CircleID

Do you know how big your web site will be in a year? You probably can guess, or even estimate how much traffic will hit your site in a year. You can also use millennia proven methods such as Crystal Balls, Animal Entrails or even Coffee Stains. But seriously, it is no easy task evaluating site growth in this volatile, ever changing economy.

A key element in the scaling of any web site is the Application Delivery Controller (ADC). This device not only balances HTTP requests among the web servers but also offloads many of their tasks such as TCP management, content encryption, content compression and caching. In addition, an ADC allows for better site control via a rich set of layer 7 traffic control and content modification rules.

How do you choose an ADC looking forward?

There are several strategies of doing that and I will look into their pros and cons.

• Estimate the largest expected growth and buy a device that can handle that much traffic. That is probably not a good idea because you pay a lot more money upfront, money that could be better used on other things, and money that should your projections prove optimistic, need not have been spent at all.
• Buy a device that meets your needs now, and replace it with a bigger device when the need arise. This approach offers better budget control but it has its drawbacks — doing a forklift replacement is a disruptive procedure that incurs a lot of operational and financial overhead on your organization.
• A variant on the previous approach is to buy the device that you need now and add more devices as are needed. Forklift replacement is not needed and it is less disruptive but still three issues remain.
  
  • Repartition your site. When more than one ADC handles your site, you will need to divide your site into several partitions.
  • Manage multiple devices. Depending on your vendor's management tools, managing several devices may become a load on your IT department.
  • Redundancy Cost. The standard practice with ADC's is to pair them into active/passive pairs for high availability purposes. As you add devices, they are added in pairs, which mean that 50% of your investment is unused.
• Yet another variation would be to buy the device that you need today, and when you require more horsepower, add more devices to it, turning it into a virtual device that is the sum of all its parts. Here at Crescendo Networks we call that architecture HyperScale and believe it offers the following benefits:
  
  • Repartition your site. No need to repartition your site since you still operates a single entity that sees your site as a whole.
  • Manage multiple devices. Again, you only manage a single virtual device, with the communication to the physical devices handled for you behind the scenes.
  • Redundancy Cost. With several physical devices sharing the load, a more economic N+I high availability scheme can be employed.

In summary

There is a challenge in predicting your site's future needs. This reflects on your decision for purchasing the right Application Delivery Controller. Choosing an architecture that allows painless growth can save you a lot of money and hassle when your site grows.

By Amit Fridman, Vice President Engineering at Crescendo Networks

Related topics: Cloud Computing, Data Center