Specialty Web Servers
by Don Larson Whether webmaster of a cutting-edge commercial site or member of a corporate IS team, you are constantly running the gauntlet between budget concerns, infrastructure constraints and the next big thing perpetually looming on the horizon.
To forge third-generation Web sites and stay on top of the "third wave" of the Net, you must provide dynamic and interactive applications, create collaborative environments, and foster a sense of community. Unfortunately, the technologies needed to create such sites-streaming media, real-time communication, virtual worlds-rarely come without a price. Beyond their costs in monetary terms, new applications and media types usually mean new software, special servers and, before you know it, more hardware and bandwidth. Meeting these prerequisites, however, is only half the battle with many types of media, which often demand new content production and extensive management after they're up and running.
To further complicate matters, new players are constantly cropping up in the special server market. Once, with relatively few companies in the niche, Web managers faced a fairly straightforward decision: either put your money on proprietary technology, or stick with the more traditional means of delivery already in place. After all, the one nagging question lurking behind special-purpose servers is: "If I can simply post an .ra or .wrl file and people can access it through my Web server, why should I invest in a proprietary server?" Now, with giants like Netscape and Microsoft entering the mix by actually giving away streaming media and conference servers as part of their respective suites, the question becomes even more basic: "Are even free servers worth investing valuable time and energy?"
Questions and factors like these have long led those responsible for a site's future directions to question whether the payoff of added functionality and interactivity that special-purpose servers bring to a site is worth their price-both in terms of money and resources. To help answer the questions, Web Developer� scans the spectrum of special servers-which do everything from multimedia delivery to conferencing, collaboration and even virtual communities-to see whether they tip the scales on the side of pain, or gain.
Streaming media are still the hot technology where multimedia on the Web is concerned. Whereas single-track media (such as audio or video) was once the standard, the latest trend is to move towards true multimedia applications: those that can deliver time-based, synchronized content containing audio, video, images, animation and even script commands in a single data stream.
The "old" players-RealAudio, VDOnet and Xing, for instance-are still in the game, of course. But even veteran streamers like Progressive Networks (RealAudio) are trying to leverage their existing market share by introducing their RealMedia Architecture (RMA) platform for delivering multiple data types.
More recently, Internet giants Microsoft and Netscape shook up the streaming server market when they began bundling special media servers-NetShow and Media Server 1.0, respectively-with their server suites. And new entries, like VXtreme's Web Theater, have garnered attention by introducing even higher compression and enhanced delivery technologies for streaming video content. Despite the recent emergence of these platforms, the promise of delivering applications like illustrated audio and video training programs, guided tours of a Web site and other true multimedia presentations over the Internet and intranet has many content developers very interested. In this section, we'll take a comparative look at the features of each platform and discuss some of the potential drawbacks that these new technologies present.
Before detailing the specs on each platform, however, let's take a quick look at the general benefits of special-purpose media servers by answering that nagging question: "Why not just send it through the Web server?" In fact, many of the products just mentioned, including RealAudio and Netshow, support Web server workarounds (what Progressive Networks calls "HTTP Pseudo-streaming"). However, as we explained in our special multimedia issue (see "The Dark Side of Multimedia,"
Web Developer�, Sep./Oct. 1996), Web servers utilize "bursty" transmissions to send files.
"Bursty" refers to the fact that the data is transmitted, at random times, in chunks which take up all available bandwidth temporarily. HTTP daemons, therefore, do not make optimal use of network capacity, or allow you to control the rate at which particular files are sent out, as many of the special media servers do.
Furthermore, by streaming through an HTTP daemon, you use TCP/IP to deliver the content. Because TCP/IP is a "reliable" protocol, it delays transmissions while it recovers lost packets, causing significant interruptions in the broadcasts and thus a poor quality of reception. On the other hand, "unreliable" streaming protocols like Real-Time Streaming Protocol (RTSP) and Real-Time Protocol (RTP), which are layered on top of older protocols such as UDP, automatically compensate for packet loss and deliver a best-effort quality broadcast.
Our multimedia issue also introduced you to the basic platforms of established streaming media vendors such as Xing, VDOnet and Progressive Networks. However, these companies have recently added new technologies to their server packages. Xing now offers a PlusPack upgrade for its StreamWorks server, which adds propagation capabilities for distributing streams across multiple servers. Similarly, the 2.0 version of the VDOLive server, along with better video quality and faster compression, has an added "storybook mode" for synchronizing audio and image tracks.
