Tuesday, October 06, 2015

My take on the 21 Inc bitcoin computer announced recently

Since the launch announcement of the Bitcoin computer for $399 by 21.co, there have been a lot of questions and debate about what does it really trying to achieve in the bitcoin ecosystem. The FAQ on the 21 Inc website is not useful to answer a lot of questions floating around, though this may change with time as launch date comes nearer. In the interim, I have compiled the following FAQ which may be of use if you (like me) are also wondering about the answers to a lot of unknowns. Please contact me for additions/corrections or comments.

Why use Raspberry Pi for the hardware when I can so the same using my own computer or a laptop? Why would anyone in their right mind pay 400 dollars for a raspberry pi + an API wrapper?

The 21Inc Bitcoin Computer is a fully functional bitcoin miner, a full bitcoin node and a micro-payments server. It has been rightly pointed about by many others that it is not economical using it as a standalone bitcoin miner – one source says you will be able to mine only $38 worth of bitcoins a year, and even that is not certain. 21 Inc has also acknowledged this fact. Another source says that the hardware included with this device is only worth $250 and thus the $399 is actually the price of the hardware + the included API and micro-payment server.

In my view the aim of 21 Inc is to enable hobbyists and developers come on to the blockchain and start using the (21 Inc + bitcoin) ecosystem to create applications and trade on it via micropayments. 21 Inc is maybe aiming to create a bitcoin platform on which a lot of applications are built using their APIs. If there are a lot of people having the 21 Inc device, it would be very easy for them (with the right kind of applications built on top) to transact with each other using bitcoins for micropayments. At this point in time, to my knowledge, there is no single platform which enables bitcoin transactions using a standard API which allows applications to be built on top of it without charging any transaction fee (happy to be corrected if wrong).

Ok so why not build a library of APIs and publish it instead of selling a piece of hardware? The issue with that is that it will have to integrate with a host of bitcoin wallets where the end users will end up paying a transaction fee. It will still not address the issue of privacy and anonymity (as the wallet provider knows who you really are).

If this device gets smaller and faster (which it should with time due to Moore's law) then you could potentially plug it to your cellphone or refrigerator as a USB drive and thus enable e-commerce on this basis. In the future, this will also enable the IOTs (Internet of Things).

In my view bitcoin micropayments with minimal transaction fees is the key objective of this device.

When there is hardly any Bitcoins I can mine using this device, then what's the difference between using this computer and running a full node on a normal computer? Why include a miner in this in the first place?

Let us look at the alternatives of being able to mine (remember that the 21 Inc computer is a full bitcoin node, meaning it is a miner and wallet rolled into one):
  • If we have an API sitting over the top of a bitcoin wallet enabing payments, we have no anonymity and privacy. There will be an additional transaction fee which will be a killer deal for micropayments. In fact this is not a viable option as using a wallet means you have handed over your details to the “centralized” wallet provider who can track each and every payment you send or receive. You can also look at an option of using multiple such 21 Inc computers which you use to be truly anonymous and decentralized instead of hooking yourself on to a central wallet.
  • Using your computer to mine bitcoins on its own and take/send payments is possible, however you will have to set up a micropayments server and wallet to get started, which leads to issues as I described above for the wallet, and challenges in setting up a micropayment server to work with minimal transaction fee. A developer does not need to spend 20-25 hours setting this up and then buying bitcoins from other sources, which is worth atleast $150 in time saved – the assumed saving you can make if you go the DIY route.
  • The most important reason in my view to include a miner is to enable a large ecosystem of miners which will process micropayments at nearly zero cost. Most miners will not process micropayments today and this will mean tha those transactions are delayed and slow. The 21 Inc computer will enable a lot of mining power to be allocated for micropayments.

Why does not 21.co include a set of BTC – say worth $50 with this device so its easier for people to get started? Why cannot I buy BTC myself to get started?

I can see two reasons for them to do this:
  • It is easier for them to process returns and credit card chargebacks for their customers, as in theory someone can ask for a refund of the price paid after transferring the bitcoins to himself. There is maybe an option in the future to launch a non refundable cash only device with a built in bitcoin wallet in the future.
  • If you were to buy the bitcoins yourself – say worth $100, and want to climb the micropayments bandwagon, you still need a mechanism for transacing with minimal transaction fees. There are additional issues with using a wallet as I have outlined earlier.

