After my recent switch to OS X, I fell in love with TimeMachine and the transparent backups it provides.  I even have it backing up to my Windows 7 file server wirelessly so I literally do nothing when I am at home to get seamless unattended backups (not even plugging in an external HDD).

The one issue I have found in this setup is that my VirtualBox disks (Ubuntu Linux and Windows 7) get copied completely after each backup.  This is less than ideal.  3.5GB per backup adds up very quickly!  Here I describe how I created a ‘base’ (immutable) virtual box disk and let VirtualBox create an additional differencing disk on top of the base disk.  After this, TimeMachine is only copying very small chunks of data rather than the whole 3.5GB disk!

I found quite a few posts online related to this topic, including this forum post, this forum post/tutorial about VirtualBox VDIs, this forum post, and this StackOverflow topic.  Each of these contain good information.  However, this forum post was pretty much exactly what I was looking for.

To summarize the process, we first must create the base VDI disk and mark it as immutable.  This forces VirtualBox to create a second differencing image of changes beyond the immutable base disk image.  The second step is to change the mode of the differencing disk, toggling autoreset from ON to OFF.  By default the differencing disk only applies for one run of the VM.  In otherwords, any changes across restarts will be lost.  Turning autoreset off fixes this.

Here are the steps I’ve extracted from that discussion, including commands that need to be run:

1. Create an immutable .VDI disk.
   1. Detach the base VDI from the virtual machine in the VM settings.
   2. Make the VDI immutable by running the following command:

   VBoxManage openmedium disk -type immutable DISK_FILE_NAME.vdi

   3. Reattach the now immutable VDI to the VM.
2. Tell VirtualBox to use a differencing disk explicitly. Click the check-box that says ‘Differencing Disk’ in the VBox settings for the VDI that is immutable.  Note the GUID of the differencing disk that gets created.
3. Disable autoreset for the differencing disk:
   1. By default VBox will reset the state of the machine between restarts.
   2. Disable this behavior by toggling the autoreset setting:

   VBoxManage modifyhd --autoreset off DIFFERENCING_DISK_GUID

After running through these steps you should be able to start your VM, make some changes, and restart.  The changes on disk that you made should be persistent.  It would be nice if VBox would make this particular use case all GUI-based, but at least it’s possible.  Regardless, now TimeMachine only copies the differencing disk which is significantly smaller than the base disk image.  Please let me know if this works for you in the comments!

Update (2010-5-10 21:47): If you make significant changes beyond the base disk image, it is possible for the differencing disk to grow beyond the size of the base image itself.  In this case, it may be worth it to collapse or merge the changes into the base disk image and start the differencing disk fresh.  To do this, you make a clone of the differencing disk image, use the clone as the new base, and use the process above to start fresh with a new differencing disk.  I used this forum post as a basis for this process.  The clone command is:

VBoxManage clonehd DIFFERENCING_DISK_GUID NewBaseDisk.vdi

Distributed Lookup Tables – Alan LaMielle – NCAR SIParCS 2009 from Alan LaMielle on Vimeo.

This is a talk I gave that summarizes my work over the summer of 2009 in the SIParCS program at NCAR in Boulder, Colorado. I looked at reducing the non-scalable memory footprint of a set of lookup tables in an atmospheric model within CCSM—the community climate system model. Our approach was to utilize a feature of MPI 2 called one-sided MPI by distributing the lookup tables across MPI tasks. One-sided MPI allows one to remotely access memory of other MPI tasks.

• 1 large can brooks (or kunners) chili beans
• 1 large can tomato sauce
• 2 cans dark red kidney beans (add one for more beans)
• 1 can (15 oz) diced tomatoes
• 1 lb hamburger
• onions
• chili powder
• tobasco sauce
• other spices to taste

Brown hamburger, and onions.  Combine all ingredients in CrockPot and cook on low for 6-8 hours.

After a few days of debating whether or not to buy a Nexus One, I finally gave in a bought one.  Rachael ended up getting a Droid Eris (Verizon) for Christmas, so in the course of a few weeks I gained access to two fantastic Android-powered phones.  I had been compiling a list of mobile application ideas and the fact that I now have the hardware to test on motivated me to start looking into Android development.

