The explosive popularity of Web browsers creates unprecedented high demands for decentralized and communication-based business applications. Since the Web technology is evolving, either the way people access information from the Net or how organizations run their businesses through the Web has been changing dramatically. This paper focuses on the key issues to the success of a Web server which has to enable the enterprise to meet the requirements in today's network-centric computing environment, including integration between the Web and legacy data, authoring and managing tools, and scalable server technology.
There is no doubt that the World Wide Web is the most popular and exciting medium to access Internet/Intranet because of its well-adopted, platform-independent, and easy-to-use interface. The universal acceptance of Web technologies increases the demands for communication-based applications. In the past, all leading-edge commercial applications with high performance and reliability such as online transaction processing system (OLTP), information processing system and decision support system were run on legacy systems. In fact, most of today's mission critical real-time transactions still rely on those legacy data, be it customer order records, financial information, or employee data. The challenge for today's enterprise Information Technology is to integrate legacy system with the Web interface. By doing so, they can efficiently broaden the scope of communication without the complexity and high cost associated with the traditional two or three tiers client/server model. The deployment of Web enablement or so called "legacy linkage" is very crucial for the enterprise to utilize the ubiquitous Web environment without re-engineering the database and application codes that reside on a reliable and scalable environment (such as in mainframes). To bridge the gap, database vendors, Web server and application developers have been providing various middleware solutions. The infrastructures of those solutions are based on some common ground but with different approaches. The first part of this paper will talk about the legacy system connectivity.
With the ability to unleash legacy data on the Web,
it then comes down to the issues of manageability and serviceability
to account for a successful site. Web server needs to work with
excellent authoring and developing tools in order to offer a high
quality of service. These tools must be powerful enough to design
and maintain all the static and dynamic information with ease,
yet to be simple and straightforward to reduce the development
costs. Meanwhile, the Web community is exploding. Millions of
people and families are involved with Web surfing and shopping
in their everyday life; more and more large scale commercial and
e-commerce Web sites are going to be launched. The true test for
the Internet is gathering on the horizon. Web servers must be
able to provide service to a large amount of concurrent users,
i.e., it's the scalability that will dominate the competition.
We will introduce some authoring tools as well as scalable Web
server technologies in later sections.
Figure 1: shows a high level system model of a scaleable
Web service with legacy linkage and authoring tools support. We
will discuss each of these pieces in this paper as follows: In
Section 2, we analyze the Web middleware architecture and some
approaches currently available. In Section 3, we introduce the
Web authoring tools and their components. A few scaleable Web service
technologies are described in Section 4. The paper concludes
at Section 5.
2. Web Middleware Architecture
A middleware is a gateway program between applications
and databases. Web middleware connects the legacy Information
System and the Web server by allowing them to exchange messages
and data. Normally, the middleware architecture includes a manager
component that allows applications and database-specific drivers
to call upon a standard Application Programming Interface (API)
that acts as a pipe for accessing these resources. There are some
well-known industrial standards such as Open Database Connectivity
(ODBC) from Microsoft, Distributed Computing Environment (DCE)
from Open Software Foundation, Distributed Relational Database
Architecture (DRDA) from IBM, Enterprise Database Access (EDA)
from Information Builders, and Java Database Connectivity (JDBC)
from JavaSoft. Take Microsoft's ODBC interface as an example,
it defines a minimal common API for relational databases and is
based on Object Linking and Embedding (OLE) technologies. On the
contrary, OSF's DCE is heavily based on Remote Procedure Call
(RPC) for UNIX.
In a general Web access model, as depicted in Figure
2, the Web server must be able to interact with legacy IS, where
the precious data is, in addition, to communicate with front-end
browsers through the HyperText Transfer Protocol (HTTP) over Internet's
TCP/IP protocol. The fundamental techniques to interact with the
front-end browsers could be Common Gateway Interface (CGI), Server
Side Include (SSI), Application Programming Interface (API), and
Mobile/Component-based Code (e.g. Java, ActiveX). On the back
end, it usually requires a connection gateway to deal with network
protocols (shown in Figure 3), such as the System Network Architecture
(SNA) or Advanced Peer to Peer Communication protocol (APPC) for
IBM hosts. The Web middleware can provide either a data or interface
gateway for the back-end, depending on the functionality. It can
be a stand alone server or, be integrated as part of the Web server.
Note that TCP/IP has always been the native communication for
UNIX systems. Therefore, legacy UNIX system and mainframe can
support generating HTML and linking with Internet browser clients.
IBM has already supported TCP/IP on both MVS and VM as well as
HTTP, which makes mainframe itself, a Web server. IBM System/390
thus transforms the mainframe into a powerful client/server platform.