Will this be a closed or open API from 21.co?

My assumption is that the API from 21.co will be a closed API. They can not make any money with an open API. They may definitely end up open-sourcing some parts of it though to increase adoption and accelerate applications sitting on top of it.

What is the business benefit for 21.co to do this? How do they make money?

It is like asking how Google was supposed to make money with just a search engine before they figured out how to monetize the ads. I think the aim they have is to be a dominant and controlling player in the bitcoin ecosystem. They may have some ideas on how to monetize this as well in the future, but I cannot think of any at this time other than they making money @ $150 per device for their software. They may also be banking on figuring it out along the way if this grows as Google did.

What are the use cases you can think of for the 21 Inc computer?

Some use cases that come to my mind are:
  • Today advertisers pay websites for placing ads for the attention of the users visting them. Why can the advertisers not pay the users for their eyeballs directly a portion of the ad palcement charges (cost per click) if both the adertisers and users can do micropayments using bitcoin?
  • Trade my spare bandwidth or cpu power for bitcoin without bothering about tracking the payments received.
  • IOTs and sensors will get enhanced over time with miniaturization of this device. They offer a variety of use cases, like an individual selling his medical parameters like heart beat rate which by itself is of low value but aggregated data is of value to a lot of people. Another example can be a weather sensor selling data in a locality which when aggregated nation-wide can be of great value to many agencies.
  • You can reward peers for posting your links on social media, or pay the people who read your opinions. Its like a decentralized ChangeTip with zero payment processing charges.
  • From one source : Imagine a datacenter of network switches that have one of these installed. These devices could then get paid for their service, maybe based on how much bandwidth they successfully route. Then they pay for their electricity cost. Any device that gets too far into debt would be "killed". Any device that accumulates enough wealth could "reproduce" by ordering more of itself or copying it's software to one of the "dead" switches.

Disclaimer: I have no financial or commercial relationship with 21.co. I have no friends or family members working there. All sources have been attributed and if someone has been missed, happy to incorporate.

Thursday, December 11, 2014

How do you compare Impala, Vertica, Teardata and IBM BigInsights?

A lot of us have been challenged to perform product comparisons for Impala, Vertica, Teradata etc.  The question you may be asking can be:

  • What is a good use case to use Impala (or Vertica)?
  • What are the pros and cons of using one over the other?
  • What kind of reporting tools exists for these?
  • What is the hardware each of these runs on?

I have collated (after trawling through a lot of websites and product documentation) a comparison matrix.  I hope you find it useful.

Tool Licensing Model Hardware  MPP Columnar Database In Memory? Reporting Capabilities Visualization Tools
Hadoop/Hive/Pig *Apache Open Source x86 linux based commodity hardware Yes No No Standard Tableau, Qlikview, SpagOBI, Spotfire
* CloudEra support Supports opensource as well as commercial tools
Spark * Apache Open Source * x86 linux based commodity hardware Yes No Yes Standard Tableau, Qlikview, SpagOBI, Spotfire
* CloudEra support * Works on top of Hadoop Supports opensource as well as commercial tools
Vertica License for Software and Applicance,~ 1.1M for 100TB *Specific Hardware Yes Yes No Standard Supports ODBC/JDBC to allow tools to connect in addition to VSQL, does not have out of the box applications.
IBM Infosphere BigInsights Per node based licensing x86 linux based commodity hardware Yes Limited, runs on top of Hbase No Enhanced Comes with out of the box applications for analytics, predictive modelling etc. 
Teradata Aster per TB  Applicance/Engineered system Yes Yes No Standard Aster Discovery Platform integrates with several leading BI and reporting tools such as Tableau, MicroStrategy, Tibco Spotfire, IBM Cognos and SAP BusinessObjects
Actian Per node based licensing x86 linux based commodity hardware Yes Yes No Standard Tableau, Qlikview, SpagOBI, Spotfire
Supports opensource as well as commercial tools
Impala * Apache Open Source * x86 linux based commodity hardware Yes Yes No Standard Tableau, Qlikview, SpagOBI, Spotfire
* CloudEra support * Works on top of Hadoop Supports opensource as well as commercial tools

Please flag any errors or corrections to me. 

Wednesday, December 10, 2014

Why is hotel booking from MakeMyTrip or GoIbibo cheaper than booking directly for hotel?