New to the RealAudio 3.0 server are a number of features designed to reduce bandwidth consumption, including support for IP Multicasting, which shares a single data stream with multiple listeners, and RA's Splitter Technology for stream propagation. Even more substantial is Progressive Networks' introduction of the RealMedia Architecture (RMA), a streaming media platform with which the company hopes to leverage its existing base of RealAudio products. Utilizing a server component called the RealMedia File Filter, a RealAudio server is able to convert and stream a variety of third-party data types, including audio, video, text, images, animations and even MIDI.
Media Server currently offers streaming support for two common file types, WAV and AU, a fact which is sure to bring a smile to the face of any content creator. The server also supports RealMedia and Netscape's new file format, LiveAudio. Media Server also offers browser-based administration through a similar interface as other Netscape servers, as well as sitewide performance and traffic statistics.
Available as a free add-on to the Internet Information Server (IIS), Microsoft's NetShow platform delivers functionality similar to that of its closest rival, Netscape's Media Server, with the noted exception of its support for video. Like Media Server, NetShow supports content ranging from graphics and URL "flipping" to script commands, as well as technologies like IP Multicasting.
It is in the areas of server administration and content creation tools that NetShow sets itself apart from platforms like RMA and Media Server, however. The NetShow Server can be configured and monitored through the Internet Service Manager included with IIS. This feature gives you enhanced control over the network bandwidth utilization by enabling you to set the maximum content throughput per file and per server, as well as the maximum number of simultaneous clients that can connect to any server. Managing the server at this level has the effect of throttling file transfers to a specific bit rate, so that network capacity won't be "flooded" by simultaneous connections or multicasts. NetShow also includes performance monitoring and traffic logging capabilities for improved server management.
Netshow is bundled with its own production tools, including a GUI tool, the ASF Editor, for creating synchronized files in Microsoft's Active Streaming Format. Even more impressive, although hardly surprising, is the third-party support NetShow has received from some well-established multimedia tool vendors. Only months after the platform's release, companies such as VXtreme, Vivo, VDOnet, Sonic Foundry and Macromedia have already announced support for the ASF format.
Finally, relative newcomers VXtreme recently burst onto the streaming video scene with the introduction of their Web Theater package. Offering improved scalability, the Web Theater server is able to deliver 10 to 20 frames per second at 28.8 Kbps and over 30 fps at 1 Mbps. VXtreme also offers better delivery through advanced "adaptive congestion control" algorithms that dynamically detect changes in available network capacity and provide the best quality of video possible at that bandwidth.
Not to be left behind, Web Theater is also able to produce and serve illustrated video, complete with synchronized text, graphics, HTML and applets. The package comes complete with Web Theater Producer, making VXtreme's platform one of the few to include a full-featured video capture and multimedia syncing authoring environment. As with NetShow, the Web Theater Server gives you significant facilities for server management and administration. For those planning to manage large amounts of video content, an optional interface to an Informix Illustra Server database is also available.
As you can see, these platforms provide the tools and technologies you need to create an exciting new breed of rich multimedia experiences on the Web. Several even give you the means by which you can better manage and control available network resources, even as you add a bandwidth-intensive mix of new content. But for every silver lining, as the saying goes, there's a dark cloud. And hanging over these streaming multimedia platforms are issues such as price, network upgrades and the relative immaturity of the current tools and technologies.
As another saying goes, to play you've got to pay. With the exception of Microsoft's platform, this is true of the media servers included above, all of which are sold on a per-stream basis. Although many of the vendors offer relatively cheap starter packages and a variety of scalable solutions, many real-world applications require robust configurations, able to support up to one hundred simultaneous streams. For the software alone, at their current pricing structures, this 100-stream setup would cost over $11,000 for both RealAudio and VDOLive, and almost $17,000 for VXtreme. Annual support, upgrades and production tools are extra.
Also extra is the additional hardware and, in many cases, bandwidth required to support these streaming media servers. Theoretically, add-ons like Media Server and NetShow-and even some of the per-stream servers-could run on the same physical machine as the Web server for lightweight applications. For optimal performance under anything but the lightest load, however, all of the vendors recommend a separate box for their server.
Along with additional software, many of the new technologies designed to optimize network capacity, such as propagation and IP Multicasting, require auxiliary hardware or upgrades. IP Multicasting, for example, is only possible on subnets in which the routers and switches have been upgraded to support it. For Internet applications, this means access through a provider participating in multicast-enabled nets like the MBONE; intranet and WAN apps would have to add their own multicast-enabled routers, often in addition to the existing production routers (for more specs on IP Multicasting, see www.ipmulticast.com). Finally, the capacity to stream live feeds may require additional hardware and software packages, as in the case of Xing's StreamWorks Transmitter and NetShow Live.
