Network Administration Got You Scared?Here's what you need to know |
|
From "Special Report", Access to Wang, January 1995 |
|
[ Prior Article ] [ Return to the Catalog of articles ] [ Next Article ] |
The Wang VS and other minicomputers are becoming less dominant in corporate computing, and it's likely that networks - particularly networked personal computers - are a large part of the reason.
The abundance of cheap, powerful PC software has changed our expectations and those of our end users. Many administrators have tried to ignore this change, continuing their VS-based applications with no attempt to integrate them into the PC environment; the results are isolated applications, rekeying of data, difficult translations between unlike systems and discrepancies between information generated by the "legacy" applications on the host system and reports prepared using ad hoc tools such as spreadsheets.
Network administration has many VS administrators worried, and rightly so, since it is more complicated than conventional system administration. But even if your primary business applications run on a Wang VS or other host system, you can still benefit by consolidating information and other resources on a network.
I will concentrate on the types of activities typically required of a system administrator: setting up the host (server), new workstations (clients), users and applications; diagnosing performance problems; and supporting remote connections. The specific commands and processes for each environment will differ slightly according to the components that you select for your network.
To begin with, what is a network?
When asked this question, most system pros would name a number of popular Network Operating Systems, a cabling approach or a protocol; but these are only a part of the story. I define a network as a common architecture that connects multiple systems, allowing interaction between applications through a common protocol.
By this definition, any set of PCs communicating with a VS or Unix host could be considered a network, since there is interaction (two-way communication) using a shared syntax, even if the medium is a simple host-to-PC file transfer system. Batch file transfer between systems over switched lines (dial-up) could also be considered a network, as the many users of UUCP transfer (Unix-to- Unix-Copy Protocol) will attest. Naturally, Local Area Networks (LANs) also fit this definition.
I use the term architecture because a network embodies many facets of connectivity, including physical connections, protocol and the capabilities of the system's Network Operating System. The connections between systems may be local (twisted pair, coaxial cable, etc.) or logical (switched lines, packet networks). Protocol choices include TCP/IP, Novell's IPX, Microsoft's NetBEUI and simpler file transfer protocols, such as XMODEM, ZMODEM, Kermit and UUCP. Commercial Network Operating Systems like NetWare, AppleTalk, LAN Manager or Banyan's Street Talk provide a command syntax and the ability to share disks, printers and other resources.
The combination of all of these elements adds value to the systems involved, including these benefits:
Disk drives, printers, modems and other resources can be shared by all users.
PC applications can be run from a single copy of software, minimizing support problems and simplifying most upgrades.
Files can be transferred or shared.
The central file server can be backed up to ensure that critical information is not lost when a workstation fails.
Now let's look at how network administration differs from other system administration.
The steps for setting up a new VS workstation include installing cable, adding the device to the system configuration, adding the user to the access control (security) system and connecting to the host machine. The same steps exist in a networked environment, with additional work in configuring the workstation.
Most networks installed today use Unshielded Twisted Pair (UTP) wire to connect workstations and printers to the remainder of the network. Similarly, most VS installations have shrugged off the difficult dual-coaxial cable of old - the Wang "928" - in favor of UTP and baluns (coax adapters).
The primary differences between these connections are the quality of the medium and the type of connectors used.
Wire used for network purposes should be designated Level 4 or better, the same level recommended far Wang balun connections. In contrast, Level 2 is used far most telephone connections. Level 4 wire is designed fat speeds up to lOMbps (the speed of most Ethernet networks), and Level 5 will support up to 100 megabits ("fast Ethernet"). The difference in cost is insignificant, so many organizations use Level 5 exclusively.
Network connections typically use an RJ45 connection, similar to the RJ11 connection used by telephones but larger, supporting eight wires instead of two or four. (Only four of the eight wires are used to establish a network connection.
An RJ45 "biscuit" (telephone jack) is installed near the workstation and a line cord is used to connect this jack to the workstation. The wires are run to the central wiring closet and attached to the network through a hub, often with a patch panel in between to simplify future changes. This connection to the hub or patch panel serves the same purpose as a connection to the backplane of a VS, adding the device to the rest of the network.
