Server Load Balancing: Not Just for the Big Guys Anymore
Apr 14, 2010 5:00 AM PT
Today Web and application infrastructure continues to expand applications for e-commerce and communications with customers, partners and internal employees. For those responsible for deploying, maintaining, and supporting these network applications, there is an increasing need to get their highest performance out of data center equipment while simplifying the complexity that causes excess overhead and unnecessary costs.
The cost associated with network downtime can run six figures and above on an hourly basis. If not controlled, downtime can also lead to customer dissatisfaction and lost business. These consequences are unacceptable in a market where it costs ten times more to attract a new customer than it does to maintain a current one.
One of the greatest challenges to any organization is the maintaining of continuous and economical Web and application infrastructure and uptime. Business-critical Web sites and intranets face many challenges, from user traffic spikes to server bog down to handle compute-intensive SSL transactions, large and numerous data file requests and IO and memory overloads and hackers try to flood servers with malicious attacks.
Today's global economy relies on internet sites to run optimally 24/7, while IT staff demands 100 percent operational-ready solutions. The operational readiness, maintainability and durability of today's Internet sites are directly tied to the myriad of equipment all connected together with the data center infrastructure. Now managing Internet user traffic and applications is the key element in delivering high availability.
Listen to the podcast (10:50 minutes).
Putting Idle Systems to Work
Application delivery controllers (ADCs) and server load balancers are devices that optimize Web and application infrastructure, providing high availability, high performance, and flexible scalability and secure operations all while streamlining IT costs. ADCs simplify the management of network resources and optimize and accelerate user access to diverse servers, content and transactional based systems.
Approximately 10 percent of the application servers in a typical data center are not used efficiently today. Yet these systems still require ongoing backups, maintenance, and licensing costs not to mention the energy costs associated with running them. However, application delivery controllers and server load balancers can make these under utilized servers highly optimized.
These appliances can help reduce data center costs and optimize data center operations with: maximizing server utilization, optimize investments in existing equipment, improve performance by distributing user traffic loads among multiple Web and application servers, optimize server and application resources by efficiently allocating traffic across servers based on application types, increase site performance by distributing user traffic based on actual server resources and cost effectively scale server capacity to keep up with growing traffic demands.
One of the key elements that an ADC and a load balancer provide is that it reduces the single point of failure associated with having only one server. Until recently, even basic server load balancing was essentially cost prohibitive for smaller-sized companies. Today, these are not only affordable but the consolidation of load balancing, content switching and server offload capabilities for SSL, features like caching and compression really provide smaller businesses with very cost effective out of the box solutions that are priced from about $1,500.
Data center managers know that reliable Web and application infrastructure is crucial with fast and reliable access to information and transactions. ADCs integrate four important functions of site management: high availability, server off load, user traffic distribution among server resources and application security. The key is to have the right product to fit your specific infrastructure needs. For enterprise organizations, companies with hundreds or even thousands of servers, integrating what's called "best of breed" network infrastructure is pretty commonplace. However, for smaller businesses even with little as two servers, "best of breed" can be expensive and overkill given the focus on enterprise-specific features and functions. There are vendors that offer full featured load balancing and application delivery appliances that include advanced capabilities like content switching, low cell persistence and server affinity, SSL cache and compressions all at a price that start around $1,500.
What to Look For
One of the key components in dealing with increased and unpredictable Web traffic and enabling of high availability for both Web-facing and Internet applications is the application delivery and load balancing appliance. Key features to look for in a load balancer include: support for layer 7 functionality like content switching, application health checking, L7 persistence, SSL acceleration and off-loading, advanced features like caching and compression and integrated security such as IPS.
So far we have been discussing the benefits of a server load balancing and application delivery controller as a dedicated purpose-built hardware appliance. As organizations are starting to leverage virtualization technology to extend and optimize their application server infrastructure, a relatively new breed of load balancing products are appearing on the market in the form of a virtual appliance.
These virtual appliances provide many -- in some cases, all -- of the same functionality of the hardware-based appliances but with the added benefit of being tightly integrated with the virtualization platform's hypervisor. That enables network administrators to really leverage their virtualization infrastructure for maintaining high availability and optimized performance of their applications.
As a virtual appliance, a virtual load balancer removes the need to maintain hardware and reduces costs associated with power consumption, cooling, rack space constraints and other environmental dependencies of hardware-based appliances. Typically a virtual load balancer can be quickly configured and deployed, and accelerate the service provisioning and simplify ongoing management which further helps in reducing operating expenses and increasing ROI.
Virtual appliances are designed to work with specific platforms such as VMware or XenServer and can be installed as a guest operating system on a virtual machine and cloned or relocated being tightly integrated with a specific virtualization platform. Administrators can leverage the management tools of that hypervisor to also manage their virtual load balancing appliance.
Virtual or Traditional?
The relative performance of a virtual load balancing appliance is highly dependent on the amount of physical resources, such as CPU memory, that is allocated towards that particular virtual machine supporting the load balancing application. So it is very important to consider the amount of available physical resources that could be allocated toward that virtual machine and whether those resources are enough to sustain the required throughput.
Some vendors have designed their virtual load balancers to function as evaluation or development platforms allowing administrators to test the functionality of their applications and load balanced environments. For production and business-critical applications, these vendors want you to purchase their hardware-based platforms, while others do offer virtual load-balancing appliances that are built to support full production operations. So the decision to go virtual or traditional hardware route will depend on specific application and performance requirements. In either case a server load balancer or application delivery controller can offer significant improvements to web application performance and enable high availability of application server infrastructure.
You no longer have to spend tens or even hundreds of thousands of dollars on expensive ADC appliances. There are vendors that provide full-featured, high-performance load balancing and application delivery solutions starting well under US$2,000.
Peter Melerud is the cofounder and vice president of product development for Kemp Technologies.