After getting the Android SDK setup I needed to connect my N1 and push an application to it for testing on real hardware.  I found some Android documentation that discussed connecting up an Android device over USB.  However, after following the steps on this page, I was still seeing strange output when running adb devices:
List of devices attached
????????????    no permissions
This led to some searching which turned up a blog post.  Basically, the Google Android team hasn’t added the Vendor ID for the Nexus One.  Apparently HTC’s USB Vendor ID isn’t correct.  So, I followed these steps to fix it:

1. Create/edit a udev rules file:
   sudo vim /etc/udev/rules.d/51-android.rules
2. Add the following line to this file:
   SUBSYSTEM=="usb", SYSFS{idVendor}=="18D1", MODE="0666"
   (Note the vendor ID of 18D1. This was changed from the HTC vendor code of 0BB4.)
3. Restart udev using either
   sudo reload udev
   or
   sudo service udev reload
4. Connect your Nexus One.
5. Run
   adb devices
   and you should see something like
   List of devices attached
############ device

Hopefully Google updates the documentation to include the Vendor ID 18D1 rather than making us hunt for this number ourselves.  Enjoy!

A few weeks ago a friend of mine gave me her dying iPod Classic.  She noted that it had started acting erratic after being dropped one too many times.  The device would skip playing parts of songs and lock up periodically.  She noted that since I’m usually one to tinker with and hack on devices, she figured I might find a good use for it–more so than she had for it at least.  Well, I’ve definitely found a good use for it by swapping out the original hard drive for flash-based solution.  The device works (nearly) like new.  Here’s how I did it.

I started the process just looking for a replacement disk.  I found various sites that sold them, but they all ranged from $120+.  I wasn’t looking to put too much money into this project.  Also, I didn’t like the idea of having a moving-parts iPod anyway since the whole reason I obtained the device in the first place was because a non-solid-state disk had been dropped one too many times while in operation.  For these reasons, I began searching for a solid-sate solution that would allow me to revive my broken iPod.

Some searching yielded the site of an interesting adapter that converts the ZIF interface of the original hard disk to one that accepts Compact Flash cards.  Additionally, the adapter was designed for exactly this purpose: converting a hard disk iPod to a solid-state device.  This was exactly what I was looking for!  After spending ~$25 on the adapter, and picking up a 32GB Kingston CF card from amazon for ~$80, I had located all the hardware I would need for the project.

Before ordering the adapter and CF card, I did want to make sure I could take the iPod apart without ruining it.  I found a few good tutorials, with the best being one at iFixIt.com.  Also, I discovered other posts about doing exactly what I wanted to do.  This site is a bit more hardcore than I was looking for (he was hand soldering connections, but the upgrades are for much older generation iPods).  I used ideas for taking apart the iPod from this post (rather than specialized tools, I used a razor blade and a few very small flat-head screw driver bits).  Using all of this, I successfully managed to take the iPod apart (without breaking anything!) and was confident I could use the hardware to complete the upgrade, so I placed the orders.

iPod, dead hard drive, ZIF to CF adapter, and 32GB CF card

Dead (dying) hard drive

iPod Classic

ZIF to CF adapter made by Tarkan Akdam (2)

ZIF to CF adapter made by Tarkan Akdam (1)

Once both the adapter and CF card had arrived, actually doing the upgrade was pretty simple.  The trickiest part of the process was actually taking apart the iPod.  The very small ribbon cables are probably the most delicate parts and getting the out of their respective connectors and putting them back in required a little patience and gentle-handedness.

Once I had inserted the CF card into the ZIF to CF adapter, I placed the adapter into the iPod housing, hooked up the various ribbon cables, and snapped the two halves of the case back together.  I connected the device up to iTunes and, after reformatting the new ~30GB volume, I had a working solid state iPod Classic!

After loading up the iPod with music and playing with it for a few weeks, I only have a few complaints.  First, the device still locks up periodically when I’m using it.  This is mostly when I first start playing a new set of songs.  I suspect that this is due to the slower CF card that I picked up.  Second, the CF card is only 32GB, so I can’t fit all of my music on it.  However, I think in time CF card prices will come down and I’ll be able to swap the current card out with a 64GB or 128GB card.  Other than these two complaints, I’m very happy with this hack.  I ended up with a 32GB flash-based iPod classic for ~$100.  Check out the full set of pictures below.

