Hadoop’s Sweet Spot

In my previous post, some took issue with my previous article exhibiting a negative bias toward Hadoop and here I’d like to correct that perception.

Over the decades, many computing ecosystems have come and gone.  In the evolution of data management, integration and analytic tools, the story is the same.

Big Data Delivery: Is the Advantage Technology or Methodology?

“All truths are easy to understand once they are discovered; the point is to discover them.”
– Galileo Galilei

Unless you’re someone who’s been in a biosphere or in some way completely isolated from the tech world the last few years, you are probably familiar with Hadoop and the excitement it’s created for being a high velocity, high volume data acquisition and advanced analytics ecosystem. There are numerous reasons given for organizations to move to Hadoop based solutions. Two of the more common benefits stated by advocates of Big Data ecosystems are:

1. Low cost commodity hardware and low (or no cost) software licensing.
2. Shorter delivery timelines.

Putting Together a High-Performance Macbook Pro on the Cheap*

Macbook Pro with Crucial 16-gig memory and Samsung 512-gig drive

SUMMARY: Purchase a base model Macbook Pro 15″ Laptop and add 16-gig of RAM and 512-gig SSD from 3rd party afterwards and save $861 over purchasing an a similar Apple configure-to-build system.

DETAILS: My early 2008 Macbook Pro has been a real workhorse over the past 4 years but with some of upcoming projects I needed a system with a little more oomph to run comfortably. It’s beat up pretty well from traveling literally all over the world with me and still runs well. My initial thought was to just add a solid state drive (SSD) to give it a bit more speed, but already being maxed out with 4-gig of RAM and needing more for some of the work coming up, I decided to get the new mid-2012 Macbook Pro with the maximum RAM and largest SSD available. A decent tax refund arrived to use on the purchase but after configuring the $1,799 Macbook Pro with 8-gig of RAM and the 512-gigabyte SSD the price tag came to $2,899 from the Apple Store — with tax around $3,133 … ouch.

Looking at the usual online places to purchase Apple products, for a fully configured system with maximum RAM and largest SSD from Apple the prices were not much better than purchasing directly from Apple. Poking around the interwebs researching it seems that the way to go is to buy a base mid-2012 model and upgrade the RAM and Disk on my own.

After looking around online, the components for the High-Performance Macbook Pro purchased are:

  • Base mid-2012 Macbook Pro 15″ from MacMall via a link from Appleinsider.Com Mac Price Guide for $1,674 shipped.
  • 16-gigabyte RAM module from Crucial.Com for $89 shipped.
  • Samsung 512-gigabyte Solid State Drive from Amazon.Com for $509 shipped.
  • $2,272 Total Price to the door, about $860 less than a similarly configured system from Apple out the door.

Macbook Pro with Crucial 16-gig memory and Samsung 512-gig drive

CHOOSING THE SOLID STATE DRIVE

There are a number of popular options when considering a Solid State Drive for the Macbook Pro at the time of this writing:

  • Crucial M4
  • Intel 520
  • OCZ Vertex
  • OWC Mercury
  • Samsung 830

After a lot of research, contemplating, consideration and a coin-toss or two the Samsung model 830 512-gigabyte Solid State Drive ended up winning out. The main criteria for consideration were:

  • Reliability: Many message boards contained posts regarding the various drives I was researching. The bottom line on this point is that it seemed that the Samsung was pretty reliable. A couple of the brands listed about seem to have a larger number of messages with people saying they had to send theirs back for another or that the SSD died after a week or two. Samsung and Intel seemed to have the least amount of complaints of dead-on-arrival or unit failures.
  • Compatibility: The Samsung 830 SSD disks use the Samsung PM830 controller while the other brands will use SandForce, Marvell or other controllers. The PM830 or variant is the controller in the SSDs in the Macbook Pros currently being shipped by Apple. I figure that of the 3rd party disks, the one with the controller closest to what Apple is using is likely to cause less trouble.
  • Price/Performance: The Samsung is neither the fastest nor the slowest in the various read and write benchmarks you will see at anandtech.com, tomshardware.com and other websites. It sits in the middle of the pack of most benchmarking lists. But it is neither the cheapest nor most expensive SSD either. The price performance is there and looking at the differences, even the benchmarking websites will say that the real-world differences will be negligible for most peoole.
  • TRIM Support: Most new SSD disks come with internal garbage collection to free up space from delete files when the disk goes idle, but TRIM commands issued by the Operating System are more efficient at making sure that deleted space gets freed up for use again. Research on the message boards showed no reports of people having difficulty enabling TRIM for the Samsung 830.

There are multiple package configurations for the Samsung 830:

  • Drive Only: MZ-7PC512B/WW: This is just the SSD drive. For a mid-2012 Macbook Pro this is the one you need.
  • Laptop Kit: MZ-7PC512N/AM: Has spacers to fit in laptop drive bay as well as a USB-SATA cable and software (for Windows) to transfer data.
  • Desktop Kit: MZ-7PC512D/AM: Has a 2.5″ to 3.5″ bracket so the drive can be installed in a desktop computer as well as a USB-SATA cable and software (for Windows) to transfer data.

