SEM test bed tribulations

analysis
Sep 23, 20052 mins

When the best-laid test plans go awry, it makes for some interesting reconfigurations

We certainly didn’t make things easy for the vendors in this SEM test. By the time we were finished, our test infrastructure had been replaced, our test equipment was showing signs of failure, and our client devices were calling in sick — the ones that would communicate at all.

Fortunately, we had resources to request emergency shipments of new network infrastructure (thank you, Extreme Networks), and we quickly replaced the old switches that went toes up with Extreme Summit 400-48 boxes.

Because all of the vendors had set up their products for the network we’d planned — rather than the network we ended up with — we got to see how each of them could incorporate new items into a network. We specified some unusual items, such as a firewall none of them had seen (the Ingate 1400 SIP-capable firewall), so we also got to watch them figure out how to go from raw logs to operation.

And just when we seemed to have things on track, our previously rock-solid Spirent WebAvalanche started having heat problems. This may have been due to a few too many round-trips between Oliver Rist’s lab in New York and the Advanced Network Computing Laboratory at the University of Hawaii. In retrospect, maybe we should have bought a hardened shipping case for it after all.

Still, we were able to get our simulated enterprise up and running, even if it was somewhat different from what we’d originally planned. For every test, we loaded the required agent software on the requisite operating system using a ClearCube blade system, with blades running Windows Server 2003, Windows XP, and Suse and Debian Linux. We also created standard images of each machine so we could be sure we were starting from the same base configuration. All of this was monitored by the open source Snort listening through a Gigamon Systems multiplexed network tap.

When we were satisfied that everything was working, we ran a series of standard test scripts to generate events for the SEMs. Because the University of Hawaii seems to be a favorite target for China-based hackers, we also had a constant level of real intrusion attempts for the Snort IDS to deal with.