Technical innovation in the enterprise usually means finding the best product to solve a particular problem. But individual software packages and standalone hardware components rarely provide an all-in-one solution. That's why Windows NT Magazine is spotlighting several organizations that combined in-house development, third-party products, and of course, Windows NT to produce unique and innovative solutions to technical problems in the enterprise.
For the second annual NT Innovators awards, the editorial staff of Windows NT Magazine asked readers to share their innovative and unique approaches to using NT. We accepted nominations on our Web site and selected the most exciting users of NT and third-party products in 1997.
Setup Manager, Sysdiff, and the Super Duper Retro Fixer
Imagine that you have 1200 workstations and servers to upgrade and customize for various job functions. You're upgrading not only the workstation hardware and the operating system, but also all the applications running on these workstations and servers. To complicate matters further, these 1200 workstations and servers are in 75 locations throughout the state—and some of the locations are extremely remote. Now, do the entire upgrade in a few months.
If this scenario sounds like one of those impossible case studies from Microsoft's Networking Essentials certification exam, think again. It's precisely the challenge that was facing systems consultant Hank Oen and the rest of the networking staff at APS, a power company in Arizona.
In 1997, APS simultaneously upgraded its network from Windows NT 3.51 to NT 4.0, distributed fixes, upgraded from MSMail to Exchange and from Office 4.3 to Office 97, and deployed CA Unicenter TNG to get more control over technical resources. Using Microsoft tools, third-party tools, and innovative homegrown network-testing solutions, APS's networking staff accomplished the upgrades with little disruption to the network users. The staff developed stable platforms of operating systems and applications, rolled out the platforms, and issued upgrades and patches with as little user involvement as possible.
Two resource kit tools helped the networking staff create the installation scripts to perform the upgrades: Setup Manager and Sysdiff. With Setup Manager, the networking staff created answer files to perform the unattended installations. With Sysdiff, they implemented a standard desktop across all machines.
The Sysdiff process involved three stages. In the first stage, the networking staff prepared the base installation and took a snapshot of it. In the second stage, they installed all the standard applications (including an email client and a telephone book) and made some changes to the Registry. Sysdiff then compared this setup with the snapshot to create a file noting the differences. In the last stage, Sysdiff used the differences file to reproduce the setup on the other machines.
Messing up 20 user configurations with a buggy installation is a mishap; messing up 900 user configurations is a disaster. So to avoid such a disaster, the next step for the networking staff was to test the 40+ user-installable business-unit application installations and software upgrades. To test the configurations, they used imagine LAN's ConfigSafe and Chicago-Soft's DLLaGator. These tools let staff members fine-tune the way they loaded the applications to achieve better application stability.
ConfigSafe protects a computer from buggy applications. In a version-controlled test environment, the network staff used ConfigSafe to compare the system before and after application software installation so that they could see exactly what the software was doing to the system, including what files the software installed and where it installed them.
DLLaGator's function is more specific: It creates a relational database of all file and module names that access a specific DLL. With this tool in the test environment, the networking staff determined which DLLs the applications shared and which DLL version the applications loaded; that way, they could make sure they used the most recent DLL version.
No matter how much testing occurs, an installation rolled out to nearly a thousand clients is bound to require some fixes. At first, as users filed bug reports, the networking staff created the patches and applied them to subsequent builds. After about the seventh iteration, however, the staff rolled all the patches into one executive file (internally known as the Super Duper Retro Fixer) and added the file to the user logon scripts so that the workstations prompted users to apply the patch when they logged on. The staff ran into some minor problems with the Super Duper Retro Fixer when a couple of Windows 95 machines became confused at logon and tried to apply the NT patches, which corrupted their Registries. But the problems weren't serious, and the staff quickly corrected them.
APS's installation innovation didn't end with the use of NT tools to create, fine-tune, and repair unattended installations. To make the applications and network as solid as possible, the networking staff created application simulators and network-testing devices to test network traffic patterns, packet load, and full-load database transaction response times. When you're running a WAN that encompasses most of the state and includes microwave relays, frame relays, and fractional T1 links, you must test the network thoroughly.
The upgrades aren't over yet. Oen expects APS to implement more changes in 1998, including the addition of new customer information systems, human resource systems, and financial systems.
