When I started my career, my assignment was singular: Support CICS/VS on the mainframe. My clients were few, because online processing was relatively new, but two of the clients had grown into major players in online and were significantly expanding outside their home offices. In fact, these clients had offices spread across a substantial portion of the U.S. and considered their online systems a major competitive advantage, so they wanted to provide their remote offices with online systems’ access. This allowed the branches to be much more responsive to policyholder inquiries and reduced the workload for a central office customer service.
The only way to extend CICS access was through the use of existing telecommunications facilities in the form of voice telephone lines, using modems that operated at either 2,400 bits per second (baud) for point-to-point lines (a line dedicated to one office) or 1,200 baud for multipoint (a line shared by multiple offices). This translates to 300 characters per second or 150 characters per second, respectively. For screens that were often 1,500 to 1,900 characters in size, response time ranged from tens of seconds to minutes, hardly brisk response time. Yet compared to mail, customer response time was lightning-fast.
It was obvious existing telecommunications facilities were insufficient to support online access long term, and data/voice had to be segregated. IBM had created a network design tool that would be considered primitive these days but groundbreaking then. I was assigned to create some network designs, and a lot of considerations like capacity (especially full-duplex vs. half-duplex), response time degradation point and error rates had to be calculated by hand. It was my first real deployment of queuing theory, a great opportunity to work with the tool’s developers, and with the client’s help testing, we provided remote offices with a new, invaluable tool.
We’ve Come a Long Way, Baby
Arguably, networks have evolved faster and farther than computers, because networks have undergone a complete transformation more than once. Early on, almost all media was copper-based using electrical signals, but now fiber-optic links use photons to carry voice and data, and wireless uses radio or microwave technologies for signaling. Quantum transmission may use techniques such as quantum tunneling or quantum entanglement, and there’s really no end in sight as to how communications—network is an insufficient term anymore—may continue to evolve. It’s foreseeable the day will come when it’s all integrated into the human body, far-fetched though that may seem.
Network Design Considerations
Many more considerations go into a network design today than when I designed that first network; networks are a lot more complicated and varied. Initially, response time and capacity were the primary concerns; today, it’s security. Cost is always a concern, and complexity has mushroomed. Fundamental issues to investigate when contemplating the design or redesign of a network include:
Do It Right the First Time
While there’s no sure-fire way to design a perfect network (and there’s no such thing as a perfect network), doing due diligence, using the described tools, infrastructure, procedures, documentation and guidelines as discussed will result in a vastly superior network to one that just evolves as technology, business processes and organizational objectives materialize.
Project planning, a network design discipline, a deep and comprehensive understanding of the organization, in-depth research, attention to detail and hard work will result in a much more responsive and reliable network. Vendor tools and guidance can clarify areas of confusion, and vetting tools such as simulators can comprehensively test and validate network performance and reliability in great detail, allowing administration to not just design for the present, but also for the future.
For the initial network and any substantial redesign, document the network both graphically and textually. The adage “a picture is worth a thousand words” is especially apt for network design, because a network is a topographical entity, so a map—or group of maps showing magnified views—is especially useful when planning or making network changes. Lines between network nodes can be annotated with information like media type, number of physical or logical links, bandwidth, length, attached routers or other equipment, etc., can be invaluable. Supporting documentation, such as written descriptions of details like contact names and numbers, applications, cabling descriptions, shutdown and startup procedures, recovery and problem resolution tools, and feature lists fill in the holes the map omits.
Security can never be strong enough. Virtually every business these days is on the internet, and almost all these data streams include sensitive or personal information. The danger lies not only in data theft, but also the injection of false data into data streams. Encryption is the basic form of protection, and with the right encryption algorithm and sufficient encryption key protection, most data streams can be well-secured. Implementation of a secure infrastructure such as Payment Card Industry compliance is a must.
The network has to perform well. The data volume that’s carried by networks today is magnitudes greater than that of even a decade or two ago. Consequently, network bandwidth has to be magnitudes greater as well, and key to that is the establishment of a comprehensive performance monitoring and management methodology. Incorporating efficacious forecasting into this infrastructure allows for anticipation and non-disruptive bandwidth expansion, and a response time monitoring scheme with supporting enforcement adds controls to ensure interactive traffic isn’t overwhelmed by batch or other low priority traffic, also providing an early warning of approaching capacity shortages.
Do your homework before making any changes. When network changes are required, thorough research and planning comes first. Tailorable, internet-accessible network design checklists can be customized to a company’s specific needs, and an internal library of checklists should be established and maintained. Prior to a substantive design or physical change, carefully review the checklist and use it to produce a project plan and schedule staff. Vendors should be consulted and users should be notified in advance.
Implement and employ a network design tool. Numerous, good network design tools are available, so after gaining a good grasp of the current network design, choose the tool that best fits current circumstances. Existing applications and distribution of function should be carefully examined and documented, so special requirements of different nodes can be assessed and unique procedures or network reconfiguration tasks can be recorded. The final network should be a best fit for the nature of work it supports.
Establish a robust problem determination and resolution framework. When something breaks or goes wrong, equipment should be in place to identify the failing or malfunctioning component, and procedures should exist that provide guidance in fixing or resolving the breakdown or error. A wide variety of hardware and software diagnostic tools are available to assist problem resolution, and training should be provided to network support staff on tool usage, and, when the problem’s in vendor’s hands, how to contact and work with the vendor.
Redundancy eliminates errors. An additional technique to reduce the impact of errors and failures in a network is to build redundancy into the network design. When a network component fails, a backup or partner component can take over the workload, albeit at times in degraded mode. Redundancy doesn’t always mean duplicate hardware and software. Sometimes it can represent functions, such as re-routing network traffic or shutting down optional processes like maintenance functions or utilitarian processes. These redundant process steps should be thoroughly tested and readily available to network support staff.
Thoroughly vet a new or modified network design. Once the design process has been completed, it should be thoroughly appraised and authenticated. Sometimes tools of this nature are part of a network design product, and if so, this is a significant product advantage. The vetting must be comprehensive and provide extensive information regarding throughput, longevity, integrity, reliability and response time. Some products have simulators, and these facilities have the advantage of being able to run different specifiable scenarios and volumes, response times and a wide spectrum of information.
Do It Right the First Time
While there’s no sure-fire way to design a perfect network (and there’s no such thing as a perfect network), doing due diligence, using the described tools, infrastructure, procedures, documentation and guidelines as discussed will result in a vastly superior network to one that just evolves as technology, business processes and organizational objectives materialize.
Project planning, a network design discipline, a deep and comprehensive understanding of the organization, in-depth research, attention to detail and hard work will result in a much more responsive and reliable network. Vendor tools and guidance can clarify areas of confusion, and vetting tools such as simulators can comprehensively test and validate network performance and reliability in great detail, allowing administration to not just design for the present, but also for the future.