A final factor in the downside of these new server technologies has to do with issues surrounding the relative immaturity of the platforms, especially where streaming multiple data types is concerned. This involves every component of the platform, from server and client to production tools. Although many elements of the RealMedia Architecture are supported in the 3.0 release of RealAudio, for example, full support for the new platform will require upgrades of both the server and player. In addition, both the RealMedia File Filter server component and the client-side player work in conjunction with licensed components, which means third party plug-ins for each new data type you support. This stream of new plug-ins will demand continual upgrades and management, although the initial release of RMA does offer native support for a number of third-party vendors, including Futurewave, Narrative, OLiVR and Iterated Systems.
Furthermore, although plenty of tools for creating multimedia content exist, what is decidedly lacking are tools for combining and synchronizing different media types to create these next-generation multimedia applications. Apart from Web Theater's Producer and NetShow's ASF Editor (which was relatively unimpressive in the initial release we previewed at Fall Internet World, although an enhanced version has since been released), few of the other vendors have introduced their own full-featured authoring tools for developing this type of content. Until server, client, tools and third-party support all coalesce, managing a multimedia server could be a demanding responsibility.
Conferencing and Collaboration
Nothing fulfills the promise of interactivity and dynamically updated content like conferencing, or chat, and discussion forums. Chat was once restricted to the "underworld" of the Internet, found only in the form of cryptic command-line user interfaces connected through Internet Relay Chat, or IRC, channels. However, a new breed of client-server conferencing software recently brought real-time chat rooms to the Web. By extending chat servers to make them Web-compatible and replacing the old interfaces with more user-friendly versions, products like Paralogic's paraCHAT and The Palace's Commercial Server have quickly raised the standards of Web-based chat.
However, the latest trend in conferencing and collaboration is complete, solutions-based platforms. Here, as in the case of special media servers, Netscape and Microsoft have once again made their voices heard by providing turnkey "community" packages that included both chat and discussion group servers. Discussion group servers also provide the means for implementing two-way communication through applications like bulletin boards and private user groups. A close look at the pros and cons of each product will help you determine which solution best fits your needs.
Paralogic's paraCHAT provides the most basic type of conferencing, text-based chat. Although Paralogic provides a free, advertising-based hosting service, the licensed version of the server offers you control over the software, including full customization of both server and client, moderation, advertising and even micropayment capabilities. paraCHAT is written entirely in Java, which means the server will run on any platform that supports a Java Virtual Machine and the client will operate within any Java-enabled browser.
The primary feature that this Java-based architecture delivers over HTML-based chat servers is speed. Since HTML-based chat uses HTTP, each new message sent by a user requires a new socket connection, along with a GET command and often a CGI program to retrieve the new status of the page, or room. Paralogic's method creates a single socket connection, and each new message is sent down this same pipe. paraCHAT is also more efficient where server loading is concerned. "We run hundreds of simultaneous sessions per server, and we have multiple instances of the server running on the same hardware as a fairly busy Web server," reports Paralogic's president Vijay Vaidyanathan. "And we still see over 60 to 70% idle time."
The Palace's Commercial Server represents the new face of conferencing: virtual chat. Like text-based chat on cartoons, real-time virtual worlds like The Palace take chat "rooms" literally, allowing users to assume the shape of avatars, roam through graphical environments and talk to each other via pop-up dialogue boxes. Of course, due to its highly graphical interface, virtual chat places greater demands on network resources and content development than does text-based chat. In addition to the overhead introduced by avatars, movement and graphical messaging, one of the biggest factors in resource consumption involves multiple downloads of the initial graphics, or "props" package, which each new user must have to view the different rooms. In addition, users can upload their own props, which must also be distributed to other users by the server. In both cases, The Palace utilizes caching on both the server and client to minimize the load on network resources. Administration features allow you to keep the size of the cache files manageable.
Unlike text chat, which (as in the case of paraCHAT, can take as little as five lines of HTML to implement), virtual chat requires extensive content development. Although the time needed to create virtual rooms depends upon their number and complexity, a spokesperson for The Palace estimated that average development time ranges from one to three weeks. The majority of this time is devoted to creating the background images that form each environment. To further extend the worlds, developers must master a proprietary scripting language, called Iptscrae, although the server also supports C language plug-ins and Shockwave movies.
Extending these single-server platforms, Netscape's Community System and Microsoft's Commercial Internet System provide complete solutions on which to build interactive applications, including both real-time conferencing and discussion forums.
