Content Delivery Networks

Cisco's Content Networking Architecture: Components

Through internal developments and acquisitions of breakthrough content networking technologies, Cisco has built a comprehensive architecture for optimizing Web site performance and content delivery. This architecture comprises the five essential technology building blocks along with two complementary technologies that provide the foundation for all of Cisco's existing and future content networking solutions. Table 5-1 describes each one of the components and complementary technologies that make up CDNs. This table provides an overview of each of the components. A more detailed explanation appears later in the chapter.

Table 5-1 CDN Components

Cisco is the only vendor that offers all five CDN solution components. By adding these components to their existing IP networks, organizations can quickly benefit from a variety of new, high-value e-services.

The foundation of a CDN, as listed in Table 5-1, is a highly available Layer 2 and 3 infrastructure. This infrastructure is made reliable by many of the protocols that organizations already run such as HSRP, Spanning Tree, OSPF, EIGRP, and BGP.

After this infrastructure is in place, a content delivery network can be overlaid onto the foundation to begin to allow intelligent networking to take place at the upper layers of the OSI reference model.

Intelligent Network Services

The first component, Intelligent Network Services, is the most readily available and, in many cases, is the enabler for the other components. Intelligent Network Services are tightly integrated with the Layer 2 and Layer 3 infrastructure because the software features for Intelligent Network Services such as QoS, multicast, VPN, and security are available on Cisco routers and switches. For added functionality and performance, external appliances can be purchased to give a company even more intelligence and increased performance.

For example, multicast is a tremendous enabler for an enterprise network because it allows the distribution of one large video file to be made with one stream rather than multiple unicast streams. Companies can easily deploy an e-learning or corporate communications solution without requiring an upgrade of all connections to support the multiple streams because a popular e-learning presentation can be sent once (multicasted) to many people simultaneously.

Content Switching

Content Switching is an essential category to many CDN architectures because of its capability to increase the entire architecture's performance, scalability, and redundancy. You'll hear the term "load balancing" used almost as a synonym with content switching. There is only a small technical difference between load balancing and content switching. However, both accomplish the same outcome—namely distributing requests across multiple servers.

The distinction to be made for load balancing is that the distribution of requests is done with less intelligence than content switching. This intelligence is based on the 7-layer OSI model. Load balancing is handled at Layer 3 and Layer 4, which means that the forwarding decisions are made on the IP address/port number relationship. In most cases, for web switches load balancing occurs on HTTP TCP port 80 and SSL TCP port 443. Figure 5-4 depicts how the load balancer distributes requests across multiple servers, creating more scalability by reducing the load on one server.

Figure 5-4 Load Balancer Reduces Load Through More Even Request Distribution

The growing volume and sophistication of e-business-related traffic required a solution that provided increased performance and features. Content switching answered the challenge because of its capability to gather knowledge of the user, device, network, and content. While doing all of this, it is still able to provide deep packet inspection and parsing by using an appliance- or module-based architecture with the increased hardware performance coming from application-specific integrated circuit (ASIC) or network processor-based packet handling.

Content switches are capable of answering these questions:

• Who is requesting the content?

• What content the user is requesting?

• Which server has the content?

• Which server that has the content is best available to serve the content back to the user?

With policies that are implemented at Layer 5 through 7 of the OSI model, the content switch is capable of specifying the policy-based web traffic direction based on full visibility of URLs, host tags, and cookies. This gives a company the flexibility to make forwarding decisions with even more intelligence. Figure 5-5 shows how the Layer 7 content switch is able to distribute the request based on more information, adding more intelligence into the network.

Figure 5-5 Content Switching Basics

Cisco Content Services Switches

The Cisco CSS 11000 series content services switches are Layer 5/7 aware and provide a robust front-end for web server farms and cache clusters. With unique features for e-commerce, web hosting, and content delivery, the Cisco CSS 11000 series is an important piece in Cisco's end-to-end content networking solution. Cisco's CSS 11000 series switches, powered by Cisco Web Network Services (Web NS) software, offer unique services, including the following:

• Directing web traffic based on full visibility of URLs, host tags, and cookies

• Enabling premium services for e-commerce and web hosting providers

• Strengthening DoS protection, cache/firewall load-balancing, and "flash-crowd" management

Cisco Content Switch Module

The Cisco Content Switching Module (CSM) is a Catalyst 6500 line card that balances client traffic to farms of servers, firewalls, SSL devices, or VPN termination devices. The CSM provides a high-performance, cost-effective load-balancing solution for enterprise and ISP networks. The CSM meets the demands of high-speed Content Delivery Networks, tracking network sessions and server load conditions in real time and directing each session to the most appropriate server. Fault-tolerant CSM configurations maintain full state information and provide true hitless failover required for mission-critical functions.

