A web-scale application requires high performance and massive scalability to support a large number of simultaneous users with an outstanding user experience. They must be scalable on-demand while delivering great performance, agility and resiliency. Today’s modern applications either have hyper-scale requirements from day 1 or may grow rapidly in the future. It requires a scalable, future-proof application delivery platform for load balancing, web application firewall and application analytics, which is immensely challenging. Businesses are always on the lookout to save on their operational footprint, while providing the best in class service for their end customers.
How are web-scale applications deployed – Legacy Mode
Traditionally, the approach to deploy an application follows these practices
- Application servers are fronted by custom–built, very expensive hardware load balancers
- Capacity is statically provisioned and managed via the use of multiple Active-Standby load balancer pairs
This legacy approach does not work well to achieve a web-scale application deployment, as you can clearly see that it presents some daunting challenges and inefficiencies
- Underutilization of critical datacenter resources such as power and rack space
- Overprovisioning of capacity, as the standby appliances are never utilized
- Complex operations as each entity needs to be independently managed and operated
- Error prone maintenance includes capacity management and rolling upgrades
These need to be addressed to truly realize the full potential of delivering a web-scale application and enabling businesses to use intelligent load balancing as part of in their digital transformation journey.
Enabling the software-defined vision – Intel Powered
Intel has invested heavily in its R&D and over the years, Intel’s Xeon Scalable processors have leapfrogged the commodity server performance landscape, offering highly powerful servers with various instruction offload capabilities: crypto via AES-NI as an example. This has enabled servers built with Intel’s Xeon Scalable Processors to rival the performance of custom–built specialized hardware. Additionally, Intel’s network NIC adapters coupled with DPDK packet pipeline, provide high network throughput and packet processing for network thirsty applications. This has become a key enabler in designing modern, agile, scalable data centers and clouds, where racks of high-performance servers are deployed to tackle a myriad of tasks.
With the hardware substrate now built on scalable high–performance Intel servers, software solutions can deliver much more than what typically was done through specialized hardware boxes and offer incredible agility, performance and scalability, thus, enabling a true software-defined infrastructure both for data centers and clouds.
Web-scale application delivery – Avi Networks
With Intel’s high-performance CPUs and VMware NSX Advanced Load Balancer (formerly Avi Networks), you can easily deploy and manage capacity for modern applications in any environment (data centers and clouds) with an intelligent, elastic elastic, intelligent load balancing fabric.
Multi-cloud load balancing – Nobody likes to micromanage
Avi Networks platform is built on a software-defined architecture. Avi Controller is the central control plane that governs application deployment through application-specific policies.
Avi Service Engines are the distributed data plane with the configuration pushed to them by the Avi Controllers. They also collect real-time inline traffic metrics to form an intelligent closed loop of analytics.
Avi Controller (often in a cluster of three for HA) manages the entire lifecycle for the applications as well as the infrastructure, which includes but not limited to
- Creating/Deleting the Avi Service Engines as needed based on the application capacity requirements. In other words, scaling in and out the load balancing capacity on-demand.
- Programming the correct networking configurations and allocating IP addresses to the Avi Service Engines to forward traffic
- Providing non-disruptive, granular rolling upgrades of the Avi Service Engines
An intelligent and elastic fabric – Say goodbye to overprovisioning
Avi Networks enables application delivery in a truly elastic, active-active fashion. Load can be evenly distributed across all Avi Service Engines. Avi Service Engines act as Lego blocks which can scale either vertically and/or horizontally to realize the expected application scale. This allows for efficient capacity utilization.
This elastic application services fabric is fully self-healing. Avi Controller will automatically create additional Avi Service Engines if additional capacity is needed or if any existing Avi Service Engines fail. It can automatically rebalancing load without affecting in-flight traffic.
Avi Controller delivers autoscaling capabilities based on KPIs such as network packet rate, throughput, CPU utilization etc. Application capacity pool can automatically grow and shrink based on capacity utilization. This unleashes the full power of theelastic load balancing fabric.
Multi-Cloud – Make any app run on any cloud without a ceiling!
Avi Controller natively integrates with all major infrastructure providers to provide fully automated application delivery. Orchestration is provided for private datacenter providers such as VMware/OpenStack, public clouds AWS/Azure/GCP as well as container orchestration platforms such a K8s and OpenShift. This allows any app on any cloud to achieve any scale with the same fully automated, zero-touch, elastic application provisioning. See this whitepaper on how to achieve 1 million SSL TPS in public cloud.
Analytics – Take the ‘trouble’ out of troubleshooting
Avi Service Engines sit in line with applications and collect important KPIs which are then processed by the Avi Controllers to provide a rich set of application insights. These are delivered through a large array of metrics and application logs. Avi Controllers also provide seamless integration with all the leading monitoring/automation solutions such as vRO-vRA/ Ansible / Terraform / Splunk / Grafana / appviewX. This enables application admins to observe, triage and secure faster.
Third-party Load Balancing Benchmark Report – How to Achieve 1 Million SSL TPS
Avi Networks has partnered with Intel to showcase the power of a truly scalable, high-performance, elastic application delivery platform. Principled Technologies – an independent third-party testing company demonstrated load balancing of over 1 million SSL TPS (transactions per second) for a web application. You can learn more about the test in this whitepaper.
To learn more, join the upcoming joint webinar with Intel on May 20, 2020.