It has been a while since I have visited the issues with my server. Well it has been a while since Silicon or Rosewill have contacted me. I don't know, maybe I was too hard on them, but the support conversation I was having with them just stopped one day with this:
That was the last I ever heard from them. I just didn't know how to respond, other than to explain to them that when I move 400MBs of files, it shouldn't take 1.5GB of RAM. Yeah, my answer isn't near as awesome as their assertion that programs use RAM. So I guess they win this battle. Silicon just has yet to respond at all, but I read a disclaimer on their site that they don't do end user support.
One peculiarity that I noticed was that when transferring files over the network, the RAM usage remains constant. The CPU usage goes up, understandably, it takes a lot of work to compute parity data, but the RAM holds steady. It seems there is a leak, or just a poorly programmed component of the driver, specifically in the DMA section. Good luck letting Rosewill know, they will probably respond that when programming, RAM and CPU usage goes up...
So instead of just sucking it up and running a RAID 1 or 0+1, I decided to do the unthinkable. As you can see above, I am using the stock Windows Dynamic Drive RAID 5.
What is interesting, is the windows setup suffers from the same strange RAM issues, which supports my theory of drivers and not hardware level problems. So why am I using it? Well, it doesn't exhibit the cyclic redundancy checks I would experience moving 300+ GBs around using the RAID cards setup. The true reason is that I was tired of waiting 17 hours every time I wanted to make another RAID 5.
So there you go. The server is set up. It is running, and currently serving up 800GB of data to my surrounding network. My disk usage surged 300GB as soon as I got the server set up but it has steadied as of late. I will post another entry in April when I get the other 6GB of RAM for this thing. What else should you do to combat a memory leak? Thanks for reading