As a technology guy I’m always interested in how things work under the covers. I recently switched to a MacBook Pro because frankly I was tired of chasing the dragon with Linux on my workstation and I wanted something that Just Worked(tm). One of the very first things I did was get Apple’s Time Machine working. Like many others I decided to not go down the Time Capsule route, because I already have a very nice NAS (3x2T RAID-5 running Debian 6.0 and several virtual machines) and a very nice Wireless Access Point and frankly I don’t need to throw that all out just to get a Time Capsule . Using one of the many sites out there I was able to get netatalk running and all the magical Bonjour services advertised to get the MacBook Pro to believe my Linux box was in fact a valid Time Machine backup destination. Afterwards I started poking around inside Time Machine to see what made it tick.

Time Machine Internals:

Time Machine leverages a technology similar to snapshots. It keeps incremental backups of your filesystem inside a sparse filesystem image on an AFP share that the pretty 3D animated browser lets you navigate through to perform restores. If you mount the disk image outside of Time Machine you will see the various ‘snapshots’ contained in the root directory arranged by date and time of the backup. Underneath the covers the files are stored in 8 MB chunks which are created and deleted as space is required or freed. The ‘snapshots’ are file-based, so if you change a part of a file the entire file is re-transferred. The transfer is triggered by a change in the file’s mtime (UNIX for Modification TIME) attribute so if you were to say take an ISO image and use the UNIX ‘touch(1)‘ command to update the mtime to ‘now’ the file will be re-transferred next time Time Machine runs even though the entire file is identical to the already backed-up copy.

For the skeptical:

aramaki@15:11:03 ~/Downloads (535)$ md5 SW_DVD5_Office_Mac_Standard_2011_English_MLF_X16-99088.ISO
MD5 (SW_DVD5_Office_Mac_Standard_2011_English_MLF_X16-99088.ISO) = cd180db30b9851da6a55545457bc36aa

aramaki@15:11:28 ~/Downloads (537)$ md5 SW_DVD5_Office_Mac_Standard_2011_English_MLF_X16-99088.ISO
MD5 (SW_DVD5_Office_Mac_Standard_2011_English_MLF_X16-99088.ISO) = cd180db30b9851da6a55545457bc36aa


As you can see the mtime has been updated:

aramaki@15:14:14 ~/Downloads (541)$ stat -f "%m" SW_DVD5_Office_Mac_Standard_2011_English_MLF_X16-99088.ISO
1335366688

aramaki@15:14:25 ~/Downloads (542)$ date -r 1335366688
Wed Apr 25 15:11:28 UTC 2012


And there are no extended HFS+ attribs on the file:

aramaki@15:14:33 ~/Downloads (543)$ ls -lO SW_DVD5_Office_Mac_Standard_2011_English_MLF_X16-99088.ISO
-rw-rw-r-- 1 mernisse staff - 972087296 Apr 25 15:11 SW_DVD5_Office_Mac_Standard_2011_English_MLF_X16-99088.ISO


And yet, Time Machine is currently transferring it as if it was changed.

The other caveat to the architecture is that if you (like me) use File Vault 2 on Lion, the backup you make via Time Machine will be unencrypted. So you will want to take precautions to secure your network (wired and wireless) as well as physically secure the device upon which you store your Time Machine backups. In the case of using a Linux server to emulate a Time Capsule (as I am doing) eCryptFS or dm-crypt can provide security for the storage backing the Time Machine volume.

Disaster Recovery:

Time Machine provides a nice way to easily backup your Macintosh workstation. This is probably enough for most people but for the more paranoid of us the obvious flaw is that it provides no disaster recovery options. If your house burns down, or a power surge blows up your electronics you are completely sunk. This is where using a Linux server in lieu of a Time Capsule really shines.

Now because the disk image is comprised of 8MB chunks, you could simply use a file-based replication technology like rsync or duplicity to mirror the Time Machine destination volume offsite. The drawback of course being that your minimum transfer is going to be 8MB/hr and will likely include a LOT of duplicate data.

The alternative is to utilize something like NetApp‘s OSSV technology to make de-duplicated, block-level transfers of the Time Machine volume offsite. This would be much more efficient over the WAN as it will overcome the file-based snapshot technology limitation of the underlying Time Machine engine and will provide a much higher storage efficiency in the destination volume.

Since I already use OSSV to protect my virtual machine infrastructure here at home and in the colocation facility that I host my public-facing IT gear at, I intend (when I find the free time…) to test using OSSV to protect my Time Machine volume and get some real world numbers to show the bandwidth and storage reduction involved. My impression after looking under the covers a bit is that it will likely be significant.

