Previous Table of Contents Next


Chapter 41
The Hidden Costs of Free TCP/IP

Walker, Richer & Quinn, Inc.

As free TCP/IP software becomes increasingly available, data center managers have new choices to make. Evaluations must consider whether free software provides the connectivity, features, performance, and ease of use required in today’s organization. In addition, “free” software can incur costs in unexpected areas. This chapter uncovers the critical issues when the decision is being made to use free TCP/IP software, discusses interconnectivity issues when introducing a vendor’s TCP/IP software into the organization, and examines the implications for an AS/400 environment.

INTRODUCTION

The growth in TCP/IP is explosive. All major hardware vendors, including IBM, have endorsed TCP/IP. By 1998, it is projected that 59% of all midrange systems, including UNIX and AS/400s, will be running TCP/IP, and desktop installations of TCP/IP will rise to more than 50%. As TCP/IP becomes the established networking standard, management tools and new applications will become even more important, along with proven reliability and strong technical support. Free TCP/IP stacks may be satisfactory for a casual user, but they are less likely to fulfill the needs of complex corporate environments, now and in the future.

Even though a free protocol suite may appear to fill all the organization’s needs, many data center managers have been surprised to discover that that is not always the case, particularly in corporate environments. Not only do free suites usually lack the robust functionality of purchased TCP/IP software, but they also often involve hidden costs. Because an application is only as reliable as the protocol suite it runs on, it pays to review some considerations, outlined in this chapter, before selecting a product.

In addition, companies that analyze TCP/IP stacks should look at both architecture and applications when they choose a stack for their complex enterprises. Architecture makes a difference to the ultimate functioning of the stack in such areas as mobile connectivity and configuration flexibility. Kernel applications, those that are critical to the smooth functioning of an enterprise network that are integrated into TCP/IP, vary widely among stacks. This chapter discusses the importance of kernel applications and identifies specific issues for the data center manager to examine.

Finally, AS/400 vendors and customers have been paying close attention to the proliferation of TCP/IP. Many organizations have opted for the promise of open networking with less expensive implementations and information sharing with other systems, across a common backbone network. Unfortunately, with the promise comes compromise. This chapter illustrates the differences between TCP/IP and IBM’s proprietary networking technology, Advanced Peer-to-Peer Communication (APPC) and illustrates the limitations of TCP/IP in an AS/400environment. The chapter concludes with suggestions for maximizing the benefits of TCP/IP without sacrificing the comprehensive nature of an APPC connection.

SELECTING FREE TCP/IP PROTOCOL SUITES

If the organization is considering a particular TCP/IP implementation, data center managers must ask questions to determine whether the software is designed for the needs of large, corporate networks. Free data communications protocol suites, also known as stacks because they involve more than one protocol, should be examined closely; the following sections discuss critical questions data center managers must ask to uncover possible pitfalls. (Exhibit 1 offers a checklist for identifying high-value TCP/IP software.)

Exhibit 1. A TCP/IP Checklist for Identifying High-Value TCP/IP Software


Software designed for corporate environments:

  Full corporate connectivity with TCP/IP, UDP, LAT, IPX/SPX, SLIP, CSLIP, and PPP.
  VxD/DDL design for performance and reliability.
  Management tools (SNMP MIB II, an event viewer, and diagnostics).
  Support for all existing name and address resolution protocols and current dynamic addressing.
  Full-featured applications, such as VT420/320, FTP client and server, TN3270E and TN5250, Internet and intranet applications.
  NFS capabilities.
  Open connectivity — 100 percent Windows Sockets compliance to ensure compatibility with third-party and Internet shareware/freeware applications.
  Ease of use remote and mobile computing — with features optimized for remote computing (dynamic link recovery, reductions in protocol overhead, increased throughput).

A product backed by a company with solid experience in host systems, networking, and desktop programs. Knowledgeable and ongoing free technical support that includes online services and fax-on-demand information. Complete, easy-to-read documentation. Limited site license agreements that allow free updates for at least one year. Favorable ratings in independent comparisons by industry publications and evaluators.

What Are the Costs of Using a Stack That Is Not Designed for Complex, Heterogeneous Environments?

For the casual user, a free TCP/IP stack may provide all the necessary functionality. If the organization has several networks and users, however, the requirements may far outstrip what a free stack can provide. In this situation, data center managers must consider a stack that is compliant with current standards, that does not sacrifice access to legacy systems, and that delivers the required capabilities for mission-critical computing. Capabilities include:

  Full-enterprise connectivity. True open computing means networking all the organization’s systems, so software must include not only TCP/IP but also UDP, LAT, NS/VT, IPX/SPX, SLIP, CSLIP, and PPP. Selected software should also connect to UNIX, Digital, IBM, and HP hosts. Otherwise, the costs of adding connectivity can mount fast.
  Full-featured applications. The data center manager must check for advanced VT420, TN3270E, TN5250 emulation, FTP client and server, and Internet and intranet applications, to name a few. Accepting a stack without these applications means the organization will have to pay extra for them.
  Ease of use. If the software is not easy to use, the organization must commit additional resources to training and support. Command-line driven operations are not intuitive, so other features to look for include graphical operations and context-sensitive help. If the interface has a familiar look, users can become productive much sooner.

What Is the Cost of Using a Stack That Lacks a Robust Networking Architecture?

Costs result from a multitude of inefficiencies. Areas the data center manager should examine closely are performance, name and address resolution, remote and mobile computing, and open connectivity. Managers should:

  Look for networking software that offers a combined VxD/DLL design. A cooperative implementation can give the superior performance of VxDs and the reliability of a DLL. Of course, the true test is in the organization’s own environment; managers must be sure the stack works with all hardware, including notebook computers, and all applications. The alternative could be a hodge-podge of fixes, which not only cost more but also affect performance.
  Be sure the host name and address resolution protocol currently in use is supported. For example, some implementations ignore DNS. The support must provide access to all applications that the organization plans to run. If not, the data center manager could be in for a significant amount of work. He or she will have to map all the nodes in the network to their specific addresses manually, which can be a significant drain on the budget. Dynamic network address assignment is another consideration. For example, some stacks support only DHCP, whereas organizations may also use BOOTP. In this case, the stack must support both.
  Consider developing needs, especially in the area of telecommuting and mobile computing, because today enterprise connectivity increasingly means connecting from anywhere. Free stacks are not likely to be designed with the specialized features needed to support connections to enterprise computers in remote offices or hotel rooms, or over cellular connections. Data center managers must look for a stack that is optimized for these uses, including dynamic link recovery to re-establish a lost connection, and features that reduce costly overhead and increase throughput. Without these optimizations, the organization could have to pay for additional software for these users, as well as incur unnecessary charges for telephone or cellular connections. In addition, establishing one standard inside and outside the office for all connections will simplify support. These optimizations provide efficiencies on internal networks, as well.
  Check support for open computing standards such as Windows Sockets(WinSock), because most corporate computing environments do have a mix of hardware and software. Most vendors claim to support WinSock, but only a stack that is 100% WinSock compliant has passed all the Windows Sockets API Tester(WSAT) without error. This achievement is the data center manager’s assurance that the stack will work reliably with applications designed to the WinSock standard, including third-party and Internet freeware/shareware applications. No assumptions should be made about any vendor’s support; the manager should feel free to ask for information about its WSAT results. Ensuring that applications are supported can save money, and spare staff the resulting frustration when network applications fail due to poor WinSock compatibility.


Previous Table of Contents Next

Copyright © CRC Press LLC