Facebook Friends and Frenemies Report - Why We Add and Remove

There's a new Facebook research report from Nielsen McKinsey's NM Incite titled Friends & Frenemies: Why We Add and Remove Facebook Friends that's pretty interesting. The report looks at the factors that help Facebook users decide whether they want to add someone as a friend or remove an existing person the their friend list. Here's a few details:

• Knowing someone in real life is the top reason cited for friend-ing someone (82%)
• Offensive comments are the main reason someone gets the boot (55%)

Additional report details suggests that real world interactions drive online friendships. Meanwhile, sales-oriented and depressing comments help drive friend removals. Facebook etiquette also plays a role, with updating too often, too little or having too many friends a consideration for some Facebook users.

Regarding gender, the report research indicates that men are more likely to use social media for careers/networking and dating – while women use social media for a creative outlet, to get coupons/promos or to give positive feedback. More men add friends based on business networks or physical attractiveness and women are more likely to friend based on knowing someone in real life or remove them due to offensive comments.

Here's an interesting infographic from the report.

What's a T3 Line?

In my last post I described what a T1, also called a DS-1, line was. Most of us have also heard the "T3 Line" term used. Let's take a look at what a T3 or DS-3 line is.

DS-2 Signal
Before we can describe a DS-3 line, let's first take a look at a DS-2. In that last post we figured out how each DS-1 signal (T1 line or circuit) carries a bit rate of 1.544 Mbps. Four 1.544 Mbps digital DS-1 signals are multiplexed into one DS-2 signal. If we have 4 DS-1 signals per DS-2 signal and each DS-1 signal is 1.544 Mbps we can calculate:

Adding overhead consisting of timing and synchronization bits brings the DS-2 bit rate to 6.312 Mbps.

DS-2 Formation

DS-3 Signal

Each DS-2 signal carries a bit rate of 6.312 Mbps. Seven 6.312 Mbps digital DS-2 signals are multiplexed into one DS-3 signal. If we have 7 DS-2 signals per DS-3 signal and each DS-2 signal is 6.312 Mbps we can calculate:

Adding overhead consisting of timing and synchronization bits brings the DS-3 bit rate to 44.736 Mbps.

And..... 44.736 Mbps.... that's a T-3 line!

T1 Lines - What They Are

Most of us have heard about "T1" lines. We know they are some kind of (expensive) communications line you can get from one of the telephone companies. It turns out T1's are part of the Digital Signal (DS) Level System. 

Back in August, I wrote a post titled More on CODECs: Quantization + Sampling Rate = A PCM Wave. In that post I described how a piece of an analog signal is quantized and companded and then given an 8 bit binary code in a process referred to as encoding. From that post, we know to convert an analog signal to a digital signal the analog signal is sampled 8000 times per second and, after matching the instantaneous voltage sample level to one of 256 discrete levels, an 8 bit code is generated for each sample. If we multiply the sample rate by the bit code we get:

(8000 samples/second)(8 bits/sample) = 64,000 bits per second (bps)

So we can say a single analog voice channel, after conversion from analog to digital, requires 64Kbps of digital bandwidth. This 64Kbps is referred to as Digital Signal Level 0 (DS-0) and is the basic building block or channel for the existing digitally multiplexed T carrier system in the United States and the digital E carrier system used in Europe. 

Voice calls are digitally multiplexed using either time division multiplexing or statistical time division multiplexing. Calls are grouped in a way similar to frequency division multiplexing. Let’s look at how this is done.

Digroups or DS-1 signals

Individual analog voice call channels converted to digital and require a bit rate of 64 Kbps each. 24 64 Kbps digital voice channels are multiplexed into digroups or DS-1 signals. If we have 24 DS-0 signals per DS-1 signal and each channel is 64 Kbps we can calculate:

Adding overhead consisting of timing and synchronization bits brings the DS-1 bit rate to 1.544 Mbps - that's a T1!

DS-1 Formation

DS-1 Overhead

We’ve described the process of encoding where an analog signal is sampled 8000 times per second, quantized into one of 256 discrete signal levels, companded it is then given an 8 bit binary code. After a single analog signal sample has been encoded it is multiplexed, with 24 other encoded 8 bit sample signals. This generates a 192 bit (8 bits/sample signal × 24 sample signals) sequence for the 24 sample signals. A process called framing then adds one framing bit to create a 193 bit frame.