The Web middleware solutions mainly fall into the
following categories. First, it is a Web server that supports
major databases. All the major Web server vendors have integrated
their servers with proprietary APIs (e.g. NSAPI from Netscape
and ISAPI from Microsoft) such that they are capable of supporting
one or more commercial DBMS directly. Examples are Microsoft's
Internet Information Server, IBM's Internet Connection Secure
Server, Netscape's LiveWire Pro, and Oracle's WebServer. These
servers support popular database servers such as Microsoft's SQL
Server, IBM's DB2, Informix OnLine
Server, and Oracle7. Secondly, most popular commercial
DBMS vendors like Informix, Sybase, Microsoft, Illustra, and Microrim
all supply tools for Web connectivity. From the early stage, facility
like Informix-ESQL/C, a CGI interface kit and Sybase's web.sql
which is based on SSI and embedded SQL, to R:WEB of Microrim which
integrates the auto-generation of HTML pages from their R:BASE
Many have been working on gateway solution for accessing
legacy applications and databases. There are many varieties with
different focuses: Microsoft's SNA Server specializes in leveraging
investments in the IBM Hosts by providing LAN-to-SNA connection
gateway and ODBC/DRDA driver. Sybase's CONNECT family delivers
a middleware solution to link any data source or application into
IS through APIs. XDB's HeatShield and IBM's Net.data implement
for mainframe DB2. HeatShield is designed to offer PC-based DRDA
link with support for ODBC and JDBC driver managers. Net.Data
builds on the strong database access and reporting capabilities
of DB2WWW Connection which is a CGI-BIN application and uses Web
macro file approach combined with language environment support.
NEON's Shadow Direct Servers aim at mainframe data sources, too.
It integrates the client ODBC application with CICS (Customer
Information Control System), IMS (Information Management System),
and DB2 on the MVS server. NeXT's WebObjects supports direct mainframe
database access through "Data Source Adapter". Its Conextions
Builders Adapter Web-enables the mainframe 3270, 5250 applications.
IBM's CICS Internet Gateway, Simware's Salvo Server, Teubner &
Associate's Corridor and Attachmae's Emissary Host Publishing
System all have similar approaches to use "screen scraper"
to translate 3270 data stream into HTML on-the-fly.
A conceptual representation of middleware architecture
is shown in Figure 4 as a summary. Various component(s) included
in different middleware approaches are by no means exclusive solutions.
More innovative and integrated techniques are surely to come. At
the same time, the Intranet revolution is emerging behind the
company firewalls. Intranet is a cost-effective and versatile
solution to enhance group collaboration and productivity for enterprises.
By integrating Web browsing, legacy database/application access,
and services like groupware and email, Intranet can efficiently
coordinate department projects and workflow. Netscape, Microsoft,
and Lotus all provide their own integrated server suite to meet
the Intranet requirements: Based on ONE (Open Network Environment),
Netscape deploys a platform-independent, open standards-based
services offered by the full service Intranet named SuitSpot.
The cross-platform/database access is supported through ODBC
connectivity in addition to native support for database of Informix,
Sybase and Oracle. Microsoft, on the other hand, utilizes their
dominance of PC operation systems and continues to push Window-centric
platform. Their BackOffice is designed to improve streamline business
processes with a choice of client-server solutions that work on
the Internet. It employs SNA Server for host integration which
extends company LAN on a scaleable Windows NT platform. Lotus'
Notes Server is another next-generation Web server which fully
supports Web browses and Notes clients. It uses ODBC as a standard
interface between Notes and DBMS, and also includes specific
integration with Oracle, DB2 and Seabees SQL Server. The tightly
integration with CICS-based transaction processing systems is
achieved through the use of IBM's MQ Series for Notes.
3. Web Authoring Tool
The visual presentation and site organization play
important roles for a successful commercial Web site. Advanced
Web publishing and authoring tools can help to create attractive
Web face with a powerful functionality. It provides a WYSIWYG
design environment which requires little or even no knowledge
of HTML at all. One example is Netscape's Navigator Gold, making
easy the design task which used to be very tedious and time-consuming.
A word-processing style page editor simplifies the job to specify
proper tags all over the document. One simply enters the contents
of the pages using conventional formatting styles like variable
font sizing and text alignment, and it transparently translates
them into corresponding HTML scripts. Images can be inserted into
or removed from pages as easy as pushing a button then selecting
the name of file. It also integrates the process of uploading/downloading
pages to/from the Web server without a separate FTP program. Microsoft's
Internet Assistant for Word offers a similar functionality by
enabling Web attributes on top of the Microsoft Word.
However, the page-oriented HTML layout editor is just the first step. Since Web page embeds many hyperlinks, either a local file or a remote URL, an e-commerce site usually has a very complicated road map in which pages reference one another. To well maintain the system hierarchy, a much more powerful authoring tool is necessary. Emphasis must be put on managing
the relationship between pages in the site. The easier
to create all pages from scratch as well as to update them afterward,
the more productive the site will be. In this regard, systems like
Adobe's PageMill/SiteMill and SGI's WebMagic are sort of enhanced
media-rich page design tools. NetObjects Fusion from NetObjects
uses visual, site-oriented approach toward Web design which treats
each Web site as a whole and allows users to edit all of its aspects
from overall structure to individual page components without any
HTML coding. It provides centralized control over external file
such as images, sounds, and applets as well as site-wide elements
like headers/footers, buttons, and links. If a user edits any
one piece, NetObjects Fusion automatically updates for the entire
site. Microsoft's FrontPage 97 also keeps site-design in mind.