Hank Oen · 602-250-2001
Email: [email protected]
imagine LAN · 603-889-3883 or 800-372-9776
Chicago-Soft · 603-643-4002 Web: http://www.quickref.com
Wireless Cellular Digital Packet Radio, ATM, and MUGSHOT
Crime doesn't wait. Officers need as much information as possible as soon as possible," says Officer Ted Cormier of the Charlotte-Mecklenburg Police Department (CMPD). That philosophy is the basis of the innovative Windows NT-based communications system that the CMPD is implementing in North Carolina. With this system, officers will have realtime access to the information they need, not only to punish crime, but to prevent it.
According to Officer Cormier, the problem with the existing system is as follows: Imagine that an officer in the northern part of Charlotte stops and contacts an individual standing behind a closed business. The officer records the contact on a written field interview card and turns in the card at the end of the shift. The police department might not enter the information on that card into its files for days or even weeks. Worse yet, if someone misplaces the information, it might not be entered at all. Now imagine that 20 minutes after the first officer contacted the individual, another officer stops the same individual for the same activity. The second officer doesn't know about the first incident, and thus cannot establish a pattern.
Cormier envisions the new system changing the way the CMPD does business. With the new system (due for pilot testing this month), a WAN will connect the CMPD offices to each other and connect patrol cars to a wireless Cellular Digital Packet Data (CDPD) radio network. Officers will be able to enter data into and retrieve data from an Oracle database as necessary, whether via a standard LAN connection in the main office or a branch office, a car-mounted modem, or through a standard dial-up link. Thus, the information available to police officers can remain current. Using this system, the first officer in the example will be able to immediately record the contact. That way, when the second officer encounters the same individual 20 minutes later, the second officer will be able to search the database for any history of similar behavior, notice a pattern, and inspect the individual's actions more closely.
Although the CMPD uses an existing LAN to help its officers stay in touch, the new communication system will be farther-reaching and faster. The old network consisted of mainframes distributed among the CMPD offices connected by a 19.2Kbps circuit. Officers in the field relied on a radio network of mobile data terminals with a connection running at 4800bps. With the new network, branch offices will connect to centralized NT Server machines in the main office that provide email, file and print services, remote access services, and access to legacy applications. The branch offices will connect to the main office via asynchronous transfer mode (ATM—running at 155Mbps), so centralizing all the servers for ease of server administration won't be penalized by slow connections.
The CMPD anticipates that the higher wireless WAN speeds will also let officers in the field use the MUGSHOT system to download pictures of individuals and help the officers make identifications. MUGSHOT has been available to desktop users, but the slow radio link made it impractical for officers to use remotely.
Because the information the police department handles is extremely sensitive, NT's security is fundamental to the success of the new communications network. Three-hundred new workstations and 1100 new laptops running NT Workstation will require users to log on. This added security will help ensure that users can't bypass the password protection by hitting the power switch or disable system passwords by removing a jumper from the motherboard inside the computer.
What about supporting this network? The CMPD can't afford downtime (the database must remain available 24 hours a day, 7 days a week) and can't afford to hire staff to monitor the network all the time. Therefore, the CMPD adopted a troubleshooting support mechanism that offers two levels of support to keep the network going. At the first level are the troubleshooters, who are sworn and non-sworn personnel trained to be the first line of support for all computer problems, whether involving hardware or software. The CMPD maintains a staff of 108 troubleshooters, so that one troubleshooter per shift per district is always on duty. At the second level, if a problem turns out to be more than a troubleshooter can handle, the troubleshooter calls in a trainer for help.
With the use of an NT-based WAN, officers can access the CMPD databases from their cars, homes while off-duty, and anywhere they can get a connection. Officers out with a group of people, a community leader, or even with individuals involved in criminal activity will be able to connect to the database to get the information they need when they need it. As Cormier says, "What once took weeks and an, 'I'll get back to you,' will be accessible immediately through a secure NT network." Although the CMPD is the first police department in the US to implement this high-speed and comprehensive communications network, Cormier is sure that other police departments will follow suit in the near future.
Charlotte-Mecklenburg Police Department|
Ted Cormier · 704-525-6145
Email: [email protected]
Intelligent Call Routing, Interactive Voice Response, Automatic Call Distribution
Suppose you want to create a financial customer service network that extends across most of a continent. No matter which region or country customers call from, they need their call answered as quickly and simply as possible. Now assume you want to include features that remember who the customers are and what they wanted the last time they called, so that they talk to the customer service agent who can most likely help with their questions.
Beginning in 1995, Citicorp, a large financial services company, began implementing such an innovative solution, first for its credit card division and later extending this solution to the American Airlines Advantage Bankcard. Citicorp needed to implement a client-server network that extends throughout the US, Canada, and the Caribbean. Mark Ambrose, Citicorp's executive director of global fulfillment strategy and security, explains the goal as creating a customer experience of "one click, one call, one mile," so customers dialing in can get information as easily as possible. He wants people to be able to call in when they need to, rather than waiting for a good time to carry out a tedious and difficult chore.
