Previous | Table of Contents | Next |
Performance can be measured by reporting the volume the average number of messages delivered per hour, or messages in each hour over a 24-hour period. This measure can be divided further by indicating the type of message (i.e., text only, single/double attachments, read receipts). This information gives network managers a measurable indication of the kind of features the user community requires.
For network planning purposes, it may be useful to measure volume or system pressure, ignoring the number of messages sent and focusing on the number of total gigabytes sent per day.
Manufacturing environments have long used a tiered approach to messaging for distributing the workload of factory floor applications. As environments become more complex, the tiered approach offers additional flexibility.
An entire enterprise can be considered a single department, indicating the need for a one-tier system where clients are tied into a single server or post office. Multiple departments in a single enterprise or a single department communicating with multiple enterprises require routers and gateways to communicate with the world outside. When multiple departments need to communicate with each other and with multiple enterprises, a messaging backbone or messaging switch is called for.
Exhibit 1 summarizes the implementation scenarios discussed in this chapter.
A single department in a single enterprise will most likely deploy a one-tier messaging model. This model consists of a single messaging server or post office that provides all services. It may be as large as an OfficeVision system on a mainframe or a Higgins PostOffice on a Compaq file server running NetWare. The department need only concern itself with following corporate guidelines for networking and any naming standards.
Exhibit 2. One-tier Model.
Exhibit 3. Two-tier Model.
Caution should be observed when using corporate guidelines. It is often simple to apply mainframe conventions when standardizing PC LAN-based applications. Many large organizations tend to forget that the whole reason for deploying desktop computers is to move away from mainframe conventions (e.g., 8-character user IDs) that are nonintuitive for users. Exhibit 2 shows a typical one-tier model within a single department of an enterprise.
As the number of E-mail users grows, or multiple departments need to be connected, an organization will probably deploy multiple servers. This two-tier model can consist of integrating similar messaging systems from the same vendor or from different vendors. Exhibit 3 illustrates a connection between two departments using the same vendor software connected via application routers.
Exhibit 4. Using Application Gateways.
In a typical PC LAN environment using a shared-file system such as cc:Mail or Microsoft Mail, the router acts the same way as the PC. The post office is completely passive. When users send messages, their workstations simply copy the message to the file server as an individual file or as an insertion into a file server database. In either case the PC workstation actually does the work the post office simply serves as a shared disk drive. The router is also an active component, but has no user moving messages. It periodically moves messages from one post office to another without user interaction.
Many enterprises have different departments that have chosen their own E-mail systems without a common corporate standard. To integrate dissimilar systems, application gateways can bridge the technical incompatibilities between the various messaging servers (see Exhibit 4).
A simple gateway can translate cc:Mail messages to GroupWise. A more complex gateway can bridge networks (e.g., Ethernet to Token Ring), network protocols(i.e., NetWare to TCP/IP), and the E-mail applications.
Converting one E-mail message to the format of another requires a lot of translation. Document formats (i.e., DCA RFT to ASCII), addressing formats(i.e., user@workgroup@domain to system:user), and message options (i.e., acknowledgments to read or deliver receipts) must all be translated.
Gateways can emulate routers native to each environment. They perform message translations internally. The alternative to this approach is to place the gateway between the routers as opposed to between the post office this is not an end-user design, it is merely a function of the vendor software (see Exhibit 5).
If an enterprise is large, network administrators may want to make use of economies of scale to handle common administration, common gateways to X.400, and Internet networks. The network administration staff may simply need points in its network where it can measure progress. Gateways from each environment to every other environment can be provided, but this solution becomes costly and difficult to maintain. A better approach would be to use a central switching hub or a distributed backbone, as shown in Exhibit 6.
Exhibit 5. Placing a Gateway Between Routers.
Exhibit 6. A Central Switching Hub.
Exhibit 7. Worldwide Distributed Hubs.
The central switch or hub allows for a single path for each messaging environment to communicate with all other messaging environments. The central hub, if it is relatively inexpensive, can be expanded into the distributed model. This is often done as the aggregate system grows and requires additional performance and capacity.
However, this implementation can be taken to an extreme, as seen by the number of companies that have grown PC LAN/shared file systems beyond their original design. It is inexpensive to grow these systems incrementally, but difficult to provide end-to-end reliability. Most organizations plug the technical gaps in these products with the additional permanent and contract personnel to keep the multitude of routers and shared-file system post offices up and running.
Some organizations have taken this distributed hub approach to the point where they have multiple connections to the Internet and the X.400 world (see Exhibit 7). Some organizations offer the single message switch for their global environment, and their messages are more well traveled than their administrators. A message sent from Brussels to Paris may stop in Los Angeles on the way because of the central switching mechanism. In addition to local switching, the distributed hub allows for redundancy.
Most companies deploy E-mail systems using variations of three architectures: a common platform, where all E-mail systems are identical; a multiple backbone where each E-mail environment has its own gateways; or a common backbone where all systems share common resources. The following sections describe these architectures along with the advantages and disadvantages of each.
Previous | Table of Contents | Next |