Like NetObjects Fusion, it offers some advanced professional templates.
It does not have the tight integration between Web pages as NetObjects
Fusion does, but there are some nice tools inside a separate bonus
package. such as Image Composer and Publishing Wizard.
The trend of Web authoring is going to be more user-friendly,
visualized design tools with drag-and-drop, point-and-click, and
auto-management. Flexible and intelligent pages/files maintenance
will efficiently keep the business-related information up-to-date
and consistent. As the sites are getting bigger, fancier, and
the reference links are frequently getting updated, it really
makes a lot of difference by employing a professional authoring
4. Scaleable Web Service
Given the exponential growth of Internet and Web
community, it is increasingly more difficult for organizations to
properly predict what Web server needs in the future regarding
both resources and hardware requirements. On the other hand, large
business Web sites need far-flung, intense demands in order to
make them profitable. As a result, a Web server must be able to
grow with a seemingly endless increase in the number of user requests.
In other words, a scaleable Web service architecture is needed
for high availability and quality of service. Also, a scaleable
system can eliminate the single point of failure inherent in single-server
configuration. In this section, we will discuss the issues of
building a scaleable Web service.
National Center for Supercomputing Applications (NCSA) has defined a combination of one-to-many mapping facilities necessary for building a scaleable Web service:
Generally speaking, Web salability is based on the
flexibility to change the number of Web servers dynamically without
interrupting Web service. To achieve that, the correspondent name
mapping should be resolved dynamically too. NCSA's implementation
allows for dynamic scaleability by rotating through a pool of HTTP
servers that are alternately mapped to the hostname alias of the
Web server. It includes a cluster of identical configured servers,
a Round-Robin DNS for distributing requests across the cluster,
a distributed file system (AFS) to maintain a synchronized set
of documents across the cluster, and method to administrate the
cluster. Because these technologies are inherently platform-independent
which allows new types of server architectures to be easily integrated
into the HTTP service. Besides, the configuration is transparent
to the client.
Oracle's WebServer is based on a secure, scaleable
architecture. The core of WebServer is the Web Request Broker
(WRB), a high-speed mechanism for dispatching, load-balancing
and adding third-party server extensions. Through the unique independent
processing architecture, the WRB guarantees that third-party server
extensions will not affect other parts of the system, thus delivering
unparalleled reliability for users. An Oracle7 server may optionally
be integrated for increased data processing power and scaleability.
Oracle WebServer can translate and dispatch client information
requests directly to the Oracle7 applications server using PL/SQL,
Oracle's procedural language for Oracle7. This approach ensures
that dynamic, data-driven Web applications, such as Oracle Applications
for the Web, run much faster with Oracle WebServer than conventional
CGI-based Web servers.
With the proliferation of multimedia applications
and the need to support thousands of concurrent video streams,
highly scaleable and available multimedia servers are necessary.
IBM's solution for a robust, highly scaleable Web server is to
use scaleable parallel computing. This technology offers flexibility
by joining together from two to hundreds of computer processors
to break down complex, data-intensive jobs to speed their completion.
Parallel architecture can achieve computing power once available
only in high-end mainframes while getting the benefit of scaling
flexibility at the same time. IBM Scaleable and High Available
Web Server offers services via a Scaleable POWERparallel System
SP2 or a cluster of RS/6000 workstations. This server is built
to support a large number of concurrent users, and has support
for scaleability, high bandwidth, real-time multimedia delivery,
fine-grained load balancing and high availability. Some of the
technologies has already been used in the Web Server for 1996
Moreover, IBM's high-end parallel processing system
S/390 Parallel Sysplex is another candidate platform for scaleable
Web serving. Because in the S/390 Parallel Sysplex environment,
processing capacity can be added in granular increments; from
the addition of single processor within an existing system to
the introduction of one or more data-sharing systems. New system
can be introduced into the Parallel Sysplex in a non-disruptive
manner. Parallel Sysplex data-sharing technology enables systems
to be added to the configuration with near-linear scaleability
and nearly-unlimited capacity. Plus, IBM's commitment to open
system had made MVS compatible with UNIX and supported HTTP. It
is a promising platform to build a scaleable Web server architecture
on legacy host, utilizing the highly reliable and available environment.
Web protocols make it simple to standardize the process for information access. However, it requires well-designed integration to run a successful commercial site; including legacy linkage, powerful management and scaleability. In this paper, we have emphasized on these issues with the demonstration of current vendor solutions. They are certainly going to be the essential parts for the future of Web.