The CSM provides the following key benefits:

• Market-leading performance—Establishes up to 200,000 Layer 4 connections per second (performance might vary based on software versions being run), and provides high-speed content switching while maintaining one million concurrent connections.

• Outstanding price/performance value for enterprises and ISPs—Features a low connection cost and occupies a small footprint. The CSM slides into a slot in a new or existing Catalyst 6500 and enables all ports in the Catalyst 6500 for Layer 4 through Layer 7 content switching. Multiple CSMs can be installed in the same Catalyst 6500.

• Ease of configuration—Uses the same Cisco IOS Software command-line interface (CLI) that is used to configure the Catalyst 6500 Switch.

Features and Benefits of the CSM

Some of the features and benefits of the Cisco Content Switching Module are as follows:

• Firewall Load Balancing—The CSM allows you to scale firewall protection by distributing traffic across multiple firewalls on a per-connection basis while ensuring that all packets belonging to a particular connection go through the same firewall. Both stealth and regular firewalls are supported.

• URL and Cookie-based Load Balancing—The CSM allows full regular expression pattern matching for policies based on URLs, cookies, and HTTP header fields. The CSM supports any URL or cookie format allowing it to load balance existing World Wide Web content without requiring URL/cookie format changes.

• High Performance—The CSM performs up to 200,000 new Layer 4 TCP connection setups per second (Performance might vary based on software versions being run.). These connections can be spread across 4096 virtual services (16,384 real servers) and all of the ports in a Catalyst 6500 or they can be focused on a single port. This provides a benefit over competitors who use distributed architectures that require all of the ports to be used in order to gain maximum performance.

• Network configurations—The CSM supports many different network topologies. A CSM can operate in a mixed bridged and routed configuration allowing traffic to flow from the client side to the server side on the same or on different IP subnets.

• IP Protocol Support—The CSM accommodates a wide range of common IP protocols including TCP and User Datagram Protocol (UDP). Additionally, the CSM supports higher-level protocols including Hypertext Transfer Protocol (HTTP), File Transfer Protocol (FTP), Telnet, Dynamic Name Server (DNS), and Simple Mail Transfer Protocol (SMTP).

• User Session Stickiness—Whenever encryption or e-commerce is involved, the end user must be consistently directed to the same server (that is, the server where the user's shopping cart is located or the encryption tunnel terminates). The CSM's user session stickiness provides the ability to consistently bring users back to the same server based on Secure Socket Layer (SSL) session ID, IP address, cookie, or HTTP redirection.

• Load-Balancing Algorithms—The CSM supports the following load-balancing algorithms:

  • Round robin

  • Weighted round robin

  • Least connections

  • Weighted least connections

  • Source and/or destination IP hash (subnet mask also configurable)

  • URL hashing

• Quality of Service (QoS)—Providing differentiated levels of service to end users is important when generating revenue from content. The CSM leverages the Catalyst 6500's robust QoS, enabling traffic differentiation as follows:

  • Correctly prioritizes packets based on Layer 7 rules

  • Directs users who are paying more for services to faster or less loaded servers

• High availability—The CSM continually monitors server and application availability using health monitoring probes, inband health monitoring, return code checking, and Dynamic Feedback Protocol (DFP). SNMP server health traps can also be configured. When a real server or gateway failure occurs, the CSM redirects traffic to a different location. Servers can be added and removed without disrupting service and systems can easily be scaled up or down.

• Connection Redundancy—Optionally, two CSMs can be configured in a fault-tolerant configuration to share state information about user sessions and provide connection redundancy. In the event the active CSM fails, open connections are handled by the standby CSM without interruption. Users will experience hitless failover which is an important requirement for e-commerce sites and sites where encryption is used.