Summary:

Time Machine is a good technology. It truly makes backups easy for the general computing public and I highly recommend it to any Macintosh user (though I imagine, given what I know about the slightly-fanatical nature of the Macintosh user base I don’t have to). I wouldn’t be surprised if we see Apple offering an iCloud bolt-on service in the near future that will allow you to replicate all, or part of your Time Capsule protected data off to some centralized Apple cloud storage environment (hopefully on NetApp storage) but until that happens, and for those of us who rolled our own Time Machine backup destinations it seems to make sense to roll our own DR protection methods.

I have a NetApp FAS2020 which I am currently using to re-arrange my VM server’s RAID array and have been syncing data to it using rsync(8) to a NFSv4 share on the filer. One of the interesting things is that NFSv4 no longer uses numeric UIDs/GIDs and instead transmits the textual strings and relies on the host’s getpwnam() functions to translate it into a numeric value for the underlying file system. Thankfully Data ONTAP has the ability to use LDAP in a similar fashion to modern Linux hosts[1].

I am running Data ONTAP 7.3.6 and my LDAP configuration looks like this:

toaster*> options ldap
ldap.ADdomain
ldap.base dc=ub3rgeek,dc=net
ldap.base.group ou=groups,dc=ub3rgeek,dc=net
ldap.base.netgroup
ldap.base.passwd ou=users,dc=ub3rgeek,dc=net
ldap.enable on
ldap.minimum_bind_level anonymous
ldap.name xxxxxx
ldap.nssmap.attribute.gecos gecos
ldap.nssmap.attribute.gidNumber gidNumber
ldap.nssmap.attribute.groupname cn
ldap.nssmap.attribute.homeDirectory homeDirectory
ldap.nssmap.attribute.loginShell loginShell
ldap.nssmap.attribute.memberNisNetgroup memberNisNetgroup
ldap.nssmap.attribute.memberUid memberUid
ldap.nssmap.attribute.netgroupname cn
ldap.nssmap.attribute.nisNetgroupTriple nisNetgroupTriple
ldap.nssmap.attribute.uid uid
ldap.nssmap.attribute.uidNumber uidNumber
ldap.nssmap.attribute.userPassword userPassword
ldap.nssmap.objectClass.nisNetgroup nisNetgroup
ldap.nssmap.objectClass.posixAccount posixAccount
ldap.nssmap.objectClass.posixGroup posixGroup
ldap.passwd ******
ldap.port 389
ldap.servers ldap.internal.ub3rgeek.net
ldap.servers.preferred
ldap.ssl.enable off
ldap.timeout 20
ldap.usermap.attribute.unixaccount unixaccount
ldap.usermap.attribute.windowsaccount windowsaccount
ldap.usermap.base
ldap.usermap.enable on


This basically mirrors my nss_ldap configuration and has allowed me to easily copy my RAID array over to the filer so I can wipe and rebuild the underlying RAID layer.

NFSv4 Reference:
RFC-3530
NetApp TR 3085

[1] http://media.netapp.com/documents/tr-3464.pdf

The other big change has been my further move towards mobile computing. I am actually writing this on my BlackBerry PlayBook while enjoying the delightful weather and watching people come home after a long day celebrating my heritage ;) My fixed computing has been relegated to recreational tasks and only the most serious work related functions. I think more so than any advance since the Internet itself, mobile computing technology is going to drive the future of communications and human interaction. The cloud model may be a big buzzword in marketing these days but for me it is nothing more than an optimization of the already decades old client server model, whereas the platform and location agnostic nature of mobile computing is truly a revolutionary movement.

I still recall how liberating it was the first time I got a BlackBerry smartphone and was able to be effective at working not only while out of the office, but on the other side of the planet. The decoupling of location from the ability to do ones job is worth so much more, not only to businesses but to the individuals who get to enjoy life unchained from a desk.

Now if only I had the freedom to fly more… That would be something.

The last several lessons however are leading up to my first solo which is arguably the most important mile stone in a pilot’s training for any class of certificate. As such the concentration has been on me executing take off, traffic pattern and landing successfully and if I’m honest I’m really starting to struggle and get frustrated by it. I don’t know what the root cause is at the moment, part of it is trouble maintaining alignment with the center-line of the runway, which sounds easy and trivial (especially with a 32′ wide plane on a 60′ wide runway that is almost 3/4 of a mile long) but I can assure you it is not. So it kinda works out that I get overwhelemed, frustrated, and end up over-controlling the plane…

I can drive my car with two fingers, or a thumb simply hooked on the side of the steering wheel and maintain position within a hand-full of inches at 65MPH with ease. I can do the same in a boat, but put me on final towards a runway and I’m ham-fisting the aircraft, complete with sweaty palms and 180bpm heart-rate. Time, which is supposed to be a constant seems to accelerate and all the tasks I need to complete blur together… did I make the radio call for the turn to final, am I high, flaps, air-speed, throttle, descent rate… whereas I can drive a car at over a hundred miles per hour, in traffic, with barely more than a resting heart rate, doing 80MPH in this plane is a whole different story.

