Online Activity - F5 Networks Help Consumer Co-Op Get Web in

Since 1938, Recreational Equipment Inc. (REI), the nation’s largest consumer co-op, has been supplying outdoor enthusiasts with high-quality outdoor gear and clothing. REI, based in Seattle, operates 55 stores in 23 states and actively supports conservation and recreation causes.

Three years ago, REI launched a new online venture – ( Now, the biggest outdoor store on the Internet, offers an unrivaled selection of outdoor products and outdoor resource information to customers.

Recently received the highest rating of any sporting goods site, in a ranking published by a leading national consumer magazine. The REI Web site includes more than 17,000 products and 75,000 individual items for all levels of outdoor enthusiasts. The site also features more than 45,000 pages of in-depth product information, expert gear advice, online clinics, a community bulletin board system and a complete adventure travel service.

The site’s regular categories include camping/hiking, climbing, cycling, paddling, snow sports and travel. Most recently, began including fly-fishing and fitness categories. plans to continue introducing additional online categories to its Web site, based on customer feedback and demand.

Getting Out on the Net
Known for excellent customer service and its staff of outdoor experts, REI needed to provide the same high-quality shopping experience for online customers as for in-store customers. To that end, would need to build into its e-commerce infrastructure – two UNIX-based Web servers, as well as a database server – an intelligent, load-balancing solution that addressed key issues, such as content verification, scalability, persistence and fault-tolerant reliability.

Like other thriving Web sites, REI needed to address all elements of Internet Quality Control, including high availability, high performance, security, manageability, open systems, enforcement of policy management and global data management. If its venture were to compete successfully in the online market, REI would require a site that consistently and reliably performed business-critical applications.

After researching several solutions, selected BIG/ip, from F5 Networks because its products prevent system failure, provide timely response to user requests and data flow, and cost-effectively manage an organization’s Internet infrastructure.

"BIG/ip not only offered the capabilities we were looking for, but F5 Networks was the kind of vendor we preferred to work with – a company that not only knows its own equipment, but knows e-commerce," says Robby Robertson, network analyst at’s customer requests can be intelligently distributed, regardless of traffic irregularities, because of BIG/ip’s combination of hardware and software-based algorithms. This approach is unique compared to those vendors that concentrate merely on optimizing performance and focus on hardware-only solutions. While these other products excel at performance, they do little to help availability.

On the other hand, BIG/ip is a traffic and content application management solution for a single Internet site or data center that optimizes both availability and performance for servers, caches, firewalls and VPN gateways, without installing additional software to’s servers. High availability ensures that’s servers, applications and network devices will always be available for its customers, and load balancing involves redirecting incoming site traffic to provide optimal performance for its site.

The BIG/ip Controller provides high-availability load balancing to any specific application or set of applications. This ability also simplifies’s administration and management of server resources because it is not necessary to maintain a separate server farm for specific load-balanced applications. utilizes BIG/ip’s automatic and intelligent management to control the Web site’s Internet traffic and content, which also provides the site with complete performance that includes reliability, scalability, speed and management. BIG/ip is a separate unit that sits between the network and server array to continuously monitor each server for service availability and performance, and route incoming queries to the most available server. By intelligently allocating traffic throughout the site, BIG/ip eliminates server overload conditions that may slow’s performance.

Scaling Difficulties
Single points of failure are eliminated by BIG/ip to provide fault tolerance for By distributing customer requests across a group of servers, content and applications are always available. Dual BIG/ip boxes provide additional fault tolerance with automatic fail-over in less than one second. These features are important to REI’s needs because the Web site needs a service that can route client requests to the most available server and reroute requests away from downed servers – or servers that are not responding accurately.

Server failure is largely prevented at because BIG/ip detects server and application failures while directing traffic to functioning servers and applications.

Another important BIG/ip feature is its mirroring capabilities. Mirroring provides seamless failover of’s client connections and persistence records from an active BIG/ip Controller to a standby controller. This allows a user session to continue even if the Web site’s primary BIG/ip Controller fails.

In addition to the Static Modes load-balancing algorithms, such as Round Robin, Priority and Ratio, BIG/ip also provides with Dynamic load-balancing algorithms. The Dynamic Modes include Least Connections, Fastest, Observed and Predictive.

Least Connections passes a new connection to the node with the least number of current connections. Then BIG/ip probes the application infrastructure, verifies the availability of the servers for that traffic, and ultimately prioritizes the traffic destination. The Fastest Mode passes a new connection to a node based on the fastest measured response time of all currently active nodes. In Observed Mode, connections go to the server with the best balance of fewest connections and fastest response time. Predictive Mode reacts by sending connections to the currently improving server with the best performance ranking.

The BIG/ip Controller provides layer seven management through its Extended Content Verification (ECV) and Extended Application Verification (EAV) features.

"These features enable us to test our Web pages to make sure that the applications, as well as the boxes, are running," explains Robertson. "With BIG/ip’s verification capabilities, we can be sure that customers are receiving the content they expect."

