Manage Learn to apply best practices and optimize your operations.

Network load balancer and ADC options: What's best for you?

When considering an ADC platform, take the time necessary to ensure the prospective system meets your requirements.

Editor's note: In part two of our three-part series on how to buy application delivery controllers, we'll help you understand feature sets and platforms. Part 2 picks up where part 1, Choosing the right ADC, left off.

When buying an application delivery controller (ADC), the key differentiators between $2,000 and $150,000 systems are the feature sets supported and the system performance. Let's explore some platform options and then focus on three different network load balancer or ADC deployment scenarios. Each scenario is equally valid, but each is progressively more sophisticated and requires greater processing power to deliver. Make sure to find the one that best represents your requirements, then use that to induce prospective vendors to tailor their options for your environment.

When the first Layer 4 switches appeared in the 1990s, even less intelligent Layer 2 switches and Layer 3 routers were expensive. To provide both intelligence and high performance, Layer 4 switch pioneers like Alteon Networks (sold first to Nortel and then to Radware) and Foundry Networks (now part of Brocade), created custom silicon to run the Layer 4 functions.

Custom silicon can mean higher prices

Custom silicon, whatever its performance, typically translates into higher product prices. The time and effort to develop it must be factored in, and the relative small production of the chip doesn't permit economies of scale.

Additionally, with logic essentially "burned in" to the chips, major changes could require users to wait for the next generation of silicon to be designed and manufactured in order to get significant feature upgrades.

While it is unwise to make your decision based solely on the platform, it is certainly an important factor.

While some vendors still build systems based on hardware that is proprietary (i.e. expensive), at least in part, many Layer 4 and ADC vendors have taken advantage of the tremendous price and performance improvements of Intel processors, which have dominated the processor market over the years, in conjunction with the wire speed Gigabit Ethernet interfaces available in high-production/low-cost merchant silicon.

Not only do vendors -- and therefore, buyers -- benefit from the massive economies of scale of Intel processors, but systems that use only standard hardware can provide value-add functionality in software. These system upgrades can often be installed to existing hardware, providing new features sooner and without the need to "open the box" to swap out hardware components.

While it is unwise to make your decision based solely on the platform, it is certainly an important factor, as it can impact both the cost and complexity of future upgrades. And with all things being equal, going the open route is probably a good idea.

Choosing from three ADC deployment options

When deploying ADCs, you typically have multiple options with respect to the deployment platform. This will vary by vendor, of course. The main ADC equipment options are: proprietary or standard, physical or virtual, and on-premises or off-premises.

If a vendor has built its ADC using custom silicon, you'll be running it on its proprietary platform. Vendors that provide all of their functionality in software (as opposed to providing it via custom silicon) running on standard Intel processors certainly bring you the benefits of a less expensive platform -- whether or not they deliver an integrated environment or just software for you to deploy on a server.

An additional benefit of buying ADCs built on standard hardware is that you can then decide to position your ADC as just another virtual machine (VM) running on your VM host. Understand that you will need to allocate sufficient resources to this VM so it doesn't become a bottleneck. With proprietary hardware-based approaches, you can't use common virtualization environments like VMware to virtualize your ADC.

Determining the correct location for ADCs

You may have a choice of where to locate your ADC. Because the ADC functions best when local to the server, positioning will be dictated by where you locate your servers. If the servers are located on-premises, that's where the ADC should be placed. If you have implemented servers in the cloud, you will need to look at using an ADC service offered by your cloud vendor and collocated with your servers.

Finally, because ADCs are all about handling traffic, the first place to look for scalability is in their Gigabit Ethernet (GbE) or 10 GbE connections. In fact, in tests recently conducted by The Tolly Group, each ADC used a minimum of two bonded 10 GbE links -- with four bonded 10 GbE links required when used in tests simulating larger data center or service provider deployments.

Next: In part three of our How to Buy series on application delivery controllers: How to benchmark ADCs to assure performance.

Next Steps

Part 1: Honing your application delivery platform: Buying the right ADC

Part 3: Performance benchmarking your ADCs

This was last published in September 2014

Dig Deeper on Network application performance

Join the conversation

1 comment

Send me notifications when other members comment.

Please create a username to comment.

Are you considering physical or virtual ADCs in your organization?