Myth Busters: Load Balancers Need Specialized Hardware

Grant Swanson
Posted on Jan 6, 2016 12:35:12 PM

The Application Delivery Controller (ADC) market is very competitive, where every vendor claims that they have a performance advantage. Most IT Pros who need high performance ADCs and Load Balancers start by looking at specialized hardware appliances in order to meet their requirements. Then performance testing for L4-L7 throughput takes place and folks quickly realize that as more features are turned on, the performance quickly degrades. Forward-looking ADC buyers are now starting to look at software solutions that can run on commodity x86 hardware and that offer better performance at a fraction of the cost of specialized hardware. 



If you choose the path of specialized hardware for 1 Tbps of throughput, you will undoubtedly spend a small fortune.  For example, the leader in the ADC market is F5 and their Viprion 4800 would be a choice for this type of throughput requirement. With no features turned on and the system running on ASICs with L4 traffic, you would need two fully loaded 8-blade chassis. As you turn on features and CPU cycles get eaten up, you will inevitably need to add another fully loaded chassis. Now you are looking at a couple of million dollars! 

Now if we take a look at 1 Tbps of L7 traffic, 3 fully loaded Viprion 4800 chassis - $3M list; with support, the 3-year TCO is near $5M, and that's without high availability (HA). 


One of our customers recently purchased software load balancers from Avi Networks to deploy on bare metal servers. They were stunned by the fact that we were able to meet their performance requirements of 1 Tbps throughput on commodity hardware while also getting unprecedented application insights through the  Application Performance Monitoring, Application Security, and zero-touch autoscaling. With complete, centralized visibility to application performance and transactions on their load balancers distributed across their entire data center they were able to drive down their troubleshooting efforts from days to mere minutes. What's more, they are now ready to drive a consistent application delivery architecture across their private and public clouds with an ADC solution that simply works identically across any environment. 

The power of application services delivered on commodity x86 servers is real and it is here. With the performance now available on x86 servers, you could achieve the same 1 Tbps throughput on a just few hundred CPU cores. A 2-socket, 20+ core server costs less than $10K, so you can do your math! We are talking TCO savings greater than 70%.  Our customer found that they would actually spend less than the cost of annual maintenance for their legacy ADC for a fully functional, performant ADC solution that works in any environment while enabling ticketless IT through self-service and automation for app owners from a fully API-driven platform. The Web Scale advantages of Amazon or other cloud giants (many of whom are in our Silicon Valley backyard) who have traditionally used x86 to build their hyper scale environments are now available to all enterprises.

If you want to take a look at how this works, join us for our weekly “demonar” on Tuesdays at 11:00 am PST.


Topics: ADC, Architecture, Load Balancing

New Call-to-action

Subscribe to Email Updates

Recent Posts

Posts by Topic

see all