I’m not sure why this happens, but it seems that every few years, this industry likes to debate the death of certain markets. Application Delivery Controllers, particularly the tried-and-true hardware platforms, have been under attack before and seem to be again. To me, the concept that you could just build ADC functionality into another product seems laughable. I agree that it’s getting easier to build basic load-balancing capabilities into other platforms, but the ADC is so much more than that now. In fact, it’s more common to have hardware platforms go through periodic hardware jumps to accommodate market transitions. Routers did this many years ago when Cisco and Juniper announced their next-gen product, and I think ADCs will as well.
In fact, earlier this year, ADC market leader F5 announced an upgrade to their entire product line for 2013 with some beefy performance numbers. The beefiest of the products, the high-end Viprion 4800 has 64 Gbps of Layer 4 throughput, and 320 Gbps at Layer 7. These performance numbers blow away the older platforms from F5 and the competitive landscape.
Now, I’m certainly not saying that everyone needs this performance, but I do believe some organizations do. Workloads need to dictate the platform, and choosing an underpowered ADC for workloads that require the performance can cause some serious network issues. Below are the top workloads that require a high performance ADC:
- Denial-of-service defense. These attacks are volumetric in nature, where, to maintain consistent application availability, the ADC must separate legitimate requests from a fire hose of bogus traffic. There’s certainly no question that the industry has seen a rise in the number of DoS attacks in the past few years, making this a must-have feature for network and cloud service providers. One of the cloud providers that I recently chatted with claimed to have sustained DoS attacks at a rate of 70 Gigs per second. Denial-of-service threats also include more sophisticated approaches in attacking specific applications, where the ADC must employ filtering of the web requests — a task which is highly computationally intensive.
- Web-intensive organizations. Some organizations, such as mobile service providers, online retailers, social media companies, and government sites deal with massive amounts of Web traffic. This alone can be akin to a denial of service attack, and a lack of preparedness can take down high-profile sites. One example of this was the first Victoria Secret fashion show broadcast over the web. There were so many web requests that the sheer volume of requests took down the website. As Internet use becomes more pervasive, the amount of network traffic to be managed and shaped for these sites invariably consumes the capacity of older hardware platforms.
- Consolidation of network hardware. ADCs today have far more capabilities than just simple load balancing. Critical network services such as DNS, SSL offload, WAF, and compression were once delivered by individual appliances. Today, these services can be consolidated onto an ADC. While consolidation simplifies administration, cuts power and space requirements, and can lower capex over time, running multiple services on one platform can exhaust the capacity of the ADC in high-volume traffic environments. A high-performance ADC can run multiple network services simultaneously while providing sufficient headroom for handling peak workloads.
- Future-proofing. This might seem a little boring but it’s crucial to have a platform that can scale with the company. Projecting what a company’s data requirements will be into the future is a difficult task, and is generally underestimated. Given the continued meteoric rise in traffic, it’s prudent to invest in sufficient performance, not just for today but also to meet the demands the ADC will face over it’s useful deployment lifespan — typically 3-5 years.
ADCs come in many shapes and sizes and there’s no right answer to which one organizations should use. Let the workload dictate the platform and for those high performance workloads, like the ones listed above, choose a hardware platform that can meet the requirements of today and tomorrow.
Image credit: marc kjerland (flickr)