skip to main |
skip to sidebar
I had the distinct displelasure to have one of the hard drives in one of my fleet of MacBooks replaced today. This was a previous generation unit less than one year old-- one of the Penryn models immediately prior to the release of the current Unibody MacBooks. Just to be clear-- not an old POS by any stretch of the imaginaion. For regulatory reasons, I had to remove the hard drive and replace with a new one.
This operation took:
- Hiring an approved Apple Tech to do the swap, so that such an operation would not void the warranty.
- The removal of the battery, which I'll concede is a Very Good Idea in general when working on a notebook, but not required for most other brands/models.
- The removal of a body panel unrelated to the hard drive (Memory-- just to access some screws)
- The removal of 27 screws in total
- The use of two laptop-size screwdrivers-- a Philips and a Torx head (for no apparent reason.)
- The removal of one retention bracket holding one half of the hard drive.
At this point, the system was largely open, with the entire keyboard side of the bottom enclosure off. When reversing the process, you also have to take care to align tension clips in the case properly. Also, the thin aluminum shell of the pre-billet models is prone to minor dents which complicate servicing vs. purely cosmetic scratches in most other notebooks sustaining similar damage.That's not acceptable. Business class notebooks should release a hard drive tray with no more than 2 screws, and the tray itself typically mounts to the drive with 2 more. Pop, pop, in and out.Considering that drive removal/replacement is necessary and required for all sorts of good reasons, this design ethic is beyond reproach. I understand the "I want it to be pretty" BS, but in the real world, you can hit serious problems when a critical and basic task is denied to you.
It's no secret that flash storage is getting cheap and Linux can be tweaked to boot pretty fast. Put the two together, and you get implementations like Splashtop, most commonly commercialized as ASUS ExpressGate and Phoenix' HyperSpace close behind in the race to market. The combined cost is low enough to be able to justify for most systems beyond the budget end of the spectrum. The goal is to be able to get a very basic OS with several key apps up and running in five seconds or less. We'll revisit those apps later.
The whole idea that we can have even a basic OS up and running in 5 seconds is amazing enough, but the concept of using part of the computer without taking the time to boot into a full featured OS isn't entirely new. By the late 90's, several manufacturers offered rudimentary audio controls on the outside panel of laptops so that they could be used as extremely large and overpriced CD player. That's not so much of a real OS, but baby steps... By the Athlon XP era, some of the highest end motherboards from manufacturers like DFI included enough extra BIOS functionality to boot into MemTest86-- we're getting closer, and probably the first precursor on the Phoenix side. Baby step number two.
I point these two out because you see a fundamental difference between these two approaches. One is a pure media/entertainment option, while the other is about geting work done in a very specific way. Having a second OS to boot into to troubleshoot, test or repair is a common and critical need when doing various computer tasks.
Maybe this made more sense when a full OS boot took two minutes and various suspend options were half-baked. I won't say that there are never problems with suspend and hibernate, but modern OSes and hardware have come a long way. I can bring back an already booted OS fairly quickly.
The current implementations favor entertainment and communication apps. There are certainly times when having Firefox up and running quickly would help troubleshooting, but I suspect there's a better way to go than getting quick access to Skype and MP3s. Building a troubleshooting-oriented OS would help tremendously more than a media mini-OS.
Imagine having the ability to:
Work with hardware before a hard drive is installed or with a blank hard drive before the OS is installed.
Have a trustworthy OS to assist with data recovery when the installed OS isn't working properly.
Have an OS to boot into in order to perform diagnotics away from the installed OS.
Boot into a "safe" OS while attempting overclocks without the risk of damaging an installed OS.
A Media-centric OS gets you a browser to help look stuff up when troubleshooting, but not too much more than that. I respectfully submit that having a work-oriented OS is far more useful than a limited media OS.
As a supplement to the earlier post on Google's Servers & Datacenters, here's an astounding video.
It's just amazing how out-of-the-box the datacenter thinking is, and yet how scalable and reliable it is as well.
We see a lot of the software side of Google as the glamorous side-- Google Earth, GMail-- hell, even Blogger. It's good to shine some light on the awesome work the IT Infrastructure guys do.
This one's not an April Fool's joke. Google has finally presented their server design as well as their modular datacenter plans. It's quite a bit different than most large datacenter providers, and I think there are some really good lessons both on how to design a datacenter as well as the individual server units-- both in what to leave in and what to take out. The goal here was to help other companies with power saving initiatives, but we got a lot more information than just that.
I think the most important take-aways are:
Simplify your systems! The design cuts to the bare bones of what is needed. They're then able to expand with lots and lots of simple units. The designs look an awful lot like a motherboard on a tray that they were using in the beginning.
The nodes are really, truly generic. They've gone through 5 generations of these simple systems, and it doesn't matter what they do. Each one is a compute unit with no differentiation.
Since Google has the ability to build their own systems really and truly to spec (including entirely custom motherboards) they can do away with unimportant bits to save cost and add costs where necessary (e.g. the motherboad does the 5v step-down instead of the power supply-- a few dollar feature that is worthwhile for their specific design as it saves on power efficiency in the PSU, saving more than a buck or two in the long run.)
While Google is rumored to be not interested in repairing servers (they were famously supposed to leave broken servers in place and not bother fixing them), the most failure-prone stuff is very easy to service.
Batteries on each unit vs. a large UPS is a brilliant move. These guys won't last long enough for battery aging to be a big deal, and even if it is, they're small, discrete, and easy as hell to service compared to any UPS battery. Saving space, money, weight, etc. on UPSes really helps.
The whole containerized, customized datacenter design is so radical, I'm not really sure I fully appreciate it yet. It's got all the basics of a hot aisle system, but fantastic density. I'm going to have to think about this one it's so radical.
Anyway-- very cool stuff Google.
Posts