To make this intelligent call system work, Citicorp will use GeoTel's Intelligent CallRouter software with Melita International's call management system PhoneFrame CS for outbound calls, running on top of Windows NT Server. Intelligent call routers (ICRs) in Texas and California contain the information that the call system needs to route customer calls to the right customer service agent. The system determines where to route the calls according to expected wait time and agent skills.
An Automatic Call Distributor (ACD) determines which ICR server is best equipped to take the call. Although the call routing is currently based on a database of call history tied to the customer's name, it will soon tie into Interactive Voice Response (IVR) platform voice recognition software that can determine who the caller is and, based on information logged during previous calls, what the caller is likely to be interested in. After the customer elects to speak to someone, the call goes to an ICR server that routes it to the customer service agent best able to handle the call based on the information the ICR server receives from the ACD. To ensure that the ICR servers in Texas and California can communicate quickly, they interconnect with a dedicated T1 line. Other server communications don't have quite the same urgency, so the rest of the servers are part of a frame relay network, with a 64Kbps circuit connecting each ICR server. Local servers connect via a 10Base-T network.
Why centrally locate the ICR servers? Ease of administration is key to Ambrose's plan of creating a personalized customer service experience, and Citicorp can best accomplish this ease of administration by keeping the ICR servers at two sites. As Ambrose points out, the larger the peer-to-peer network, the more difficult it is to implement, because you must connect each site to all other sites. The multiple-master domain model that the Citicorp call center network uses requires fewer links and simplifies administration.
Citicorp depends heavily on NT's enterprise-level scalability. Although the Citicorp call-management network doesn't have many servers (about 50 spread across 9 locations, including ICR servers, peripheral gateway servers, and administrative servers), it has a lot of call volume to contend with: about 15 million calls per month, and that traffic is just voice calls. When the system opens up to Internet traffic as Ambrose plans to implement during 1998 with voice-over IP, the traffic levels will be even higher. This call management network also needs to be fast, so Citicorp can minimize the amount of time that callers have to spend waiting. Database transactions that determine call routing have to be quick.
All in all, Ambrose is excited about the way the GeoTel call-routing software and NT Server are working together to make a seamless, uniform, customer service system, and Ambrose is excited about the future additions that these technologies will make possible.
Mark Ambrose · 212-599-4193
Email: [email protected]
GeoTel · 978-275-5100
Melita International · 770-239-4000 or 800-635-4821
Network Streamlining, Application Integration, and the Patient Web
Hospital-based network administrators will appreciate the innovations that Ed Bianco, chief information officer, implemented at Lowell General Hospital in Massachusetts. For a long time, Lowell General's network was a hodgepodge of NetWare, UNIX, and VAX-based LANs. But the need for interoperability; the need to support 32-bit, 16-bit, and legacy UNIX applications; and the need for TCP/IP support on the local network helped Bianco realize that a change was necessary. He decided to migrate the network to Windows NT. "VMS is going away, IPX isn't a viable protocol for what we're doing, and UNIX is losing market share," said Bianco. "NT isn't the end-all, but it's getting there." Not all workstations on the network run NT; some lower-end systems are still using Windows for Workgroups 3.11 and Windows 95. However, anything mission-critical (such as a nurse's or physician's workstation) is using NT 4.0.
When you look at the complexity of the network, you can easily understand why Bianco chose a single-vendor solution. Lowell General is a 208-bed facility occupying 7 buildings on a medical campus. These 7 buildings house the network's 500 users. Redundant 20-strand fiber and Cabletron MMAC-Plus switches connect the network, which runs at 100Mbps.
Lowell General's large network needs to work as seamlessly as possible, which isn't an easy task given that the hospital uses a variety of software to automate many hospital functions, such as accounting, radiology, managed care, and billing. As a result, the hospital's platform has to support as many of those packages as possible. Although not all the hospital's software runs on NT yet, Bianco and his staff are migrating from the VAX/VMS and UNIX applications as quickly as possible.
Bianco and his staff used their innovativeness not only to streamline the network and integrate old applications, but to develop a new application for the network. They created the Patient Web, a secure Web-based clinical repository.