I have been using the travel portals extensively for booking travel for friends and family, and one thing that has always amazed me is that if you end up contacting the hotel directly for extensions or upgrades, the rates they charge are always higher than the ones on MakemyTrip/GoIbibo/Yatra etc.

Why is it so?

The hotels get a guaranteed/high volume from MakeMyTrip.  Due to this volume they offer discounts.  These volume discounts are not applicable in case someone approaches the hotel directly.  The hotels care about MakeMyTrip more than direct customers and are happy if you come via MakeMyTrip rather than come directly. 

Tuesday, March 13, 2012

Why Costs and Tracking Error Are Not the Only Key Items for Choosing a Suitable Index Fund.

There are many Index Funds available in the Indian market that tack the Major Indices like S&P CnX Nifty and the BSE Sensex. Typically investors end up choosing the fund with the lowest expense ratio and tracking error. However I would here try to illustrate that these are not sufficient by themselves to help make you make the best informed choice.

Index Funds are a good bet for a long term investor so if you are looking at short term trading tips or gains using them then this article is not going to offer much help.

The index funds buy and sell individual stocks in the same proportion as the underlying Index. Over a period of time these individual stocks declare dividends which can be accumulated over time and invested back in the underlying index. This is the concept of the “Total Returns Index” which basically reinvests all the declared dividends in the index. As an example for the Nifty the Total Returns Index data is as below (source here) :








NIFTY Total Returns Index







A simpler way to understand this is that if all the dividends declared in the companies comprising the Nifty were reinvested in Nifty itself from 30-Jun-99 to 1-Mar-2012, the value of the Nifty would have been 6409.8 (which is 1187.7*6870.6/1256.4) instead of 5339.8. We seem to be getting a dividend yield of nearly 1.63% over the long term in the Nifty.

While an investor looks at the expense ratios and tracking errors of index funds, somehow this aspect of dividends and total returns is totally ignored in the decision making process. In order to highlight this further, I have taken the performance for all index funds launched before 2005 tracking the Nifty to be able to take a long term view of their performance. All data is as on 7 March 2012 using Value Research.


Index Fund




3 YR


5 YR


7 YR


Expenses (%)


Principal Index Fund








UTI Nifty Index Fund








Franklin India Index Fund








SBI Magnum Index Fund








ICICI Prudential Index Fund








HDFC Index Fund – Nifty Plan








Birla Sun Life Index Fund








LICMF Index Fund – Nifty Plan








Tata Index Fund-Nifty Plan








ING Large Cap Equity Fund








Canara Robeco Nifty Index Fund
















Nifty Total Returns








Goldman Sachs NIFTY BEES (ETF)







Let me make some observations on the data above:

  • Most of the index funds are not even able to match the performance of the Nifty. (I will discuss the special case of ICICI Prudential later.)
  • In addition to collecting the expense ratios which are very high, the funds are coolly pocketing the dividends as well since they are falling behind the index. I do understand there is a need for the funds to keep some cash and there are trading costs, but the numbers speak for themselves.
  • To my knowledge HDFC Index Fund is benchmarked to the Nifty Total Returns Index, not the Nifty but it is not even able to match Nifty in any period.
  • ICICI Prudential Index Fund seems to be outperforming the Nifty – even the Total Returns index in-spite of having a very high expense ratio. I would be very skeptical of investing in this fund as this seems to be an actively managed fund.
  • The ETFs like GS Bees, Kotak Nifty are benchmarked to Nifty total returns, not just Nifty. As a case in point GS BeeS regularly declares dividends as over a period of time the accumulated dividends make the NAV higher than 1/10th of Nifty. In case of Kotak Nifty and Kotak Sensex fund, the dividends are accumulated and so the NAV of these funds are higher than that of the underlying units by a small margin. See below (source here)

So to summarize:

  • Look at the dividends angle (by checking if your fund is benchmarked to total returns index and is delivering as promised.
  • Ensure you have lower costs and tracking errors.
  • You may be better off with a fund tracking Nifty total return index having a higher expense ratio than a fund tracking just the Nifty with a lower expense ratio.
If you are investing in an ETF, you require a Demat account. Additionally you are required to pay brokerage charges while buying and selling these. These charges are over and above the expense ratios of these funds.

For various reasons these index funds have not picked up well in the Indian market unlike the west, and Manshu does make some valid points on this aspect.

* I have used the data from June 1999 as that’s the one available on NSE website for Total Returns Index.