And I think that’s part of the frustration. I’m competent in activities that seem similar at least when measured by speed or requirements of precision, but this is still difficult. I am sure that time and practice will bring the confidence and calm that I’m used to feeling in other activities to this and I tihnk admitting this to myself is certainly going to help but at least for the time being… it’s frustrating and difficult.

And deep down inside, that’s part of why it is so great. Things worth doing shouldn’t be too easy.

A brief summary of my configuration:
Lenovo Thinkpad T500
Intel Core 2 Duo P8700 @ 2.53GHz
8GB 1066MHz DDR SDRAM
Intel 5100AGN Wireless
Intel GMA x4500 video (ATI graphics disabled for battery life issues)
WD 160G 5400RPM drive, with dm-crypt /home partition

To start with, I upgraded my 10.10 installation by letting update-manager -d do its thing. The process was mostly painless (a couple of customized config files needed manual intervention but that’s to be expected) and took a little over an hour to complete. Upon reboot the system was up and running without any issues which is a testament to the continued work of the Ubuntu developer community. It continues to astound me how trouble-free the various betas have been at least from the dist-upgrade front.

I’ve been using the new Unity interface for a day or two now and while I’m happy that I don’t hate it as much as I did in the alphas.

The Good:
It is fast and stable at the moment. I think compiz is a little more power-hungry on my laptop than metacity was, but not enough that it is setting my lap on fire. At least from an upgrade perspective I have not found anything that doesn’t work that used to (I maintain a local apt repository of various changes and customizations for my systems, nothing that I had working in 10.10 broke in 11.04 beta). The work on the indicators has been very good for the most part, the Network Manager applet is quite a bit better than the previous versions.

The Bad:
Compared to the new Gnome 3 interface Unity seems a little clunky, probably due to the netbook heratige of the design though I am glad that I still have the minimize buttons. The launcher/dock/whatever thing doesn’t always hide, so it sits there taking up space on screen for no good reason and you can’t move it from the left hand side which is not where I’d put it if I had the option. The Apple-like menu bar is a little irritating, but really only because I think it is highly derivative and I am sad that we couldn’t come up with something better from a design standpoint. I also find the dash to be annoying to use but I don’t really know why. Finally I’m not sure why the new date/time indicator had to lose the weather. I travel a lot and had many different locations defined in the previous applet and having an at-a-glance view of the weather was a nice feature to have.

Conclusion:
I was really worried that I’d either have to skip this release or continue using the Gnome 2 panel interface after looking at the first couple of alphas but overall am pretty happy. I will try Gnome 3 when (if?) it comes to Ubuntu and see how it stacks up. I think it’s a prettier and cleaner interface but only time will tell which I like more. The amount of work that the Ubuntu community put into this release is pretty amazing and the fact that it landed in as good of shape as it did is a really huge accomplishment.

Obligatory screenshot:
Ubuntu 11.04 Beta 1

Update, 18/April/2011 It would be really nice if they would fix this bug before shipping 11.04.

I enjoy international travel. I think that the more of this world that we can see the better off we are as people and the better we come to understand ourselves. This is certainly not my first time out of the country nor even off the continent but I admit this trip has a special place in my heart because I expect the sheer weight of the culture shock to be unlike anything I’ve ever felt before. I have been to the UK and Australia and as a native of a former British colony it really wasn’t all that different. Yes I had to drive on the other side of the car and road, yes the slang was different, and yes the money was funny colors and had pictures of the Monarchy on it but for all intents in purposes we all shared a common cultural liniage. Thursday evening (Japan Time) when I step out of the airplane I will be for the first time completely out of my element and I am excited beyond words to experience that for the first time.

Surfing Safely, from somewhere in the Universe,
Matt

As a primarily hobbiest developer I found the process to be fairly straightforward, though unfortunatly extremely Windows-centric. The primary debug workflow involves Eclipse and the BlackBerry device simulators, the latter being Windows-only. Trying to debug without the sim is a non-starter as the JavaScript engine will not throw any useful error messages while running live on a device. It is not wholly awful but I really do prefer avoiding Windows if I can reasonably manage it.