DS-1 With Overhead

The framing bits are used to keep the receiving device in synch with the frames it is receiving. Every twelve frames are grouped into a masterframe, also referred to as a superframe. Included within each masterframe is a twelve bit frame pattern from the 12 grouped 193 bit frames This twelve bit frame pattern carries a bit pattern of 000110111001 and repeats itself with each masterframe.

Masterframe

This masterframe bit pattern is used for synchronization.

Remember each channel is sampled 8000 times per second so a single frame represents one eight-thousandth of 24 individual channels or telephone calls. We can also say that, in one second a DS-1 signal transmits 8000 193 bit frames. We can use these numbers to calculate the true DS-1 bit rate which includes both data and overhead (framing) bits:

Each DS-1 signal carries a bit rate of 1.544 Mbps and.... that's a T1!

Carrier IQ - are You Being Tracked?

Last month, security researcher Trevor Eckhart published a report accusing CarrierIQ of installing malware on more than 140 million devices worldwide. Eckhart also published a video showing CIQ's software secretly running in the background and monitoring a variety of handset activity on an HTC device including key presses, browsing history, SMS logs, and location data. If you have not seen it, here's Part 2 of Trevor's video: 



Yesterday Senator Al Franken from Minnesota "reached out" to AT&T, HTC, Samsung, and Sprint Nextel after they acknowledged their use of Carrier IQ’s diagnostic software to request that they explain (within the next 12 days) what they do with the information they receive from the software.
Also yesterday, Carrier IQ released a statement saying:

We measure and summarize performance of the device to assist Operators in delivering better service. While a few individuals have identified that there is a great deal of information available to the Carrier IQ software inside the handset, our software does not record, store or transmit the contents of SMS messages, email, photographs, audio or video. For example, we understand whether an SMS was sent accurately, but do not record or transmit the content of the SMS. We know which applications are draining your battery, but do not capture the screen.

In addition, the following updates have been posted by The Huffington Post:

Grant Paul, a well-known iPhone hacker who goes by the screenname "chpwn",wrote on his blog that Apple has included Carrier IQ on the iPhone, but the software's default is disabled.  

Want to find out if your phone is secretly tracking you? Check out our comprehensive list of the devices and carriers known to use Carrier IQ.

Who And Why I Follow Back on Twitter

Catching up on some work this morning and going through new people that started following me over the past two weeks.  I've got my account setup so I get email notification when someone follows me and I look at each one, determining whether I want to follow back. Out of the 302 new followers I picked up in the past couple of weeks, I followed back only 27 this morning. That's only 8.95% and it is pretty typical.

Here's how I personally sort this stuff out:

When someone follows my feed I've got Twitter setup to send me an email notification.
I've got my email client (Thunderbird) setup to automatically move those Twitter email notifications to a separate Twitter folder. When I have some time (like this morning) I go through the notifications, determining whether I want to follow back. Here's my follow-back determination procedure:

1. I've got Thunderbird setup to preview email. The first thing I look for is a name (a person) attached to the account. If I don't know your organization and there is no name listed, I'm probably not going to follow back. Some details:

• I try and only follow back those with similar interests, these interests can be both work and hobby related. If you are a business, organization, academic institution or individual involved in Science, Technology Engineering or Math (STEM) I'm definitely following you back. I'll also follow you back if you are focused on one of my hobbies - for example - saltwater flyfishing. 
• Sorry but religion and politics are always a do not follow back red flag for me. I know many use Twitter and other social media for this kind of stuff and I don't have a problem with that. It's just not what I personally use it for.

2. If I like what I see in the email preview I'll click the link to your feed and take a look at the last 5 or so posts. If it is junk - spam, any hint of profanity, etc. Done - I'm not following you. The best chance for a follow back is if you have something posted I'm interested in. Maybe it is a short description with a link to an interesting post on the web. If it is something I really like and retweet it, you are definitely getting a follow back.

3. There are some exceptions and I typically follow back the following:

• Local businesses (not based on religion or politics). This includes my favorite Pizza shop in Western Massachusetts. 
• Known organizations, like the National Science Foundation (of course!)
• Some celebrities - how could you not follow back someone like Weird Al Yankovic? 
• Old friends and sometimes friends of friends if I can sort the connections out. 

4. Once I start following you - if I do see any spam, profanity, religion, politics I'm un-following you. I also occasionally go in and cull the list of people I follow and this is the kind of stuff I'm looking for.