Configuring the host system for a new workstation on a network is usually simpler than on the VS, since you do not usually have to specify the type of device or its physical location; networks rely on the IP address (device number) of the workstation itself to route information properly. Hardware configuration changes do not usually require you to reboot the system, as with the VS.
User setup on a host system is similar to the VS: you must edit the security files to identify the user and his or her rights. Database management systems and some other applications may require additional user setup, however, and you may need to assign privileges between systems besides those within the local system.
Setting up workstations on a network requires more work than on a VS, since each workstation is usually a standalone machine (e.g., PC or other desktop system) with its own software installation. The amount of effort required here will vary, but a minimal setup includes a local operating system, client portions of the network software, a means of connecting to host and server applications (e.g., Telnet, terminal emulation) and any applications to be run from the local workstation.
Since you can use the network to share access to applications (e.g., word processing, electronic mail), you should be able to keep the amount of local setup to a minimum, but you still must have the expertise to correctly configure and support your workstations.
Every network device - servers, workstations, printers, etc. - must be defined with its own unique network location (address) so data can be correctly routed to and from that device. Some networks also require unique workstation names, which must be set up individually. A great many network errors involve accidental duplication of these addresses, so pay attention here.
E-mail and other network applications also require administrative effort, and Network Operating Systems have their own access control requirements. The setup requirements are similar to those of a host system like the VS: user identification and assignment of access privileges.
Think of network services and support for host (terminal) access as server applications, each with individual access control and user identification schemes. If these server functions are run from the same system, these setup needs may he integrated. For example, some vendors have created Unix versions of popular Network Operating System servers (Portable NetWare, LAN Manager) integrated with normal Unix user setup, reducing the amount of administration. In most shops, though, these server functions are separated, and user setup is required on each server.
Your network is up and running, but you're receiving reports of sluggish network response. The problem could be an overloaded server, a poor workstation connection or too much activity. What can you do to try to discover the reasons for the problem?
If you suspect the server is the bottleneck, you must adjust your diagnostic approach to the type of server you are using. Unix systems, for example, have system activity monitors that can often uncover performance issues. PC-based servers are trickier, since there are fewer performance tools in that environment. Frequently, the problem is obvious: overused disk drives, too little system memory available or similar causes.
Workstation connections can use echoing tools like the Unix ping command to time server response. Ping sends test packets to a designated network device, noting the time it takes for the packet to return to the sender. Slow response might indicate a bad workstation connection . . . or a busy network.
Unfortunately, overall network activity is more difficult to track, requiring the purchase of expensive monitoring equipment and software. If you think net activity is the problem, it is usually best to find network specialists and have them survey your system traffic, much as you would hire Wang to perform a system performance audit for the VS.
Nearly all shops support some form of remote access, either a terminal using host-based applications or file transfer. How does this kind of access merge with a networked environment?
The answer here depends on what kind of support your Network Operating System offers for dial-up connections. Most PC-based LAN soffware can meet these needs but all require the user to dial into a specially-configured PC or server, making the cost prohibitive for casual use. Special software products for PC remote communications such as PC Anywhere, - from Symantec Corp. or Carbon Copy, from Microcom, and hardware products(multiplexers, bridges) can meet these needs but also require a high expense per connection.
Another approach is to use dial-up protocols that allow network packets to travel over serial communication lines. Examples include SLIP (Serial Line Internet Protocol) and PPP (Point-to- Point Protocol).
Both are commonly used today to connect users to Internet services that require network communications, but their response time is much slower due to the added overhead of the network packet activity.
In spite of these speed problems, look for further developments in the next year that will make SLIP and PPP connections more practical, including faster connections, better data compression and smarter software.
After nearly a decade of promises, the network age is truly upon us. Newspapers and magazines carry stories about the Internet, and users from all walks of life are subscribing to public and commercial network services for recreation or business purposes. Network client software has come of age, and it is finally possible to interchange components from several vendors to create a working system. There is no better time to discover the benefits of network services for your organization.
Copyright © 1995 Dennis S. Barnes
Reprints of this article are permitted without notification
if the source of the information is clearly identified