ECV is a method for BIG/ip to verify that the content is responding properly to prevent users from seeing error messages. This application ensures that customers receive the data they requested even if a server is down. Network managers verify this by sending a query to Web servers, firewalls, cache servers and other transparent devices, then examining the response that is sent back for accuracy. If there is a problem, BIG/ip proactively directs the users to a server that is responding with the correct information until the problem is resolved.

EAV is a more sophisticated version of ECV, which allows the network managers to test their site from the customer’s point of view to ensure that the correct applications and content are responding. The managers perform this verification process by developing their own external programs so they can perform multiple layers of testing to determine if the applications are available and working correctly.

Self Testing

EAV proactively tests’s transaction process, simulating a session that logs onto multiple accounts, places items into a shopping cart, verifies that the credit card authorization is working properly and completes the online financial transaction. EAV then provides critical feedback to’s network managers on server, application and content availability, which is the information BIG/ip requires to apply load balancing to maximize the utilization of resources.

The network managers can also customize EAV to check a number of’s other applications. Then BIG/ip will verify that multiple directory and/or authentication services (LDAP, Radius and DNS) are providing the correct content to customers, that the gateway services (SAA, SNA, etc.) are available, that mail servers (POP, IMAP, SendMail) are accepting connections and responding properly, and the availability of multiple interactive services (Telnet, TN3270, TN5520).

BIG/ip also excels at persistence, says Kim Muromoto, Technical Services Manager at Persistence – or the ability to ensure that customers remain connected to the same server during a transaction – is required to seamlessly process customer requests, while keeping the link between the vendor’s site and the customer unbroken. Persistence is necessary when a server has data associated with the user and the data is not dynamically shared with the other servers. However, e-commerce poses some challenges for persistence, specifically in regard to scalability and performance of a load-balancing device.

Several persistence settings (eight modes) are featured in BIG/ip, including SSL Session ID Tracking. By reading specific Session IDs from an SSL transaction, BIG/ip assures that each user is uniquely identified and is delivered accurate and timely content until the transaction is complete. BIG/ip also uses destination address affinity that redirects requests for the same content back to the same cache. BIG/ip’s wide variety of persistence capabilities also allows’s users to reconnect to the same server that they connected to in the past when they return to a site.

Both inbound and outbound traffic can be balanced for devices like firewalls or routers to ensure users return to the same device. And the nPath optional mode allows servers to bypass BIG/ip when returning traffic to clients for faster return responses.

Network managers can use BIG/ip to develop a variety of sophisticated, load-balancing algorithms to fine-tune the Web site’s performance and availability. Network-based failover allows to configure a redundant BIG/ip Controller to use a network connection to determine the status of the active controller. Network-based failover can also be used in addition to, or in place of, hard-wired failover. This is a significant feature because it gives more flexibility to the network manager. With network-based failover, redundant BIG/ip Controllers are not limited to the physical proximity caused by the 25-foot serial port failover cable. can configure BIG/ip to direct requests for specific servers to specific groups of servers and to direct requests to different groups of devices. Traffic prioritization and rate shaping are used for efficient bandwidth allocation.

Any operating system is supported by BIG/ip, which uses a variety of Internet applications and services over TCP/IP. The BIG/ip Controller supports Internet applications, such as Web (HTTP), e-mail, streaming audio and video, LDAP and other IP protocol traffic. BIG/ip also simplifies the task of accessing and controlling a number of Internet services and devices, such as FTP, e-mail, firewalls, cache and multimedia, all from a single UI. And at the network appliance level, BIG/ip extends and protects content generated by backend applications (such as databases), by actively querying individual servers at the application level.

BIG/ip is designed to protect REI’s Web site from attacks and provide protection for the servers being load balanced. BIG/ip does this by performing packet filtering for to deny access from known sources of questionable traffic and can thwart many common types of attack, such as IP spoofing or denial of service attacks using ICMP and SYN flooding. BIG/ip’s hardened and secure device resists common attacks and helps prevent hacking, divert traffic around hacked servers and quickly alerts network managers of any problems.

Maintaining REI’s site is simplified because BIG/ip reduces the need for the company to purchase increasingly larger and more expensive servers to accommodate increases in its network traffic and enables many inexpensive servers to function as a single, virtual server. BIG/ip also reduces the single point of failure and expense inherent with a single large server, and allows for the orderly addition of new servers, or the routine maintenance or upgrades of servers without disrupting service to the end user. can grow at its own pace: As demand for its products and services grow, can easily upgrade its Internet infrastructure. Each BIG/ip Controller is field upgradeable, which provides with a cost-effective solution for its unique situation.

The BIG/ip software is providing upgrades for feature and performance increases at the rate of three to four updates per year. And the hardware can be upgraded any time as new and faster process and memory become available. This method of upgrading is possible because BIG/ip’s hardware/software design allows to receive performance gains without having to simultaneously upgrade expensive hardware.

Mike Fox, Senior Network Analyst at, points out that the reliability BIG/ip provides is noteworthy. "These boxes have been extremely reliable," he says. "We just turn them on and they work – day and night. BIG/ip is instrumental in helping maintain a high-performance, high-quality Web site," adds Muromoto.