 So..... back to my experience today - only 27 follow backs out of 302 new followers..... 8.95%. Yes - there is a lot of junk out there but..... mixed in with the junk there is a lot of good stuff.

You can follow me on Twitter at www.twitter.com/gsnyder

Wavelength Division Multiplexing (WDM)

In my last legacy Public Switched Telephone Network (PSTN) post I covered Statistical Time Division Multiplexing (STDM).  In this post let's take a look at Wavelength Division Multiplexing (WDM and DWDM) methods.

As bandwidth requirements continue to grow for both the legacy Public Switched Telephone Network and the emerged Internet/IP network most of the high bandwidth backbone transmission is being done with fiber optics and a method called Wavelength Division Multiplexing or WDM. WDM functions very similarly to Frequency Division Multiplexing (FDM). With FDM different frequencies represent different communications channels with transmission done on copper or microwaves. WDM uses wavelength instead of frequency to differentiate the different communications channels.

Wavelength

Light is sinusoidal in nature and wavelength, represented by the Greek letter lambda (λ) is a distance measurement usually expressed in meters. Wavelength  is defined as the distance in meters of one sinusoidal cycle.

Wavelength Measurement

Wavelength indicates the color of light. For example, the human eye can see light ranging in frequency from approximately 380 nm (dark violet) to approximately 765 nm (red). WDM multiplexers use wavelength, or color, of light to combine signal channels onto a single piece of optical fiber. Each WDM signal is separated by wavelength “guardbands” to protect from signal crossover. One of WDM’s biggest advantages is that it allows incoming high bandwidth signal carriers that have already been multiplexed to be multiplexed together again and transmitted long distances over one piece of fiber.

Wavelength Division Multiplexing

In addition to WDM systems engineers have developed even higher capacity Dense Wavelength Division Multiplexing (DWDM) systems. Just this past week, Cisco and US Signal announced the successful completion of the first 100 Gigabit (100G) coherent DWDM trial. As backbone bandwidth requirements continue to grow these WDM and DWDM systems are significantly reducing long haul bandwidth bottlenecks.

Podcast - Why We Are Not Google: Lessons from a Library Web site Usability Study

Back in September I had the chance to interview Troy Swanson, an Associate Professor / Teaching and Learning Librarian at Moraine Valley Community College in Palos Hills, IL. In the interview we discussed a paper he published with Public Service Librarian Jeremy Green, also at Moraine Valley Community College. Here's the abstract from that paper published at Elsevier's ScienceDirect.

In the Fall of 2009, the Moraine Valley Community College Library, using guidelines developed by Jakob Nielsen, conducted a usability study to determine how students were using the library Web site and to inform the redesign of the Web site. The authors found that Moraine Valley's current gateway design was a more effective access point to library resources than a mock-up site which incorporated a central-search box on the site homepage. This finding stands in contrast to the observed trends of library Web site design that emphasizes a “Googlized” search.

Troy and Jeremy's findings are very interesting, especially if you are managing/modifying an existing site or are considering creating one. 

You can listen here:

Here's the links Troy refers to in the podcast:

The Next Level (Blockbuster article)by James Surowiecki

• http://www.newyorker.com/talk/financial/2010/10/18/101018ta_talk_surowiecki

useit.com: Jakob Nielsen's Website

• http://www.useit.com/

The Googlization of Everything (book review)

• http://www.googlizationofeverything.com/

Why We Are Not Google: Lessons from a Library Web site Usability Study

• http://www.sciencedirect.com/science/article/pii/S0099133311000280

*******

If you have iTunes installed you can listen to and subscribe to our podcasts by clicking here.

Digital Multiplexing - Statistical Time Division Multiplexing

In my last legacy Public Switched Telephone Network (PSTN) post I covered Time Division Multiplexing (TDM). I described how TDM works and why it does not efficiently use bandwidth. In this post let's take a look at Statistical Time Division Multiplexing (STDM or STATDM or STAT MUX), a much more efficient way to multiplex.

A Statistical Time Division Multiplexer (STDM or STATDM or STAT MUX) does not assign specific time slots for each device. An STDM adds an address field to each time slot in the frame and does not transmit empty frames. Only devices that require time slots get them. 

STDM uses dynamic time slot lengths that are variable. Communicating devices that are very active will be assigned greater time slot lengths than devices that are less active. If a device is idle, it will not receive any time slots. For periods where there is much activity STDMs have buffer memory for temporary data storage. 

