Shunra Software Ltd. has made Version 1.1 of its Storm hardware and software WAN emulation tool significantly easier to use. However, nagging questions about Storm?s price and flexibility could make Shunra?s software-only Cloud tool a better choice for most organizations.
Storm, which shipped last month, is a capable tool for IT managers who need to test the performance of IP-based applications before they are deployed over bandwidth-constrained WAN links. But even at its lowest-priced configuration?$12,000 for the two-port version?Storm is an expensive way to simulate WAN circuits. The four-port version is $25,000, and the eight-port edition costs $32,000.
In comparison the Adtech SX-series data-link simulators from Spirent Communications (a division of Spirent plc.) are more capable hardware-only tools and only slightly more expensive than Storm.
Any of these tools need at least some other products to simulate network traffic and to record performance results. We used NetIQ Corp.?s Chariot software and WildPackets Inc.?s EtherPeek to generate a measurable and repeatable amount of network traffic. We used both products and Fluke Networks? OptiView protocol analyzer to measure results and compare them with statistics provided by Storm.
Although our tests stuck to bread-and-butter data traffic, tools such as Storm can also be used to see how videoconferencing or voice-over-IP applications might fare.
Test before you leap
Storm tests applications before they are deployed. This is a central IT best practice because understanding bandwidth needs and application performance prior to deployment is among the best ways to ensure there are no expensive surprises after an application is rolled out.
During tests, it was a snap to install the new Storm software on a Windows NT Server. (The software also runs on Windows 98 and Windows 2000.)
As in Storm 1.0, the previous version, we used a simplified Visio tool pallete to quickly lay out endpoints (usually a collection of PCs found in branch offices), gateways and WAN connections. Endpoints are connected via gateways (routers) using a WAN connection. We could define the characteristics of these WAN connections and thus simulate a real network.
By adjusting variables such as capacity, latency, congestion, and dropped and duplicated packets, we were able to measure the effect of the network on IP applications. This is an area where Spirent?s Adtech simulators have a real leg up on Shunra?s tool, however, because Adtech boxes can be configured with a variety of data channel interfaces, ranging from RS-232C to E-3, T-3 and OC-3. Storm boxes can?t be similarly configured.
Storm uses only standard Ethernet connections to support the physical network. During tests, we put one group of endpoints on a switch and connected that switch to one of the ports on the Storm box. Another group of endpoints was similarly connected to a second port; we then conducted tests by simulating the physical WAN between the groups of endpoints. Despite this limitation, the test results were useful for getting ballpark performance numbers.