Disruption Tolerant Networking
Last night I saw a rather interesting presentation at the local IEEE ComSec chapter by a fellow from JPL. His presentation centered around Disruption Tolerant Networking. DTN is primarily being developed with networks running in space in mind, but there are some more earthly applications as well.
The basic concept is that TCP/IP, along with the rest of a standard modern networking stack, expects all nodes to be available all the time, with a relatively large quantity of bandwidth. In space, this isn't always the case. For example, look at the Mars Rovers. We communicate with them by sending commands on the ground, which go through an Earth-orbiting satellite, through space to a Mars-orbiting satellite, and finally down to the rovers on the Martian surface. What makes this connection tricky is that nothing's stationary - the satellites are orbiting the planets, the rovers are moving, and the planets are orbiting the sun. Line-of-sight is required for communication, but not always available. A standard TCP/IP stack would require the entire path to be complete, while DTN does not due to a store and forward type structure, where packets (called "bundles") are stored in local storage on the satellite until the link is available.
The video below is only 4.5 minutes long, but explains the process very thoroughly. Check it out, and be sure to switch to HD if possible:
Activation, Drivers, and DRM, Oh My!
Apparently I'm not too great at staying on-topic. However, this post feels like it has merit.
Basically, the events of the past weekend have left me missing my Mac. It hasn't gone anywhere, but after dealing with Windows for a weekend, I'm about to give up and just use my Mac.
This whole crazy debacle began about a year ago. I bought a motherboard and CPU from Fry's because the motherboard was basically free. I decided that I'd just upgrade it later, and in the meantime have a computer to use and some cash to spend elsewhere.
The time finally came where I decided that 2 Gb of RAM wasn't enough, so I went to Fry's and picked out another motherboard, an MSI P45 Neo3. I know, Fry's isn't the cheapest, but Newegg was only $15-20 cheaper and I wanted something to do this weekend. (Little did I know what was to come.) So, out I walked on Friday with it and an additional 4 Gb of RAM. Not only will Flight Simulator love the new RAM, but so will Vista in general, as well as the virtualization experimenting I'll be doing soon with Xen.
So, I set to work. An hour or two later, the board was installed, and I booted it up. Windows blue-screened. This was expected, as Windows does not like its hardware changed like that. My Gentoo partition wasn't thrilled either, so I set to work with Windows first, knowing that I was basically going to have to reinstall.
So I do my reinstall using my legally-obtained Vista Ultimate Upgrade, courtesy of the Ultimate Steal promo that Microsoft is running. Attempting to activate failed, reminding me that I purchased the upgrade and not a full version. Ok, so I perform an "upgrade" from Vista Ultimate to...Vista Ultimate. Activation works then. Why, Microsoft, must I jump through this hoop? It's rather ridiculous. A pirate wouldn't have to deal with this, but I gave you my money, so now I must suffer.
Next up is drivers. I proceed to use the drivers on the CD to at least get me to a network connection, and then I use the ones off the MSI site. Since then, I don't think I've shut off the machine in a normal way - I get tired of rebooting from hard lockups and eventually just power off instead of reboot.
Hmm, how could I possibly fix this? I bet a Windows Update will solve my problems! So, updates install, and the system reboots. Black screen. That's odd... reboot...last known good configuration...black screen. This is getting ridiculous.
Should I mention that, in there, I attempted to install Flight Simulator and was greeted by an invitation to call Microsoft and sit on hold for 10 minutes so I could activate my legally-purchased software? A pirate would have been up and running long ago.
Should I also mention that I emerged KDE with Gentoo earlier, and while the compile didn't finish due to my lack of knowledge, everything compiled without strange lockups? On a hot day, no less? This tends to rule out hardware failure, though I'll run some tests later this week to be sure. The temperatures never go into the 50s, however, so I'm pretty sure cooling is not an issue.
All things considered, even if the Mac does cost a little bit more for similar hardware to the PC world, my time is worth more. I'd rather go learn how to fly a real plane anyway.
New Year, New Blog
Just as I promised, with the new year starts a new semester, and so will my thesis project. From here on out, I'll be attempting to focus on it during my entries, and point out interesting things that I find.