Netscape's Chat Server operates through a standalone client that works in conjunction with their Navigator browser. This, in effect, adds multimedia capabilities to chat forums, as users are able to include URLs that pull up Web pages in sync with their chat. The Chat Server supports both moderated and private forums, including personal (one-to-one), group (one-to-many) and auditorium-style (one-to-very many) conferencing. The Community System's Bulletin Board Service, an extension of Netscape's News Server, enables you to add both public and private threaded discussion groups. The BBS adds functionality like a customizable HTML front-end, article indexing and searching and SSL-encryption for secure forums. Both the Chat Server and BBS offer full integration with the Netscape Commerce Platform for access control, billing and credit card processing capabilities.
Part of the Commercial Internet System ("Normandy"), Microsoft's Conferencing and Commercial Internet News Servers offer similar features to their Netscape counterparts, with two major distinctions. First, the Conferencing Server ships with an ActiveX control for the client, so chat sessions can be seamlessly integrated into a Web page. Unlike the customizable front-end of Netscape's BBS, however, the News Server works through Microsoft's news client, Athena. The Commercial Internet System also supports similar commerce features through the Merchant Server included in the platform.
Some of the same drawbacks surrounding special media servers are also applicable where conferencing and chat platforms are concerned. Like media servers, this software doesn't come cheap. A licensed copy of the paraCHAT server is $9,500, while a 1000-user Palace Commercial Server costs over $8,000. Again, support and upgrades are extra.
What demands do chat and discussion platforms put on network resources? A
Netscape white paper recommends a machine with at least 128 MB of RAM for their Community System, which requires a minimum 500 MB of disk space to install. A similar white paper from
Microsoft reports that their Chat Server supports data packets roughly half the size of standard IRC text packets, enabling the server to scale to 2,500 simultaneous chatters. The Commercial Internet News Server, the paper continues, can scale to 3000 concurrent users on a dual processor, P160 machine with 256 MB of RAM.
Bandwidth concerns are harder to detail, primarily because consumption is always a function of the number of simultaneous users. However, bandwidth-intensive applications, like The Palace's virtual chat, will place a significant burden on a data pipe. A spokesperson for the company estimated that a T1 line could support up to 500 users, although The Palace uses a DEC Alpha machine with two T1 connections for its main server, which supports peak loads of 300-plus users.
Web-based virtual reality is the ultimate representation of what Gibson had in mind when he coined the term cyberspace. Yet, while serving up "straight" VRML does allow you to create rich, immersive environments on the Web, it is not well suited to interactivity and community. That's where specialized VRML servers step in. Among their many perks, products like Integrated Data Systems' V-Realm 3D Media Server, Black Sun's CyberHub and Worlds Inc.'s Active WorldServer all deliver the one primary feature that straight VRML currently can't: multi-user virtual worlds. By providing a space in which users can interact through various methods like real-time chat, avatar interfaces, multimedia and even (as in Worlds' case) shared-state environments, VRML servers offer a robust platform on which next-generation applications ranging from education to commerce can be built today.
The catch is that VRML servers are also bundled with a significant amount of overhead. As anyone who has waited to browse through a virtual world quickly realizes, 3D environments are the most graphically-intensive sites on the Web. Even with compression, this makes for some very large files, usually on the order of 500 KB to 15 MB. Add to that chat, movement tracking, avatar and object behaviors and (as in V-Realm's case) audio capabilities, and you've got a "high octane" mix that will require a lot of additional hardware and bandwidth to support. VRML servers also have their price in monetary terms. While V-Realm is still in beta testing, a high-end Active WorldServer is priced just under $35,000.
Developing content, however, is probably the most demanding aspect of running a VRML server. Although full-featured VRML authoring tools are finally on the market, creating virtual objects and complex 3D environments is a time-consuming task that often demands the skills of a VRML expert. Servers that support the VRML 1.0 and 2.0 standards, like V-Realm and CyberHub, do enable you to take advantage of the existing and ever-expanding number of virtual worlds on the Web. However, platforms built on proprietary standards, such as the enhanced VRML utilized by the Active WorldServer, constrain your capacity to leverage both existing content and available authoring tools. To its benefit, the WorldServer does provide a feature unique to Worlds' platform by enabling a world's end users to actively participate in the construction of their own virtual objects and scenes.
The Big Decision
In the end, choosing whether to dive into the special-purpose server market will most likely be a function of both the value-added services you want to add and the level of service you have to provide. Budget concerns and resource constraints will probably overshadow many of the benefits for smaller sites and networks, especially where specialized multimedia and VRML servers are concerned. On the other hand, many high-traffic sites and large-scale corporate networks may find that implementing these servers actually makes better use of their existing infrastructure. But the final factor in the equation is this: the Web's not what it used to be. And Web servers, God bless 'em, may be slowly living out their days as the Renaissance men of the back-end.
Reprinted from Web Developer� magazine, Vol. 3 No.2 Mar/Apr 1997 (c) 1997 internet.com Corporation. All rights reserved.