An Approach To Network Qualification

1 INTRODUCTIONQ ualification and validation activities in a Good Manu-facturing, Laboratory, or Clinical Practice (GXP) environ-ment traditionally have been about ensuring that a systemhas been designed and implemented as intended. This is typ-ically achieved by employing good practice requirements -gathering, designing, and building the system - then per-forming a series of tests that independently verify the systemhas been installed properly, operates as expected, and per-forms as designed. During these activities, the systems re-quirements, design, and testing are all documented. Today s applications are distributed applications that fol-low a client server model where multiple users on differentworkstations generate and access data that is stored in a cen-tralized data repository. The diagram in Figure 1illustratestypical, distributed GXP client server model has increased the dependence ofapplications on the Network infrastructure making the net- work infrastructure an integral part of the application.

2 Thispaper provides an Approach for qualifying a Network infra-structure. Servers and workstations are purposely omittedfrom this paper for the sake of clarity and to specificallyfocus on the Network hardware and cabling 101 There are many possible Network components and net- work configurations. In general terms, a Network intercon-nects multiple host computers through some medium. Inmost corporate networks, copper cabling connects worksta-tions, servers, printers, and other Network devices to a net- work concentrator. Data is sent from one host to another viathe Network concentrator. One of the most prevalent Network technologies in usetoday is Ethernet. Ethernet uses hubs or switches as networkconcentrators, but in very different ways. Ethernet hub tech-nology is analogous to a room full of people all trying tocommunicate with each other; only one person can speak ata time.

3 It is an inefficient way of communicating and it isbeing phased out in favor of Network switches. The Network switch model is analogous to the telephonesystem in which a call is established between two individu-als. Additionally, most enterprise-level Network switches canbe managed. In other words, the switch will have a user in-terface that allows the Network administrator to configurefeatures, such as routing schemes, diagnostic alerts, networkconnectivity, and fault tolerance, just to name a few. Thesefeatures are specific to the vendor and the feature set pur-chased. Hubs and some lower end switches cannot be man-aged and are rarely used within the corporate in corporate networks, there are a large num-ber of switches spread across the facility. The facility con-tains Network closets, strategically located, each with a set ofswitches serving a specific area of the facility.

4 This designaccommodates the limitation of Ethernet running over cop-per cabling, which has a practical limitation of 300 feet. Allusers must be within 300 feet of the Network closet or theywill experience a variety of Network errors and performanceissues. The switches within each closet are interconnectedusing either copper cabling or fiber cabling. Additionally, theclosets themselves must be interconnected and can use cop-per if the Network closets are within 300 feet or they can usefiber if the closets are within 2000 feet. Network Interface Cards (NICs) are used to connect allAn Approach To Network Qualification BY ART KOZEL AND RUBEN ARMAS1network devices to the Network switch. TheNICs inside the device have a port for the net- work cable. The cable is used to connect thenetwork device to the Network outlet on thewall. The outlet on the wall is connected to anetwork cable that goes back to a patch panel inthe Network closet.

5 The patch panel is a boardwhere all the cables are terminated. From thepatch panel, another cable is used to connect(patch) the panel port to the actual on the networking protocolbeing used, the Network protocol will havesome kind of Network addressing scheme. Typ-ically, most corporations use TransmissionControl Protocol/Internet Protocol (TCP/IP) astheir base protocol - although they may useother protocols in addition to TCPIP. TCPIP isthe same protocol used exclusively on the Inter-net. The TCPIP configuration is a very impor-tant aspect of the Network because Network de-vices must have the correct TCPIP configura-tion to communicate with each other. More-over, administrators may segregate traffic usingTCPIP address schemes and Virtual Local AreaNetworks (VLANs) to improve the perfor-mance of the overall Network . Routers and spe-cial Network switches, called layer threeswitches, are used to route Network traffic fromone VLAN to aspect of a Network is the speed atwhich a Network transmits data.

6 Modern dayequipment is capable of transmitting data atleast 10 megabits, and up to 1000 megabits persecond. When connecting Network devices, thedevices must be able to negotiate a commonspeed at which they can communicate. Mostswitches will be backward compatible, whichmeans that a switch may be capable of speedsof 1000 megabits per second and will also sup-port older devices capable of only 10 megabitsper section is meant to describe a typicalcorporate Network infrastructure and provide acommon understanding of the different net- work components. There are other varieties ofnetwork types and protocols still in use of the Network must be cus-tomized to the Network currently in use. In ad-dition, there are Wide Area Networks (WANs)Art Kozel and Ruben ArmasFigure 1_____Typical Client Server Network Configurationthat use local Internet Service Providers (ISP) to interconnect the differ-ent geographically dispersed sites that are not discussed here.

7 WANs areavailable in many varieties and the reader is referred to the IEEE web-site at for more information. Network Analysis and LifecycleWhether implementing a new corporate Network or retrospectivelyqualifying an existing Network , there must be a complete Network sys-tems analysis performed. The Network analysis must investigate anddocument all aspects of the Network configuration. The easiest way ofperforming such an analysis is by following the natural procession ofthe Network lifecycle. Networks, just as traditional software applications, follow a lifecyclethat includes requirements, design, testing, maintenance, and eventually,retirement. The documentation of each of these components is crucial tomaintaining the qualified state of the Network infrastructure. Moreover,the documentation can also be used to troubleshoot issues, make smartdecisions about Network reconfigurations, determine which componentswould need to be re-qualified during upgrades or changes, and ensurethat all Network engineers have the documentation necessary to under-stand and maintain the corporate general terms, the requirements phase documentswhat needs to be accomplished and not how it will beaccomplished.

8 In addition, the requirement phase docu-ments the minimum specifications that must be metwhen purchasing, installing, and configuring a networkdevice. These requirements outline the minimum perfor-mance parameters such as negotiated speeds, networkinterfaces, error handling, reporting mechanisms, ca-bling requirements, and maximum Network segmentsize, as well as any specialized requirements. An important aspect of any requirement is that itmust be testable and verifiable. Vague requirements can-not be tested and will, therefore, cause problems duringthe testing phase. As a rule of thumb, a good require-ment is one that can always answer the question "Howwill this requirement be tested?"McCabe breaks down requirements into three differ-ent categories, User Requirements, Application Re-quirements, and Device Requirements [McCabe, 2003].

9 User requirements are those requirements that addressuser needs to accomplish the assigned tasks. These re-quirements include the ability of a user to retrieve andtransfer data to and from a server, logical and physicalsecurity, and Network accessibility. Applications will also have a minimum set of re-quirements. Applications that use multimedia compo-nents, such as video or voice, must have a predictablecapacity and delay that is required. The devices used toaccess the Network and run the different applicationshave their own set of requirements as well. These re-quirements will include items such as Network interfacesneeded to interconnect devices. For example, networkswitches that are used to connect Network closets mayrequire fiber interfaces as opposed to copper interfaces,and servers may require a per second connection of 1000megabits, while workstations may only need a 100megabits per second connection.

10 Requirements specifications must be customized tothe particular corporate environment. The items men-tioned here are not a complete and exhaustive list of re-quirements, but are intended to provide a general under-standing of what should be included in a Network re-quirements documentation provides the detailed documenta-tion that informs Network engineers how requirements are tobe implemented, and more importantly, describes the net- work architecture. Network designs typically have both atextual and graphical description of the Network being qual-ified. For example, a general Network topology diagram is aview of the Network that illustrates how all the Network back-bone devices are interconnected. A different view of the net- work may include a topology diagram that illustrates howthe different hosts and Network devices, such as printers,connect to the Network backbone.