It turns out that the drive-only will fit as-is into the mid-2012 Macbook Pro without the need for the spacer that comes with the laptop kit. The 7mm refers to the thickness of the drive, the drive is still the same length and width of 9mm drives so it lines up fine and fits well in the Macbook Pro drive bay.

Cloning the drive before installing the SSD

To get Mac OS X installed used a firewire->SATA adapter that was laying around to connect the SSD to the Mac and then Disk Utility to partition and format the drive. After the drive was formatted (GUID, HFS Journaled, case insensitive, encrypted) used Carbon Copy Cloner to clone the drive from the MacBook Pro Hard Disk to the Samsung Solid State Disk. Carbon Copy Cloner will give the option to install the Recovery Partition. Other software besides Carbon Copy Cloner like Super Duper or another Mac-based disk cloning software of your choice.

CHOOSING THE MEMORY

The mid-2012 Macbook Pro is capable of handling 16 gigabytes of memory, but Apple only provides a 4-gig and 8-gig option for the regular Macbook Pro. Interestingly the Macbook Pro with Retina Display can be configured for 8-gig or 16-gig of memory. If you want more that 8 gigabytes in your regular Macbook Pro, you must get the memory from a 3rd party. There are many places to purchase Mac memory. The two places considered for this build were:

Both Crucial and OWC have great reputations, guarantee compatibility, have a good return policy, and a limited lifetime warranty. Crucial was chosen for no other reason than it was a little bit cheaper and I had used Crucial memory in my early-2008 Macbook Pro with no problems.

INSTALLATION

Installation of the memory and disk was pretty easy. Be sure to wear a static wristband and make very sure to discharge any static so none of the components gets damaged. Using a Torx T6 and a small phillips head screw driver the following will get the memory and disk physically installed:

  • Flipped the Macbook Pro over,
  • Removed the 10 screws,
  • Popped the bottom off of the computer,
  • Unplugged the battery lead from the motherboard,
  • Pulled the 2 memory modules currently installed (4-gig RAM),
  • Installed the 2 Crucial memory modules (16-gig RAM),
  • Unscrewed the two screws keeping the disk in place,
  • Removed the current Hard Drive and Removed the 4 screws bolts in the HDD casing,
  • Put those 4 bolts back into the Samsung SSD,
  • Placed the Samsung SSD into the drive bay and screwed it back in place,
  • Replaced the battery lead to the motherboard,
  • Placed the bottom back onto the laptop chassis and screwed it in,
  • Turned the laptop back over and booted it up…

Inside of mid-2012 Macbook Pro 15"

PERFORMANCE

Captain Obvious statement here but this is by far the snappiest system I’ve had to date. The laptop boots in seconds and applications open in a blink.  When I compare my new Macbook Pro to my old one, some disk read/write numbers look like this:

  • Early-2008 Macbook Pro with 256gig 5400 rpm drive around 30 megabyte read/second and 30 write/second on an encrypted drive
  • Mid-2012 Macbook Pro with stock 512-gig 5400 rpm hard drive around 50 megabytes read/second and 50 megabytes write/second on an unencrypted drive
  • Mid-2012 Macbook Pro with upgraded 512-gig Samsung 830 SSD getting around 460 megabytes read/second and 380 megabytes write/second on a fully-encrypted drive (FileVault2).

Disk Speed Test, Samsung 830 512-gig SSD with full disk encryption (FileVault2)

CONCLUSION

I’ve been running this setup for a couple of weeks at the time of this writing and really enjoy the fact that this thing boots up fast and applications load in a blink. Some applications that were taking 10 or more Dock Icon bounces to startup (… Microsoft Office Products … ahem …) open in 1 bounce or less, basically in a blink. I think it will be very difficult for me to go back to regular hard disks after this.

The reason I got this system is that I will be running multiple virtual machines with database software in one and analytic software in another and wanted to make sure that disk contention was not going to be a problem. I got the boatload of RAM so I can run multiple virtual machines comfortably along with the Mac OS so that all systems have the memory they need to run optimally. This was almost impossible on my earlier mid-2008 system, both because of the disk and memory constraints.

I ended up getting the base model Macbook Pro, the 2.3ghz Core i7 rather than the 2.6ghz because I felt that RAM and disk were going to make more difference on performance for what I was doing than a modest CPU bump. Also the difference in the amount of GPU RAM (512 megabytes vs 1-gigabyte) doesn’t make a real difference for me as according to Menu Meters I’m usually using only 1/2 of the 512mb of video GPU RAM for the things that I do.

My plan is to keep this system for 4 years like I did the last. Hopefully when it gets to that point I can keep it running longer by replacing the current Samsung 512-gig drive with a 1 Terabyte SSD then. Seems my data needs keep doubling every 4 years so that should be just about right.

* – In the title I have an asterisk next to the word ‘Cheap.’ As we all know, cheap is relative when talking about Macs as a similarly configured windows or linux system would be much, much less. :-)

-James T Scoggins

Subscribe to RSS Feed Follow me on Twitter!