| About InfoWorld : Advertise : Subscribe : Contact Us : Awards : Events : Store |
 |
||||
|
||||
 | ||||
 |
|
|||
 | ||||
|
|
|
||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||
 NEWSLiving on the edge By David L. Margulius March 15, 2002 1:01 pm PT What began with Victoria's Secret may soon be coming to a WAN near you. Since the late 1990s, high-traffic Web sites have used CDNs (content delivery networks) from companies such as Akamai and Digital Island to help them handle peak loads and special events such as the Victoria's Secret Webcasts. Service providers, too, have long supported the technology, investing heavily in caching hardware and software that allow them to manage bandwidth utilization within their networks.
Many companies are now using CDNs to push traditional static content, frequently-accessed data, and large files (for example, PDF documents, streaming video, e-learning applications, and corporate presentations) closer to the edges of their networks where they can be more easily accessed by end-users. Other companies have begun deploying CDN technology to deliver dynamic, database-driven content for business software such as ERP and CRM. Why? Because to many enterprises, a CDN's promise of faster application performance and higher network availability is just too good to pass up. Couple that with the growing telecommuting trend and the rise in Net-enabled business services, and it's no wonder companies are mulling over the possibilities of edge networking. Building a better network A typical CDN consists of a group of servers, caches, and routing software capable of moving frequently used content closer to end-users, thus maximizing bandwidth efficiency and boosting network scalability and availability. "It's essentially a store-and-forward mechanism that relies on repeat access to generate enhanced performance," The Yankee Group analyst Neal Goldman explains. To understand how an enterprise CDN works, imagine the architecture of a public CDN like Akamai's, which relies on 13,000 servers (caches) located in datacenters and ISPs around the world. The network uses a "pull-based" mechanism to keep local caches primed with the most frequently requested content. For example, when a Web surfer connected to an ISP in Boston clicks on a photo of New England Patriots quarterback Tom Brady on CNN.com's Web site, the photo is "pulled" to the Akamai server closest to that user's ISP. Anyone else in the network who subsequently requests Brady's photo gets it from the same cache, rather than CNN's origin server, thus cutting down on bandwidth and eliminating router hops. Meanwhile, infrequently accessed content and content that has passed its "time-to-live" freshness date are regularly flushed out of the network. And that's not all: CDNs can also monitor network conditions to calculate optimal routing, minimize packet loss, balance network loads, and provide route-around and fail-over capabilities. Of course, in the enterprise the challenge is a bit different. Although the environment behind the firewall may be more predictable, and the universe of potential users better known, eCDNs (enterprise CDNs) must also provide more controlled distribution of content because the consequences of network disruption within a company are greater than on the public Internet. As with a public CDN, an eCDN consists of load balancers, caches, and proxies, all coordinated by a single software system. Typically, an enterprise will also deploy management tools to control the content flowing across the network. Publishing tools can interface with content management systems to inject the appropriate content into the network, such as material in the correct language for a particular geographical region. Bandwidth management tools allocate scarce WAN bandwidth among various users and applications. Routing tools set cache hierarchies so that, for example, if a file is not found in Kuala Lumpur, the system can check for it in Tokyo. Security and access tools restrict access to certain users or until certain times. And monitoring and reporting tools oversee who is using what and bill bandwidth costs to the right departments. Things gets tricky when you try to distribute live video, which typically consumes a huge amount of network capacity. Given that most Web video is delivered using real-time streaming protocols like RealVideo or Windows Media, the process of pushing that content efficiently to the edge of the network requires special software proxy caches that can "multicast," or receive one copy from the origin server and pass it along to multiple simultaneous viewers. This approach enables wider distribution than unicasting, which broadcasts all streams directly from the origin server. For example, Steve Jobs' latest Macworld keynote Webcast topped out at 16.5Gbps and was served simultaneously to a total of 80,000 users. Without multicasting, that would not be possible. Alternatively, enterprises