICs’ FPGA, solution testing has shown that total traffic throughput can be increased by around 30% and VE compute usage may be cut from around 80% to as little as 5%. Not only does this make the solution capable of handling greater volumes of user traffic, but it also improves system reliability while affording the BIG-IP VE extra compute cycles to dedicate to other security and traffic management tasks.
A further benefit of handling these functions within a SmartNIC is reduced latency. As depicted in figure 1, qualifying connections now effectively bypass the BIG-IP VE meaning that around 500μs can be shaved off total connection times – a small but significant saving especially for time-sensitive markets like finance and healthcare, or media organizations providing live streaming over edge Content Delivery Networks (CDN).
The BIG-IP VE for SmartNICs solution is available as an add-on to new, or existing High Performance BIG-IP VE’s and looking forward, F5 will continue to develop and deliver incremental use case support within future releases. For more information about any of the use cases touched upon here and more solution specific details, take a look at this Solution Overview (add link).
It’s an unthrottled version of BIG-IP VE that’s licensed based on the number of virtual cores attributed to the VE. i.e. instead of buying a 10Gbps VE where you’re limited to 10Gbps throughput, you can buy a 8vCPU High Performance VE where you can use 8 virtual cores and the throughput is not limited by F5, but rather by what’s possible by the compute power available. They’re available from 8 to 24vCPU in 4vCPU increments (8,12,16,20,24).