Ingredients (Sweet potatoes):

• 3 cups mashed sweet potatoes (~1 large 2lb 8oz can ‘Bruce’s Yams’)
• 1 tsp vanilla
• 2 eggs
• 1/2 cup sweetened condensed milk
• 1/3 stick butter

Ingredients (Topping):

• 1 cup ground corn flakes (or frosted flakes)
• 1/2 cup chopped walnuts
• 3/4 stick butter
• 1 cup brown sugar

Directions:

1. Mash sweet potatoes together with vanilla, eggs, sweetened condensed milk, and butter (small bits of unmelted butter will be present after mashing)
2. In a separate bowl, combine topping butter with ground corn flakes, chopped walnuts, and brown sugar (make sure mixture is mostly uniform, no large chunks of butter should remain)
3. Grease baking dish
4. Spread mashed sweet potato mixture into dish
5. Spread topping mixture over sweet potatoes evenly
6. Bake at 350 degrees for 30 minutes

The sweetness of this dish almost puts it in the desert category.  It’s very tasty–a wonderful thanksgiving companion to more savory dishes.

• 1 box family size brownie mix
• 1/4 cup flour
• 1/4 cup vegetable oil
• 3 tbsp. water
• 2 eggs
• 1 cup chocolate chips (or peanut butter chips or butterscotch chips)
• 1/2 of a jar of peanut butter (~16 oz)

Directions:

1. Combine brownie mix, flour, vegetable oil, water, eggs, and chocolate chips
2. Heat peanut butter in a small bowl for 30-60 seconds (until it can be poured out of the bowl)
3. Thoroughly mix 3/4 of the peanut butter in with brownie mix
4. Gently fold remaining 1/4 of peanut butter into mix (don’t mix completely, leave peanut butter visible)
5. Spoon dollops of mix onto cookie sheets (non-stick foil preferred), note cookies don’t spread out much during baking
6. Bake at 350 degrees for 5-10 minutes, depending on consistency of mix

Be careful to watch the cookies as they bake so you avoid burning the bottoms.  Don’t let them cook until they look like they are done.  Taking them out when they still look uncooked (but slightly crusted on the top) is fine.  The cookies will sit up and harden when cooling.  They are a bit different from traditional cookies as they are based on brownie mix.

Ingredients:

• 1 package rolled up croissant pastry dough
• 1 can fruit filling (apple, cherry, blueberry, etc.)
• Melted butter
• Large granule sugar

Directions:

1. Roll out croissant dough triangles
2. At the large end of the triangle, put a small amount of filling
3. Roll dough with filling up, starting from large end of the triangle
4. Brush butter over top of turnover
5. Sprinkle sugar over top of turnovers
6. Bake at 350 degrees for ~20 mins, or until pasty dough is golden brown

We’ve found that adding 5-6 cherries (with very little cherry sauce) is the right amount of filling.  Also, about 2-3 apple slices seems to work well.  Also, brushing the butter over top of the turnovers helps to brown the dough a bit.  The sugar makes them a bit sweeter and look pretty.  Also, as is probably true with most pastries, it’s better to let them cool on a cooling rack so the bottoms don’t get soggy (see pictures).

Following the instructions on this site, I created a new file (/etc/hal/fdi/policy/mouse-wheel.fdi) as root that contains:

<match key="info.product" string="TPPS/2 IBM TrackPoint">
<merge key="input.x11_options.EmulateWheel" type="string">true</merge>
<merge key="input.x11_options.EmulateWheelButton" type="string">2</merge>
<merge key="input.x11_options.XAxisMapping" type="string">6 7</merge>
<merge key="input.x11_options.YAxisMapping" type="string">4 5</merge>
<merge key="input.x11_options.ZAxisMapping" type="string">4 5</merge>
<merge key="input.x11_options.Emulate3Buttons" type="string">true</merge>
</match>

If you don’t want to restart, you can log out and restart the gdm and hal services.  However, I simply restarted and scrolling worked upon logging in!

The document here is apparently the original source of this information.  I’m mostly posting this for myself for future reference (and in case either of the afore mentioned pages go down).