Like nearly all Linux based operating systems, I've found ClarkConnect Home Edition to be highly stable. In fact, I have yet to see any part of the system crash, hang up, or otherwise fail over the past few months of testing. But again, what sets it apart from other systems I've tried is the configuration interface. About the only thing really missing is an online help system, but of course, this is the free version, and I suppose the company has to leave some things out in order to convince you to purchase the real thing.
Linux is a very efficient system, which is why you can use it even on older computers and still get good performance. My test platforms for this review were a Pentium II 333 with 256MB of memory and a Pentium II 400 with 512MB of memory. Both systems were very responsive and handled all the chores I assigned them easily.
I ran a throughput test using netperf on the PII 333 system with 3 network cards running in DMZ mode. There was one system, a K6II 233 running Debian Linux connected directly to the CC box DMZ card thru a crossover cable and running the netserver daemon, which is the endpoint component to netperf.
The best results I was able to achieve were obtained by pointing four other systems at the DMZ box using the command "netperf -l 30 -f M -H 192.168.1.20" on all four boxes to blast four TCP stream tests at the CC system under test. The four systems were two Pentium 75's each running Debian, a Pentium 133 also running Debian, and a P4 2.6Ghz box running Gentoo Linux.
The throughput results were 2.29MBytes/sec, 1.80MBytes/sec, 2.26MBytes/sec, and 1.45MBytes/sec which is a total of 7.8MBytes/sec or 62.4Mbits/sec.
That's slower than I expected, but the real limiting factor in this test was probably the unswitched network hub that I had connecting the 4 machines running the test. During testing, the lights on the hub were pointing to the obvious bottleneck in this case, collisions. All network cards and the hub were 100baseTX, but I'm positive that a network switch, which pretty much eliminates collisions, would show much improved results. However, as I pointed out in my previous article, the practical limits of 100baseTX are typically given at somwhere between 60 and 90 percent of the theoretical limit of 100Mbits/sec, so this is within that range.