Network Computingwww.networkcomputing.com

NetCracker Professional version 2.0 takes network design to a new level with its hassle-free setup, its ease of use and, most important, its ability to simulate a virtual network environment that provides valuable statistical feedback. While NetCracker lacks some important features, such as network autodiscovery and the ability to import data, it complements existing network analysis tools.

All It's Cracked Up to Be? Installation was a snap, and NetCracker claimed no more than 100 MB of my hard drive. The vendor recommended a 200-MHz system with 64 MB, but a Pentium II running at 333 MHz performed much better. The software called for Windows NT 4.x or Windows95/98. I installed NetCracker under both NT 4.0 and Windows98 and experienced no problems with either OS.

To create my network, I dragged and dropped a device from NetCracker's impressively large (more than 5,000) supported-device list. I would have liked to import basic MIB information, such as chassis and port configurations, from a network management file to populate our virtual network. Unfortunately, I had to add each router, switch and hub manually. Realistically, without the aid of an autodiscovery or an import option, re-creating a large or even medium-sized network would involve a lot of legwork. NetCracker says it plans to incorporate autodiscovery functionality in a future release.

Fortunately, adding network devices in the lab was simple. To build a network layout, I dragged, dropped and linked devices from the left device window. I was amazed at the level of detail in the devices on the list; a double-click on the icon brought up the device's property information, which accurately depicted such data as port type, port availability and the interface card of the corresponding ports. Database updates appear as posted, downloadable files on the vendor's home page. Setting up my test network was almost foolproof with the friendly reminders that jumped to the screen. For example, one told me that I unintentionally tried connecting devices that didn't share a common media type.

Simulations With High Expectations NetCracker supports the animation of a variety of network configurations, including client/ server, VLANs (virtual LANs), intranets, wireless, multimedia and switched-circuit. Network traffic can be generated using data, voice or video streams. So how is it possible for NetCracker to accurately run a network simulation? Each network device provides performance options for bandwidth, latency, filter speeds and forward speeds that you can configure to mimic your network. To arrange the different protocol configurations, I used the check boxes under the supported bridging, network and routing protocol sections.

In terms of tuning performance specs, the task of setting up servers and workstations was equally detailed. By dragging and dropping from different manufacturer lists of cards, I populated my servers and workstations with a combination of NICs. After adding each NIC, I set bandwidth and duplex options. I also created traffic between clients and servers by establishing a traffic profile, which specified what type of operations the system would handle. I was able to take a server and assign a corresponding profile, such as a database, FTP or e-mail server. I also allocated profiles to workstations that correlated to one of the server's profiles.

In addition, I assigned multiple profiles to multiple servers and workstations, establishing multiple packet streams between the two machines. To simplify packet tracing, NetCracker color-codes packet types based on the different kinds of traffic, and users can change the colors if they desire. It also correlates packets with different shapes depending on whether the packets are ATM cells, LAN packets, WAN packets, physical-layer data or dial-up sessions. NetCracker-connected segments also are color-coded based on media type: Coax is black, twisted pair is blue, fiber optic is yellow, multicable wire is red and radio waves are green.

After I set up a few simple network layouts, running the simulation was reminiscent of running an actual network. The packets flowed at varying rates over the different segments depending on designated bandwidths. To locate stress points within the network, I set up utilization bars next to the switches and routers, which displayed current utilization of the devices. I had the option to add other statistics, such as packets dropped and workload, to the devices' visual statistics.

To validate how realistic the simulation was, I switched TCP/IP routing from OSPF to RIP and back again, and was delighted with how the packet flows changed dynamically. Having first set routing to RIP, I was easily able to trace packet routes and determine which routers and switches were being stressed by NetCracker's indicator bars. After I rerouted to OSPF, the stress equalized itself among the different routers, as I expected. Additionally, by increasing the latency and load in different switches, routers and media segments, I determined just how well the simulation would react. I was again pleased to find what I had anticipated: more dropped packets and collisions in hubs and switches upstream.

With NetCracker, you can also produce reports. In the lab, this feature was easy to use, especially with precanned reports or with the help of the report wizard. Vital device statistics, such as utilization, workload, packets dropped and packets received, were among the options I could add to the reports. Additionally, NetCracker can export reports to HTML.

Send your comments on this article to Dave Fetters at dfetters@nwc.com.