can address the challenge of live video by offering it on-demand, thus flattening the demand curve and making the content easier to cache. How effective is that strategy? Jim Ni, director of product management at CDN vendor F5 Networks, thinks that, eventually, on-demand distribution will be an effective substitute for live distribution. "After the second CEO Webcast, the guys in Singapore aren't going to wake up [in the middle of the night] to watch it live [anymore]," Ni predicts. The City of San Diego is a case in point. San Diego now distributes all its city council meetings via Webcast, using an eCDN from Network Appliance to also deliver the videos on-demand. Fire and police training videos, as well as videos of the city's sewer system for maintenance personnel, are also available. To Allen Myers of the city's data processing group, San Diego's eCDN makes sense because it relieves the city's network of gargantuan video files from what would otherwise be a drag on response times. And as more business-related video and voice technologies come online, the need for eCDNs is only going to increase. "You're going to suck up the bandwidth with the next generation of apps real quick," Myers says. New frontiers for edge delivery With each passing day, public and corporate Web content becomes more and more dynamic or database driven. At the same time, the distributed computing revolution rolls on. In response to both of those trends, CDN and eCDN vendors have begun developing techniques to enable the caching of dynamic page elements and, in some cases, even applications. For example, a group of vendors including Akamai, Oracle, and BEA Systems has introduced a markup language called ESI (Edge Side Includes) that enables edge servers to recognize, cache, and render specially tagged dynamic content or perform data transformations at the edge of the network, based on various environmental variables. "I could look at a user's cookie at the edge of the network, note that they're a Libra, and deliver that horoscope," explains Kieran Taylor, director of product marketing at Akamai. "Without ESI, a single piece of dynamic content on the page renders the page uncachable." Other vendors, such as F5 Networks and Network Appliance, have developed their own ESI-like algorithms. And some industry leaders, such as Cisco and Microsoft, have been conspicuously absent from the ESI parade, leaving open the question of which standard will benefit from the heavyweights' support. Furthermore, to improve the performance of browser-based applications (such as ERP and CRM systems) from remote offices, vendors are working on software that can identify and serve as much of the presentation layer and underlying data as possible from the edge. "The end-user thinks the application is slow, and it's really the network," observes F5's Ni. "The first baby step is to take the 100K bitmaps, and [cache] them out there [on the edge]." Other eCDN vendors are claiming to already support the caching of XML constructs while holding the session with a remote database, and even on-the-fly transformation for wireless formats. It's all part of a return to rich-client computing, according to industry watchers such as The Yankee Group's Goldman, who predicts a rise in locally empowered applications and clients with data and application logic, as well as a built-in presentation policy. "[The CDNs] have to be able to deliver code at the edge," Goldman says. In a year or two, CDNs may even be able to deliver JavaServer Pages (JSPs), VBScript, Active Server Pages (ASPs), and possibly Enterprise JavaBeans (EJBs). "And the more the network is distributed across multiple peers, the more complex management and monitoring becomes," Goldman adds. These developments also put eCDN vendors on a convergence course with the storage world. "One of these large branch office caches can have two gigabytes of storage," notes Amit Pandey, senior marketing director at Network Appliance, which also sells storage. "That looks to me more like a file server." Finally, a raft of startups is emerging to build the next generation of edge-delivery technology -- always an encouraging sign for an emerging technology. From low-cost, peer-to-peer architectures (such as those marketed by Kontiki and CenterSpan), dynamic content caching solutions (Xcache Technologies, Chutney Technologies, and SpiderSoftware), and packet-level caching systems (Expand Networks), to "pre-fetching" tools (Fireclick), "difference send" accelerators (FineGround Networks and Speedwise) and WAN management solutions (NetReality, Response Networks, and NetScout), there's no shortage of innovative thinking these days. As long as you don't mind living on the edge. See our related illustration, "Caching on the edge". Return to our Network management package. 
SPONSORED WHITE PAPERS 
SPONSORED LINKS
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||
|
||||||||||