Many hospitals store data in various locations, so the staff can't retrieve data from one source. But with Patient Web, Lowell General's staff can retrieve all patient data from one data warehouse. For example, a nurse or physician needing a patient's file can query the SQL Server database from any building (or even from any state through the Internet). The server calls up the file and accesses all the data stored for that patient. A nurse can view the most recent information, and physicians can place orders for tests. To eliminate errors, the hospital verifies orders against the hospital SQL rules database before processing the orders. (In time, a nurse or physician might also be able to view a patient's x-rays or other stored images.) The network accesses the database in realtime, so the patient's file always accurately reflects the patient's status. This innovative front end for medical records makes obtaining and updating patient information much easier and more accurate than conventional record-keeping practices.
The hospital-based intranet isn't just for PatientWeb, either. Web-based terminal emulators provide access to VAX and UNIX systems. Online documents available from the intranet include the hospital and departmental policy and procedures manuals, physician credentials, hospital reports and statistics, a staff roster with home contact information, and descriptions of more than 5000 drugs. Practitioners can find just about anything they need at one location, rather than having to hunt down each piece of information. And, of course, putting all this information online makes it easier to keep current and accurate.
Nurses and other users like the new NT system for another reason: ease of use. Many users have worked on PCs before and were already familiar with the applications they were going to use, so they just needed to brush up on their skills. Also, going with NT means that applications are always available and that new ones will continue to emerge. If a vendor has a new NT product, Lowell General is the first to sign up.
Lowell General Hospital|
Ed Bianco · 508-937-6034
Email: [email protected]
Automated Inventory Check-In, Replicated and Mirrored Server Sets, Wireless Data Entry
Although running a major shipping company with a system of handwritten receipts is possible, you certainly wouldn't want to do so any longer than was necessary. Sea-Land Services (a CSX Corporation subsidiary that is headquartered in North Carolina, with ports in the US, Europe, South America, Australia, and Asia), used to organize the comings and goings of its 200,000 containers in 40 ports with a receipt system. However, after introducing an innovative Windows NT-based check-in and inventory system, the company has reduced the amount of time required to process check-ins, and has improved the accuracy of the check-in process. Although integrating NT 3.51 and Windows 95 into the check-in process is still underway, the innovative use of NT products and wireless networking has already shown its worth.
The check-in process can be automated, by means of a terminal automation system (TAS) that Sea-Land Services began implementing in the mid-1990s. When trucks enter the yard of a dock, they stop at a gatehouse and push a call button to be queued for inspection. When the truck comes to the head of the queue, the gate clerk inspects the container and enters the registration information into a SQL Server 6.5 database. Then, the gate clerk prints a bar-coded ticket for the trucker that tells the trucker whether to drop off a container (the boxes that go on the backs of trucks), pick one up, or both. When the trucker leaves with a container, the system outprocesses the trucker, who receives a receipt. High-volume ports can get about 2000 trucks in every day, so this automation is necessary to keep things moving. It also simplifies the process of reporting any damage to the containers and then billing for this damage.
The SQL Server databases come into play during inventory. Sea-Land Services stores containers in the yard at the docks. Making sure that the workers pick up the right container for shipment is important so that the widgets intended for Amsterdam don't go to Seattle. Therefore, an accurate inventory report is vital. However, because a yard can be more than 200 acres in size, doing the inventory by hand was both inaccurate and time-consuming. The best solution was a radio-frequency wireless network that lets staff doing inventory enter their data directly into the database, while they're still in the yard. This long-distance inventory (Windows for WorkgroupsWFWon the client side machines that plug into NT Server-based database servers) improves the database's accuracy and reduces the chance of time-consuming or dangerous inaccuracies, such as truckers picking up the wrong container.
Because downtime is very expensive, Sea-Land Services uses SQL scripts to replicate the Compaq 1500 servers that hold the database to keep a hot server on hand in case of malfunction: If one of the four servers goes down, another server can pick up the slack. Sea-Land Services doesn't rely only on replication, either: The company uses hardware RAID to mirror the SCSI server disks to keep them up and running.
Sea-Land Services currently uses servers running NT Server 3.51 and NT Server 4.0 only to store the local database information, using Sybase's MicroDecisionware (MDI) gateway to obtain information about the whereabouts of containers from the global mainframe network. However, Mark Grant and Mike Julian, Sea-Land Services LAN engineers, say that although some Sea-Land Services applications will continue to rely on the mainframe, eventually most of the global network will be NT-based. Presently, parts of the company are using legacy applications, but Sea-Land Services will phase out many of these applications and replace them with the check-in and inventory system at each port run.
Mark Grant · 904-359-3254
Email: [email protected]
Sybase · 800-879-2273