There are obvious Enterprise edges sticking out through the process though happily the new, fluffy, comsumer oriented image that RIM is putting on has really improved the developer start up process. Code signing keys no longer cost money, and getting them is basically a process of signing up in a web form and waiting a few days for the files to be delivered to your Inbox.

I ended up writing a small application that takes some of the backend data that powers my personal Status Board (similar to the Panic Status Board) and displays it on the phone along with some other useful data that I can get out of the RIM APIs.

I originally was just using jQuery to fetch the JSON from my status board but was somewhat unsatisfied by the performance of pulling the data over the macro network and loading it into the page every time the application loaded so I replaced that with an on-device cache and push update system that takes advantage of the BlackBerry Enterprise Server I have (RIM offers a free push service over BIS as well but I like the idea of using the BES’s encrypted transport) hosting my e-mail.

The push server side of the equation is somewhat poorly documented, mostly through snippits of Java and .NET code that other people have implemented and some PDFs from RIM describing the format. After some trial and error I was able to pull together a workable Python update script that fetches the JSON from my status board and pushes it to the handheld. This allows fast, bandwidth friendly, scheduled, asynchronous updates and improves the user experience quite a bit. I totally understand why RIM is pimping their push infrastructure as being a vital part of what they are calling “Super Apps”. Having the data come to the user versus making them constantly update it is kinda awesome.

I doubt this application will be useful for anyone else but in the odd case that it might be the code is available in my bzr repository, which you can view at: http://repo.ub3rgeek.net/branches/blackberry/files, the status-app directory contains the handheld application source and the push-server/status-push.py file contains the push application.

Over the next several years I endeavored to have more and more computers running for various reasons. Back before laptops became accessable for the average person I solved the portability problem by having computers all over the house so that wherever I went I had a terminal on the network that could access whatever it was I wanted. I think back in the day before AIM allowed multiple sessions I had 5 or 6 AIM names which I left logged in at each workstation. It sure made finding me, even at home, a fun logistical problem.

Sometime around 2003 I received surplus from a company that was moving an enclosed network equipment rack. Part of the problem of having something like this is the desire to stuff more and more computers into it until it is full to bursting.

Part of this desire is augmented by the fact that I tend to do various complex things with computers and sometimes you just need a number of them to get these jobs done. Thankfully computers have been getting better and better and you can cram a lot more things into them. Also since I started working with technology for a living I have started to get more and more interested in being a user while I’m at home as opposed to being a Systems Engineer 24×7. So I guess the answer to the question in this post’s title is this: I am both growing up and at the same time the faster better stronger computers are making it easier and easier to hide from the rest of the world my raging computer problem.

To help illustrate that a quick comparason: in 1998 my very first Linux server was a 486 DX4-100 with about 320MB of hard drive space, and a whopping 48MB of RAM. In 2011 my sole home server is a dual-core Athlon II X250 with 4TB (6TB raw) of disk and 2GB of RAM.

So, for your concideration, the last few years in pictures of my home network, ‘internal.ub3rgeek.net’:

2011

2010

Late 2007

Early 2007

2005

I managed to get away from the AOPA Summit only spending a few hundred dollars (resisting the urge to buy a Bose A20, somehow). I took in two extremely good and interesting panels, one on flying in and around Mexico and the other on FedEx Flight 705 by the co-pilot who was on board that day. It was a harrowing tale and I think I will always remember it. It is another affirmation that the idea of becoming a pilot is a dream that I cannot wait to finally fulfill.

It is always a good feeling getting away from home and spending time in different places. I don’t think that I shall ever lose the joy of traveling, but I am certainly going to try as hard as I can to wear myself out.

Now that I am back home the weather here has of course turned cold and I am having to fight the urge to just curl up and hide from it. I am hoping that the weather stays good enough so that I can continue my flight training.

I have logged 11 hours of dual received so far and nearly 50,000 miles aboard United in the last two years, over 25,000 this year, qualifying me for Elite status. All in all a good time.

Next year my two closest friends will be moving into his first house, and getting married respectively. If all goes well I will be visiting another friend in Japan and of course I will have to find time to fit in another trek to Australia to see my dear sister and brother-in-law, Pennsic, and whatever else comes across my plate.

Surfing Safety, traveling lightly.

I didn’t get to go to Oshkosh this year but I am making up for it by attending The AOPA Summit in Long Beach, CA in a couple weeks. In the mean time I am hoping the weather will cooperate and I will end up getting 2 or 3 more hours of flight time in before I leave. Some take-off and landing work would be excellent…