STDM Multiplexing

Each STDM transmission carries channel identifier information. Channel identifier information includes source device address and a count of the number of data characters that belong to the listed source address. Channel identifiers are extra and considered overhead and are not data.  To reduce the cost of channel identifier overhead it makes sense to group large numbers of characters for each channel together.

In my next legacy PSTN post I'll cover Wavelength Division Multiplexing (WDM).

Broadband Divide Continues

Earlier this week, the Department of Commerce's Economics and Statistics Administration (ESA) and the National Telecommunications and Information Administration (NTIA) released a study titled Exploring the Digital Nation: Computer and Internet Use at Home. The study analyzed 54,000 households using 2010 census data. Here's some details from a blog post at speedmatters.org:

• 68 % of households used broadband in 2010.

• Broadband adoption rates are slower that mobile.

• Households with lower incomes and less education, as well as Blacks, Hispanics, people with disabilities, and rural residents, were less likely to have Internet service at home.

• Seventy percent of urban households had broadband at home, compared to 57 percent of rural households.

• Less than half (43 percent) of households with annual incomes below $25,000 had broadband access at home, while 93 percent of households with incomes exceeding $100,000 had broadband.

Here's more from the study report:

• As of October 2010, more than 68 percent of households used broadband Internet access service, up from 64 percent one year earlier. Approximately 80 percent of households had at least one Internet user, either at home or elsewhere. 

•  Cable modem (32 percent) and DSL (23 percent) ranked as the most commonly used broadband technologies. Other technologies, including mobile broadband, fiber optics, and satellite services, accounted for a small, but growing, segment of households with broadband Internet access service.

• 2000s – continued to decline from five percent in October 2009 to three percent one year later.

• Over three-fourths (77 percent) of households had a computer – the principal means by which households access the Internet – compared with 62 percent in 2003. Low computer use correlates with low broadband adoption rates.

• Broadband Internet adoption, as well as computer use, varied across demographic and geographic groups. Lower income families, people with less education, those with disabilities, Blacks, Hispanics, and rural residents generally lagged the national average in both broadband adoption and computer use. For example, home broadband adoption and computer use stood at only 16 percent and 27 percent, respectively, among rural households headed by a Black householder without a high school diploma. Also, households with school-age children exhibited higher broadband adoption and computer use rates than other households.

• The differences in socio-economic attributes do not entirely explain why some groups lagged in adoption. Broadband Internet adoption disparities decrease when regression analysis holds constant certain household characteristics, such as income, education, race, ethnicity, foreign-born status, household composition, disability status, or geographic location. For example, the gap with respect to broadband Internet adoption associated with disabilities decreases from 29 to six percentage points when controlling for income, education, age, and other attributes.

• The most important reasons households without broadband Internet or dial-up service gave for not subscribing were: (1) lack of need or interest (47 percent); (2) lack of affordability (24 percent); and (3) inadequate computer (15 percent).

• Households reporting affordability as the major barrier to subscribing to broadband service cited both the fixed cost of purchasing a computer and the recurring monthly subscription costs as important factors. Our analysis of the expanded CPS data suggests that work, school, public libraries, and someone else’s house were all popular alternatives for Internet access among those with no home broadband Internet access service. Not surprisingly, individuals with no home broadband Internet access service relied on locations such as public libraries (20 percent) or other people’s houses (12 percent) more frequently than those who used broadband Internet access service at home.

The study also describes the $7 billion Recovery Act funding directed towards broadband in the U.S. Be sure to check out the complete study document linked here.

Digital Multiplexing - Time Division Multiplexing

In my last legacy Public Switched Telephone Network (PSTN) post I covered analog or frequency multiplexing. Frequency division multiplexing is now considered obsolete technology on the telecommunications network. Analog signals are more sensitive to noise and other signals which can cause problems along the transmission path. They have been replaced with digital multiplexers. 

Digital signals are combined or multiplexed typically using one of two techniques; Time Division Multiplexing (TDM) and Statistical Time Division Multiplexing (STDM). Let's cover TDM in this post.

Time Division Multiplexing allows multiple devices to communicate over the same circuit by assigning time slots for each device on the line. Devices communicating using TDM are typically placed in groups that are multiples of 4.

Each device is assigned a time slot where the TDM will accept an 8 bit character from the device. A TDM frame is then built and transmitted over the circuit. Another TDM on the other end of the circuit de-multiplexes the frame.

TDM Framing

TDM’s tend to waste time slots because a time slot is allocated for each device regardless of whether that device has anything to send. For example, in a TDM system if only two of four devices want to send and use frame space, the other two devices will not have anything to send.

TDM Framing Showing Wasted Slots

They do not require frame space but their time slot is still allocated and will be transmitted as empty frames. This is not an efficient use of bandwidth.

In my next legacy PSTN post, I'll cover statistical time division multiplexing (STDM), a much more efficient way to use bandwidth.

Verizon FiOS Buildout Essentially Done, Fixed LTE Coming

In a quarterly earnings conference call last Friday Verizon confirmed a couple of things I've been saying here for the past couple of years. The FiOS build out is basically done for now and Verizon Wireless will be offerring a fixed LTE option in direct competition with the landlind side of the business.

Here's an interesting Q&A from the VZ - Q3 2011 Verizon Communications Inc Earnings Conference Call held on October 21, as posted at DSL Reports:

Citigroup Analyst: Is there any thought of taking that non-FiOS bundle of presumably LTE broadband LTE voice,what about taking that more nationally and making that more of a national product for you versus just maybe an out of FiOS region but in territory Verizon product? 

Fran Shammo: Well, we are. And you're going to see that come in the fourth quarter with the -- what we now call the Cantenna which is not a commercial name obviously, but it's the antenna that we actually trialed with DIRECTV, which was extremely successful. And again, the benefit of this antenna is it operates the spectrum extremely efficiently. So if you look at a MiFi card or a dongle, this is very, very efficient, way above those two devices which is why it's critical to have that bundle with that Cantenna. So when we launch that you're going to see us go nationally with that type of an offer.

Analog or Frequency Multiplexing

In this post I continue discussing some of the different legacy technologies used by the Public Switched Telephone Network (PSTN). Today let's take a dive into analog or frequency multiplexing.

Analog or frequency multiplexing is now an obsolete technology in the U.S. telecommunications industry. It was used up until the early 1990’s by long-distance carriers like AT&T and MCI and is still used today in other countries. The concept of channel banks was developed for analog multiplexing and this concept is still used today for other types of multiplexing. To multiplex calls each call was given a narrow range of frequency in the available bandwidth. We know all voice call channels occupy the same frequency range – approximately 4000 Hz if we include individual call guardbands. If we want to combine a group of voice calls and separate them by frequency we must translate the frequency of these individual call channels using a process called Single Sideband, Suppressed Carrier Modulation. This technique allowed 10,800 individual voice call channels to be combined and transmitted over one coaxial copper pair. Let’s look at how it was done.

Groups

Individual voice call channels are placed into groups of 12. If we have 12 channels per group and each channel is 4000 Hz we can calculate:

This 48KHz is placed in the frequency range of 60 – 108 KHz. 

Single Group Formation

Supergroups

Individual groups are placed into supergroups of 5 and each supergroup contains 60 individual voice channels. If we have 5 groups and each group is 48 KHz we can calculate:

This 240KHz is placed in the frequency range of 312 – 552 KHz.

Supergroup Formation

Mastergroups

Individual supergroups are placed into mastergroups of 10 and each mastergroup contains 600 individual voice channels. If we have 10 supergroups and each supergroup is 240 KHz we can calculate:

This 2.40MHz is placed in the frequency range of 564 – 2.964 MHz.

Mastergroup Formation

Jumbogroups

Individual mastergroups are placed into jumbogroups of 6 and each jumbogroup contains 3600 individual voice channels. If we have 6 mastergroups and each mastergroup is 2.4 MHz we can calculate:

This 14.4 MHz is placed in the frequency range of 3.084 – 17.484 MHz.

Jumbogroup Formation

Jumbogroup Multiplex

The final multiplexing step involves combining individual jumbogroups which are placed into jumbogroup multiplexes of 3. Each jumbogroup multiplex contains 10,800 individual voice channels. I'm still amazed - 10,800 calls on one piece of coaxial cable!

Frequency multiplexing is now considered obsolete technology on the telecommunications network. Analog signals are more sensitive to noise and other signals which can cause problems along the transmission path. Those long coaxial cables make pretty good antennas. They have been replaced with digital multiplexers. In my next legacy PSTN post I'll cover how digital multiplexing works.

reference: Introduction to Telecommunications Networks by Gordon F Snyder Jr, 2002

Multiplexing - A Brief Introduction

In this post I continue discussing some of the different legacy technologies used by the Public Switched Telephone Network (PSTN). Today let's take a quick look at what multiplexing is.

Before the invention of the telephone both Alexander Graham Bell and Thomas Edison were experimenting with ways to transmit more than one telegraph signal at a time over a single wire. They both realized this was a critical piece if any communications network was to grow in the number of users.

Multiplexing

There are three ways to multiplex or combine multiple signals on the telephone network. They are analog or frequency multiplexing, digital multiplexing and wavelength division multiplexing. I'll dig pretty deep into each in upcoming legacy posts.

The SLC-96

In this post I continue discussing some of the different legacy technologies used by the Public Switched Telephone Network (PSTN). Today let's take a look at how the PSTN designed and tuned for voice communications started to change in the late 1970's with something called an SLC-96 (pronounced "Slick 96").

It's still not economical even today to run fiber into every home but Local Exchange Carriers like Verizon and AT&T have been working to replace portions of the local loop with fiber by running fiber out from the CO into a Remote Terminal (RT) pedestal box in the field called a Multiple Subscriber Line (or Loop) Carrier System or SLC-96. Each SLC-96 takes 96 64 Kbps analog voice or modem signals, converts them to digital and then multiplexes them at the Remote Terminal. The Remote Terminal is connected to a Central Office Terminal (COT) using 5 T1 (DS-1) lines. 

SLC-96 Field Pedestal Configuration

Four of these T1 lines are used to carry the 96 digitized voice channels (1 T1 line = 24 digitized voice channels so 4 T1’s are required to transmit 96 voice channels). The fifth T1 line is used for protective switching and is a backup if one of the four fails.

In my next legacy PSTN post I'll start covering multiplexing.

Hybrid-Terrestrial-Satellite Networks?

That's what Charlie Ergen at Dish Networks is putting together and it makes sense. Ergen's a former professional blackjack player billionaire currently transforming Dish to a wireless mobile video company. An October 17-23, 2011 Business Week Companies & Industries piece titled Charlie Ergen Wants To Beam You Everything does a nice job summarizing where Ergen is taking Dish. Here's some details.

Dish has spent over $5 billion this year on acquisitions of companies in bankruptcy. Here's three of the biggest:

• Blockbuster video chain - video content 
•  EchoStar set-top box maker - IP-based end device connectivity
• Hughes Communications  - satellite Internet service provider

Ergen moved fast with the Blockbuster acquisition, rolling out a Blockbuster branded movie streaming service to Dish customers last month. 

What's next? Ergen currently has a $1.9 billion offer on the books for Hulu which is currently owned by News Corp., Walt Disney and Comcast. Picking up Hulu would give Dish rights to more than one million paying subscribers.

Dish also has pending deals to pick up DBSD North America and TerreStar Networks. These two companies also own wireless spectrum which could be the real prize and an indication of things to come. What the company really needs is more wireless. Access to an existing broad wireless network (DBSD and TerreStar will not be enough) is coming so watch for Ergen to go after a provider (maybe Sprint?) or maybe cut a deal with one of the other providers (Verizon Wireless or AT&T?).

What about competition from companies like Netflix? Peter M. Hoffman from GHL is quoted in that Business Week piece "What Charlie's done is put together content and distribution. Netflix still has to rely on someone else's distribution to deliver its content."

Adding that wireless broadband piece (that Ergen needs for video delivery) could turn Dish into a viable broadband alternative to telcos like Verizon and cable companies like Time Warner. What really excites me is the access Dish could offer rural areas where there is currently not good broadband options.

New UMass President and Massachusetts Green High Performance Computing Center

I've written in the past about high performance computing in Western Massachusetts. 

I had the opportunity today to visit the Massachusetts Green High Performance Computing Center (MGHPCC) in Holyoke, MA. A small group of us toured the Center with new University of Massachusetts President Robert Caret. 
As a UMass Amherst grad (Class of 1979) I have to admit it was pretty cool to be among the group to introduce Dr Caret to High Performance Computing in the Pioneer Valley. I'd like to thank MGHPCC Executive Director John Goodhue for the invite to this special event. Here's a few pics I snapped during the tour. 

Mass Green High Perf Computing Center, a set on Flickr.

Go UMass! Go MGHPCC!