Accelerating E-commerce with 
CacheFlow Internet Caching Appliances

Overloaded Servers and Impatient Consumers

Many businesses today are realizing that the World Wide Web is an excellent tool to increase sales, locate new customers, improve customer service, and reduce the costs of selling and supporting products or services.  Electronic commerce (E-commerce) is complementing and in some cases surpassing traditional retail sales for many businesses.  Unfortunately consumers expect much more from Web-based stores than they do traditional retail facilities.  While people are accustomed to waiting in lines at stores, they are less tolerant of delay while waiting for a Web page to load.  Recent research concluded that if a page takes longer than 8 seconds to load, many users will cancel the transaction or exit the site.

Figure 1.  Likelihood of customer leaving a Web site (source: Zona Research)

On the Web every second counts.  The difference between losing 7% of potential customers and losing nearly one third is only one second.  Zona Research has estimated that these precious seconds could cause a loss of E-commerce revenue of $73 million per month.

To eliminate the lost revenue potential, and satisfy E-commerce consumers, there are four key requirements an E-commerce solution must address:

�         Throughput.  The number of transactions that an E-commerce solution is capable of handling.

�         Speed.   Delivering Web pages and fulfilling E-commerce transactions in an acceptable amount of time.

�         Reliability & Availability.  The ability to consistently serve requests and complete transactions with no service interruptions.

�         Scalability.  Building a solution that easily grows as the business grows, and is capable of handling traffic spikes.

Network administrators are challenged to build solutions that are capable of delivering the scalability, load capacity, and throughput that is needed to meet customer expectations and to capitalize on the enormous revenue potential in E-commerce.
 

Pinpointing E-commerce Bottlenecks

There are many factors limiting E-commerce performance.  Within an E-commerce site and across the Internet, everything from link speed to routers and servers adds delay to an end-user request.  Figure 2 shows how requests for Web pages must go through the router, firewall, and layer 4 switch before they reach the Web servers, are processed, and sent back out through the same network components.  Forcing user requests through these devices, each with finite throughput, can easily expose network scalability constraints and performance bottlenecks.

  Figure 2.  Typical E-commerce Network


In addition, as more and more users try to access the same content, the redundant load on the firewalls and servers for the same Web objects increases significantly.  This is particularly problematic in connection with promotions or major events.

• Redundant connections required to serve the same objects

•  Traffic surges due to promotions or major events

• Overburdened firewalls

• Overburdened servers

Redundant Connections
In an E-commerce network, a large percentage of Web objects are static.  Examining the elements of a Web page shows that objects can generally be grouped into three categories: objects that rarely change (logos), objects that change weekly, hourly, or daily (news items, promotions), and items that are dynamically generated and unique for each site visitor.  As much as 90% or more of Web objects can be static.  Delivering these objects over and over from the Web servers unnecessarily forces requests through the internal network.  This model causes delays and exposes each network device as a potential bottleneck or point of failure.

Traffic Surges
The number of online shoppers is projected to double over the next 12 months.  Couple that with seasonal shopping spikes, marketing and sales promotions, and major events and it becomes obvious that networks must be designed to handle traffic surges well beyond the average number of daily transactions. 

Firewalls
In an effort to protect Web content from malicious hackers, Web servers are often placed behind firewalls along with the rest of the Intranet.  When deployed in this fashion firewalls quickly become overburdened checking packets that contain Web objects.   Like all networking devices, firewalls can only serve a limited number of simultaneous connections.  Using these connections to serve publicly accessible Web content that does not need to be secured is a waste of resources.  Overloading firewalls with redundant data requests can cause a major bottleneck and add costly seconds to Web page delivery times.  This bottleneck has a negative impact on both Website visitors and Intranet users at an E-commerce site.

Servers
It is common to deploy multiple servers with duplicate sets of data to ensure resilience.  Servers may experience problems performing the critical functions of generating dynamic content and interfacing with back-end databases.  When overloaded serving redundant Web objects to users, servers begin losing TCP sessions and dropping packets.  Once again Web page delivery times suffer and end-user satisfaction degrades.

The conventional approach to solving these problems is more -  more servers, more firewalls, more bandwidth, more cost, more complexity, more management burden.  But now, with caching, there is an alternative.

Caching in on E-commerce

Caching offers a way to accelerate the delivery of Web content.  Historically, caches have been used primarily at the client side, to accelerate client demand for frequently requested content by storing and serving that content locally.  This is known as “client-side caching”. 

In an E-commerce environment, caches can be deployed at the  “server-side” to:

�         Offer site visitors an optimal Web site experience

�         Offload overburdened servers

�         Offload overtaxed firewalls

�         Scale the network to handle more customer transactions

�         Scale the network to handle large traffic spikes

�         Reduce capital and operating costs 

                    Figure 3. Typical server-side caching environment

In the example displayed in Figure 3, customer HTTP (port 80) requests coming in from the Internet are redirected through the router to the cache.  The cache then serves the Web objects that it has stored directly back to the client.  Requests for dynamic content or secure (SSL) transactions are passed through to the origin servers for processing.  It is not unusual for cache hit rates to exceed 90% in E-commerce environments.

Other than the obvious load reduction on firewalls and servers, serving a very high number of Web requests outside of the firewall has other benefits. Having the cache handle most of the Web requests reduces the security risks of users directly accessing servers that are inside the firewall.  Further, Intranet users who are dependent on the firewall for their network-based access receive improved performance.