Content Edge Delivery

Content Edge Delivery consists of edge content engines that seamlessly deliver web, streaming media, or static content faster than traditional servers to clients. This cached content is delivered by Cisco Content Engines with specialized software optimized for content delivery. Content Edge Delivery accomplished from the edge localizes traffic patterns. Content requests are then filled in an optimal manner, improving the end user experience and allowing rich media content over the Internet or corporate network. This is done by having that traffic retrieved from the local corporate network and not over the corporate WAN. This is known as transparent caching. Delivering a client to the most optimal content engine requires the CDN's intelligence to optimize traffic flows. Content Edge Delivery is tightly integrated with the other CDN traffic components discussed throughout this chapter. Today, Cisco Content Engines handle static and streaming media content. In the future, dynamic database content and applications will also be cached.

When a content engine receives a request, it answers the following questions before delivering the content to the end user:

• What's the best resource for this request?

• Is my current copy of this content fresh?

• If not, go to the origin server and refresh it.

• Is there QoS information that I should use or alter for these packets?

• Do I have sufficient processing cycles to handle more requests? If not, tell everyone.

Cisco's Content Engine products include the following:

• 507
• 507-AV
• 560
• 560-AV
• 590
• 7320

Content Routing or Global Server Load Balancing

Remember Figure 5-1 earlier in the chapter that shows how the size of the server store has grown? The solution, Content Delivery Networking, allows that content to be distributed close to the user so his experience of the content is outstanding rather than good. The category of Content Routing or Global Server Load Balancing, as it's also known, takes care of the intelligent routing of a user request to the best available content engine (this device stores the content). The request is taken in by the content router and the end user is redirected to the best site based on a set of metrics such as delay, topology, server load, and a set of policies such as location of content, enabling the accelerated delivery of web content and streaming media.

As the number of users accessing content on the Internet grows, providing a high level of availability and rapid response from a single location becomes increasingly difficult. The solution to this problem is content routing between multiple data centers or Points of Presence (POPs). Content routing ensures the fastest delivery of content regardless of location, providing the highest availability and site response because the content router is able to sense which content engine, content switch, or point of presence (POP) is the best available to deliver the user's request. This is where a great part of Cisco's CDN intelligence comes from. When a content router takes a look at a request from a user it is essentially asking and answering the following questions:

• What content is requested?

• Where is the user?

• What's the best site now based on site and network health-checks?

• Are there other rules to factor into this decision?

Cisco's content routing products include the following:

• Content Services Switch 11000

• Content Router 4450

• Content Router 4400

• Cisco 7200 Distributed Director

Content Distribution and Management

How does the content get to all the content engines or content POPs? The answer to that question is through the Content Distribution and Management category. The Content Distribution Manager (CDM) proactively distributes cacheable content from origin servers to content servers at the edges of the network, and keeps the content fresh. This device is a company's central store of content. The CDM enables you to automatically import, maintain copies, and configure content at the edge of the network. From the CDM, an enterprise can globally and centrally provision content to remote locations by using either unicast or multicast, the Intelligent Network Service previously discussed.

Some critical functions include the following:

• Implementing content-specific policies across a global network infrastructure

• Creating virtual CDNs to support multiple discrete clients on a shared services infrastructure

• Transaction logging for automated billing based on actual content service usage

The CDM is continually able to answer the following questions:

• What content can be pre-populated at network edge?

• How should it be distributed (Self-Organizing Distributed Architecture [SODA] or multicast)?

• Which locations should be pre-populated with which content?

• When should content be refreshed?

• How should the service be provisioned and billed?

Cisco's CDM products include the following:

• CDM 4670
• CDM 4650
• CDM 4630

Figure 5-6 illustrates what happens in Content Delivery architecture from Step 1 to Step 7, and which CDN categories (listed in Table 5-1) take care of each function.

Figure 5-6 Cisco Content Networking in Action

The following detailed list is a description of how the main components can be used.

1. The origin content is pulled to the CDM.

2. The CDM pushes the content to content engines at the edge of the network.

3. The end user makes a request for Content.

4. The request is intercepted by the Content Router, which finds the best site to service that content.

5. The content engine retrieves the content for the end user, if necessary.

6. The Content Switch receives the request and finds the best server to serve that content to the user.

7. The end user receives the content.