Today's entry, however, is not so good. Remember that idea I had a few months ago about controlling Xen machines from a control panel? Yeah, it's been done. Over and over and over. I discovered this after I got approval for my idea, leaving me with a problem - do I keep going, knowing that I'll be duplicating the work? Do I abandon the idea of writing my own and attempt to patch (or even fork) that other little project known as Ganeti? (After all, I do have a few ideas to make it even better which they don't appear to have implemented.)
Nobody said this project was going to be easy, and I just hit frustration #1 - dealing with how I thought I had a great idea, while it's actually been done. I'm sure I'll learn quite a bit in the process if I continue, but should I give up and change topics? Do I accept the fact that the CS world is filled with really smart people (such as those at Google who came up with Ganeti) and continue on anyway? After all, not every single MSCS student in the country can come up with something new.
Is the process about learning new things for oneself, or generating new things for the industry in general by building on existing research?
Exchange 2007 Sucks
I write this post as a response to a friend's Facebook wall posting. I started to write a response to him, and ended up writing this much-longer-than-allowed entry instead.
My employer has moved to Exchange 2007 twice. The first time, it was completely unusable for Mail.app users, which also happen to be 100% of our development team. Of course, that's not going to fly. We moved back to 2003, time passed, patches were distributed, and it became time to move again. We're still on it after this current move.
So, why does Exchange 2007 suck? Oh how I can count the ways.
- The CLI (PowerShell) sucks. "Get-ExchangeMailbox -identity blah ..." is incredibly, unnecessarily verbose. It's like it tries to get GUI people to realize the power of a CLI, making it needlessly complex and wordy. This could be made better if tab completion gave you a list of options, like in Bash, but instead it picks the first option. Therefore, you have a 1/n chance that it'll actually be the item you want, less so if the first command (alphabetically speaking) is one you use less often. You might as well not implement it at that point. Also, I wouldn't have a problem with the CLI if the GUI wasn't half-assed - IMAP and POP3 GUIs didn't even show up until SP1.
- IMAP is a second class citizen. This goes for POP3 as well, though we don't have many (any?) users on it at this point. Mail.app used to speed along just fine with Exchange 2003, and now it's a dog on 2007. Granted, I have thousands of email messages in my Inbox, but I never had a problem before. Please fix this, so I don't have to wait 5 minutes in the morning just to get my email.
- Creating mailboxes is slower than before. I used to be able to create mailboxes when I created the user account. This is no longer the case, requiring me to log in to another application after the user's been created just to create a mailbox for them. Why?!
- Performance in general sucks. I am 100% certain that I could serve the same number of mailboxes on a PII or PIII server running Linux and get better performance than the quad core, 4G RAM, 10K in RAID 1 server we bought specifically for this purpose. Logging in to Server 2008 is incredibly laggy, waiting for the Admin Console to open up is even slower. Pathetic. Did I mention this is on Gigabit ethernet?
- Backups? Nah. Their "backup solution" sucks. On the aforementioned Linux box, I could rsync the backups to another folder, or another server, onsite or off. Cost: $0. Exchange 2007, however, is no longer compatible with the old ntbackup system. You need to get (read: purchase) Microsoft System Center Data Protection Manager 2007, a product with a horribly long name and dumb limitations. I'll admit, the shadow copy system is really nice; I like knowing that I could restore down to within an hour of a problem. Want to do offsite storage? HAHA, I don't think so, not without a tape drive. Yes, we use disks since we're small enough that they're economical. Also, we can't send a tape over the VPN to a different part of the country, can we? Nope. Of course, you could buy Cristalink's Firestreamer product to simulate a drive and use a file as a "tape" - but that'll cost you another $250.
- SSL? Don't need that either! I don't care if SMTP sends messages in plain text - I want to secure the connections between our clients and our servers with SSL to protect passwords! What happened to this? It was in 2003...
- Added 2009-01-08: Just TRY and get your data out. So Microsoft is really pushing the 64-bit platform, releasing 32-bit binaries only for testing purposes, not as a supported production system. That's fine. However, why do I have to use a 32-bit computer with Outlook (another product that requires purchasing, now that it's not part of the CAL) just to export the data? I want my data out of your product without hoops or additional costs. Too much to ask? Sure, I could jury-rig something up to hit the POP3 or IMAP connector, but that seems a tad inefficient, don't you think?
So, as it should be plainly obvious by now, Exchange 2007, despite it's new bells and whistles (which, incidentally, cost more money), sucks at its core functionality - being a mail server.