Because as much as 90% or more of Web objects can be cached in a typical E-commerce environment, delivering these objects from the edge of the network, outside the firewall, and in front of the Web servers can deliver dramatically better response time and allow the site to handle more transactions.

The Advantages of CacheFlow and CacheOS in an E-commerce Network

All CacheFlow Internet caching appliances run the purpose-built CacheOS™ operating system. Lightweight and efficient, CacheOS leverages a number of patent-pending algorithms to minimize the response time for the delivery of both static and dynamic Web pages while ensuring the highest standards of content freshness.  As the only purpose-built caching operating system, CacheOS delivers the security and ease of installation, configuration and management that network managers require, and the fast response-time performance that end-users demand.

Appliance Architecture

CacheFlow’s line of Internet caching appliances provides crucial capabilities not found in other caching products: 

�         Extremely low administration.  CacheFlow products are designed to be installed in minutes and require practically no administration. They are self-tuning and self-healing appliances.  Other caching solutions require regular maintenance and scheduled downtime. CacheFlow devices provide the simplicity and ease of use of a true appliance.

�         High reliability and availability.  Designed for robust network requirements, CacheFlow appliances are extremely reliable. In the case of a device fault, they perform a "fast restart" within seconds, with no human intervention. No perceived loss of service is experienced, and administrators are alerted to the restart via automatic email or a page from the device.

�         Simple management.  One of the primary goals of deploying a Web cache in an E-commerce application is to alleviate the headaches associated with maintaining servers. CacheFlow appliances can be remotely managed via command line interface or through the CacheOS Java management console. 

CacheOS Highlights

As a 100% caching-focused operating system, CacheOS delivers performance and functionality not possible with general-purpose alternatives:

�         Hit rate.  When used in an E-commerce environment, caches are managing a relatively fixed set of objects.  This gives caches the ability to serve an extraordinarily high percentage of requested objects without impacting back-end servers.  The algorithms of CacheOS enable an even higher hit ratio than traditional caches by employing an intelligent object deletion policy.  Rather than simply storing the most recently requested objects like other caches, CacheOS tracks the popularity of objects and stores the most frequently requested.  This approach allows CacheFlow appliances to easily deliver hit rates exceeding 90% for E-commerce loads.

�         Response time.  While it is imperative that caches store the correct data, it is just as integral that a cache can find and serve that data quickly.  For this reason, CacheOS relies on a purpose-built object storage system and not a file system.  Object access is through a hash table in RAM, ensuring that any object can be obtained in a single disk read and that objects are delivered immediately back to the requester.

�         Reporting.  Many Website managers rely on the demographics retrieved from their server logs.  CacheOS supports standard logging formats.  For Webmasters using extended logging or non-standard formats, CacheOS also includes the ability to add fields.

�         Secure content storage and delivery.  E-commerce sites cannot afford to have hackers violating the security thresholds of their systems, or the entire business could be compromised.  In an E-commerce application when the cache is deployed outside the firewall, the need for an inherently secure solution is obvious. The development of CacheOS has always been tightly controlled to manage system isolation. This isolation keeps security at a maximum level, without sacrificing usability or new features. Other caching solutions based on general-purpose operating systems such as UNIX, NT, or NetWare have security holes through which hackers can easily break into the system. The high level of security found in CacheOS is necessary to protect the integrity of data that is now being stored and served outside of the firewall.

Real World Results:  E-commerce Customer Measurements

CacheFlow appliances have been deployed at a number of very high traffic E-commerce sites to scale the network and deliver the optimal end-user experience.  At each of those sites, network managers are realizing significant benefits including:

Server Offload
Across customer deployments, the CacheFlow appliances have consistently and significantly reduced the load on the server infrastructure.  At one customer, 92% of Web objects are served from CacheFlow appliances without involvement from back-end servers.  In the typical deployment, the network is offloaded to the point where traffic loads can increase by more than 10 times without requiring upgrades to firewall and server resources.

                       

                    Figure 4. Customer traffic flow to the Web servers before 
                           and after installing a CacheFlow appliance. 

 
Firewall Offload
The impact of routing end-user requests through a firewall or firewalls can be astonishing.  In one particular deployment of CacheFlow appliances, the firewall was handling nearly 10,000 open sessions at a time.  After installing a single CacheFlow appliance, the number of objects served by the cache was so high that the firewall load was reduced to just 150 open sessions.
                                                                          

                 Figure 5.  98% Firewall session reduction.


Page Delivery Time Improvements

Fast page response times are a make-or-break factor for E-commerce success.  For large E-commerce operations, CacheFlow has reduced Web page delivery times to well below the 8-second threshold where lost sales begin.

                
                                              
                 Figure 6.   Router response time measurements with and 
                                   without CacheFlow

 
Conclusion

Caching can deliver immediate and powerful benefits to E-commerce.  CacheFlow Internet caching appliances are unparalleled in delivering E-commerce acceleration that keeps customers coming back to transact business.  The patent-pending features of CacheOS make CacheFlow appliances a network necessity for companies deploying an E-commerce infrastructure.

To have a CacheFlow sales agent contact you regarding Accelerating your E-commerce Site, please contact: