Lightweight Low-cost Linux Laptop

I was looking for a laptop which I can use everyday in class, and occasionally during my short outstation trips. Essential software I use on most days are LibreOffice, R, Python, TexWorks, and of course Mail and Chrome/Firefox, and needed these to run fine on the laptop.  Also, I wanted the the laptop to be no bigger than A-4 size paper, be light, i.e., about 3 lbs or less, and not expensive (this ruled out all ultrabooks).

After spending couple of days browsing the Net, I came across ASUS X200MA: Intel Celeron @ 2.16 GHz, 2 GB RAM, 500 GB HDD, 64 bit, 11.6″ screen, 1 USB3, 2USB2 ports, VGA port, Ethernet port, HDMI and at a weight of just 1.2 Kgs.  You can see the details specs here.  Seemed promising, and very tempting.  I ordered myself one of these with a  vibrant blue shell at an all inclusive cost of Rs.17190.

The new laptop arrived in 4 days, and had only FreeDOS in it. Perfect. I decided to install Ubuntu 14.04. Let me summarise the experience.

  1. I created a bootable USB disk as per instructions in the Ubuntu site (used my wife’s Windows laptop for that!)
  2. I plugged the BootUSB in the ASUS laptop, turned it on. Since latest BIOS driver was already installed I didn’t have to do anything, and the USB drive was detected automatically.
  3. I got a DOS like command interface listing the options:
    • Try Ubuntu without installing
      Install Ubuntu                                                  OEM install (for manufacturers)
      Check disk for defects

      I tried 1st and 2nd options, but it just went to blank screen and nothing happens. I tried pressing ‘e’ in Install Ubuntu, which took me to set params of ‘Install Ubuntu’

    • set gfxpayload=keep
          linux    /casper/vmlinuz.efi    file=/cdrom/preseed/ubuntu.seed boot=casper only-ubiquity quiet splash --
          initrd    /casper/initrd.lz

      I googled the problem and the most common suggest was to replace quite splash with nomodesest nolapic. I tried it. Did not work. Tried various permutations such as only nomodeset, only nolapic, with “quite splash”, without “quite splash”. Still did not work. Very frustrating. After spending 2-3 hours, came across this blog which suggested to change the gfxpayload settings. The final setting that worked for me is:

    • set gfxpayload=1024x768
          linux    /casper/vmlinuz.efi    file=/cdrom/preseed/ubuntu.seed boot=casper only-ubiquity nomodeset nolapic
          initrd    /casper/initrd.lz

      After about 10-20 seconds, lines of text started to scroll, and installation was in progress. Woohoo!

  4. Next, I got the Ubuntu 14.04 installation menu screens as outlined here.  I connected the Ethernet cable to the router to get the all the green ticks on the “Preparing to install Ubuntu” screen. Tip: Keep your ethernet cable handy.
  5. The installation went smoothly. Updated the software, installed synaptic etc.
  6. However, the wifi was not working, or least I didn’t not get the menu options to access it. It seemed like the hardware driver were all installed..  After 20-30 mins of searching, I learned to do:
    $ sudo apt-get install linux-headers-`uname -r` dkms build-essential bcmwl-kernel-source

    After this, the wireless connections were listed in the Network icon on the dropdown menu at the top right tray.  I selected my router, and it worked.

Anyway, after 2 days I have got my new ASUS working, and it seems quite good (as of now).



GST to result in efficient logistics in India

Goods and Sales Tax (GST), effective from April 2016, is expected to eliminate the multiple layers of taxation that exists in India today, and create an unified Indian market. GST will combine and replace the existing Central Excise duty, Service Tax, Central Sales Tax, State Value Added Tax, State surcharges, Octroi, etc (as of today, each state has their own VAT rates, way bill requirements, and other additional taxes).  I expect the following benefits of GST on the logistics operations in India, apart from the possible tax savings:

  • Reduced paperwork: The shipper as well as the logistics carrier (trucker) need not carry the plethora of forms to get through the various state border check posts.
  • Savings in transit time & reduction in uncertainty: Trucks spend on an average 3-5 hours at a border check posts, where the carriers are checked for compliance with that state’s requirements. Further, it is not uncommon to find trucks ‘arbitrarily’ impounded, perhaps for days, at the border check posts for various genuine as well as not-so-genuine reasons. This adds an element of uncertainty in the delivery time of goods. The reduced role of border check posts will result in smooth flow of goods, and more predictable transit times.
  • Savings in Inventory costs: Additional safety stock is required at the customer end to buffer against the uncertainty in transit times. With reduction in the transit time uncertainty, the inventory levels across the supply chain will reduce, resulting in significant savings.
  • Efficient routings: The routings from sources to destinations can be made purely from operations perspective.  For example, transport from Mumbai to Rajasthan can be via either Gujarat or Madhya Pradesh, which can be decided without worrying about states’ surcharges and other requirements.  Also, circuitous routes to avoid state border crossings will become unnecessary.  More ambitiously, splitting and combining of shipments can happen enroute.
  • Improved distribution network: In the long run, the distribution network itself might change. Warehouse locations will be guided more by their ability to serve customer zones, irrespective of the state boundaries.  For example, perhaps companies in Hosur and Bengaluru can coordinate better, eliminate warehouses and operate daily milk runs.  Companies serving markets in Dahod (GJ)-Jhabhua(MP)-Banswara(RJ)  region may move towards integrated warehouses instead of multiple smaller warehouses that may exist just to comply with state norms.

The above is by no means an exhaustive list. There are other aspects that I have not touched up such as the possible growth of 3PL, benefits to SMEs, benefits to the end consumer, etc.  We need to wait for the full details of the actual GST act to understand/analyse its impact on the logistic operations in India.

“Operation Five Minutes” to book tickets

In the 2015 Railway Budget of India presented yesterday, it was announced:

We are introducing ‘Operation Five Minutes’ to ensure that a passenger travelling unreserved can purchase a ticket within five minutes. Provision of modified ‘hot- buttons’, coin vending machines and ‘single destination teller’ windows will drastically reduce the transaction time.  (Shri Suresh Prabhu, Minister of Railways, India)

The “Operation Five Minutes” initiative has been welcomed by commuters, and seen as a move towards shorter queues at ticketing counters.  Now, as an operations researcher, I see some ambiguity in the above announcement. Is the 5 minutes duration the target average time spent by commuters across all stations? Or is it average time spent by commuters at a station? Or is it the average maximum time spent by commuters at a station? Is it the target during peak hours, or non-peak hours, or an average among both? Or is the 5 minutes a sort of service level guarantee for the say 90% of the commuters?  The solutions (and investment) will depend on the specific problem statement.

Let us begin by understanding the link between ‘time spent in station to purchase ticket’, ‘transaction time’ and ‘queue length’.  Little’s Law states:

Avg. Queue Length = Avg. Transaction Rate x Avg. Time spent in System (at station).

The Law is applicable in steady state systems where average transaction rate keeps pace with the average arrival rate.  Let us suppose that it takes about 30 seconds to get an unreserved ticket (transaction rate = 2 commuters/ minute) at one counter, and on an average 20 commuters are in queue, then the average time spent in ticketing office = 20/2= 10 minutes. Now, to reduce the time spent to purchase ticket to 5 minutes, we need to increase the transaction rate to at least 4 commuters/ minute. We can achieve this in two ways, either reduce the transaction time to get ticket at each counter to 15 seconds, or have two counters where each takes 30 seconds to issue the ticket.

Let us pause here to see the enormity of the problem. A recent study showed that about 8,00,000 (8 lakhs!) passengers purchase unreserved ticket per month at a major station. This translates to about 8,00,000/(30*24*60) = 18.51 commuters per minute (=transaction rate).  The average queue lengths was about 250 commuters (considering parallel counters), which means the time spent in system = 250/18.51 = 13.5 minutes.  The implications are far reaching. On an average 1 commuter ‘wastes’ 13.5 minutes to purchase a ticket ⇒ 8,00,000*13.5 minutes = 7500 man-days lost in waiting at just one station!

Now, if the average time spent in system is 5 minutes, and 8,00,000 commuters are served per month, then the average queue lengths that is going to greet the commuters will be 18.51*5 = 92.55 commuters.  Suppose there are 10 parallel counters/ vending machines etc, the average length of each queue is approx. 9 commuters only!  Sounds really good.

This move by Railways to guarantee ticket purchase “within 5 minutes”, if implemented only at the major stations with an aim to reduce the average time, is still a significant move.  The use of technology-based solutions (hot buttons, smart phones, etc) as well as operational solutions (dedicated windows) are all needed to reduce the transaction time. However, their deployment needs to be carefully planned so that the intended benefits are achieved, and keeps pace with the very increasing number of commuters to be served per day.

Analytics vs. Operations Research

Analytics, as a buzzword, has taken over the world in the last 5 years. A quick search for its meaning shows analytics to be a “systematic computational analysis of data or statistics” (Oxford), or “the method of logical analysis” (M-W).  We have been doing such analysis for a long time: what is special now? Well, it is the combination of rigorous analytical methods with large, diverse data sets, a.k.a big data, that has caught people’s imagination. Internet of Things has made available a very very large amount of data (both structured and unstructured) . This has led naturally to the question, “Can we use the large scale data to help make the world/ state/ company/ community/ oneself better?”  — thereby bringing the use of analytical methods to prominence.  Analytics itself is now broadly talked about in four theme areas: descriptive analytics, diagnostic analytics, predictive analytics and prescriptive analytics.

Operations Research (OR), as a field, has been all about using analytical methods to improve decision making.  This long standing view of OR and the recent buzz around analytic has quite understandably led to the ‘analytics vs. OR’ d1-s2.0-S037722171400664X-gr1ebate, for both try to achieve similar things: help make better decisions using information, analytical models and technology. The relation of OR with analytics has been well put in the recent (2015) paper on “Operational research from Taylorism to Terabytes: A research agenda for the analytics age“.  They describe a dianoetic (:pertaining to though or reasoning rather than intuition) management paradigm, summarized in the figure on the right (reproduced from the above paper). They have pointed out that:

The availability of both the data, and the tools that complement it, have had significant impact on decision makers, the demands of businesses for new and further reaching forms of analysis, and indeed the central methodology of the paradigm itself.

Further, they have discussed both an isolationist approach (between OR and Analytics) and a faddist approach, and have highlighted the need for a balanced approach.  I completely agree with this view.  The OR community should play an active role in the analytics world by creating improved decision models by leveraging (big) data and related tools.

At least closer home (in India) operations research has always been under-appreciated, and this could be an wonderful opportunity (for OR) to step out of the shadows.


Michael J. Mortenson, Neil F. Doherty, Stewart Robinson, Operational research from Taylorism to Terabytes: A research agenda for the analytics age, European Journal of Operational Research, 241(3), 2015, Pages 583–595 [doi:10.1016/j.ejor.2014.08.029]