Mobile, Social & Local

Analytics and the use of the Cloud

Posted: March 1st, 2010 | Author: julio.luis.miyares | Filed under: Cloud Computing, Cloud Map Reduce |

If understanding Data is the true measure and worth of collecting it in the first place, what is now standing in the way at a minimum of the capability to crunch the data using a slew of tools adapted for that purpose? The answer is nothing , well kind of. With the publicly available services in the Cloud such as Amazon’s MapReduce, the inherent capability it there for all but the most uninitiated to make sense of the mountains of data point collected by a typical high volume website.

Of course one can just plugin Google Analytics onto their website and they are good to go with a robust solution that captures, aggregates ,slices and dices and presents pretty reports via a web form about a host of activity on a website. Nevertheless there are always instances where the need to unlock additional understanding inherent in all that data requires some additional processing and some additional tools.

One of the biggest costs for large scale data processing has usually been the need to have an operational infrastructure in place to excise, collate, aggregate and slice the mountains of data that are generated by your typical website or application usually measured in gigabytes and terabytes. Of course all that operational infrastructure is always bound to some strategic and business critical processing leaving precious little horsepower available for for those extraneous random questions that come up on a recurring basis for any business. By the time the request for that question to be answered is queued up and delivered by the typical IT group ,it is highly likely someone more nimble somewhere else has already answered the question and gotten a lead on adapting their business with the answer in hand.

With the wide availability of Cloud Services such as Amazon’s Map Reduce, the timeline to answer the question can be markedly reduced. Admittedly you still need Technologists either in house or outsourced that understand the components of Map Reduce and how to actually leverage it’s power for the needs of the business but the need to wait for precious cycles on the IT’s host complex, partnering Software engineers and Operational Admins to make the environment available with the hosts and necessary software and connectivity to actually run jobs can be short circuited to almost zero time.
You also need the Analysts that can come up with pertinent questions specific to the business they are in and with an understanding of the data elements and relationships between them that are currently captured or can be captured and can help answer specific questions. The technology is not smart enough to do it by itself.

A non technical question , are mature companies with a long standing streak of being in business availing themselves of this form of Cloud services computing? I would not expect them to have the same drive to operate on a shoestring as the majority of startups do but are Cloud services of viewed as an valid option by most IT groups in mature organizations?

Yes, more then a handful of companies outsourced their Development but especially Networking over the course of the past few decades leaving those esoteric functions to entities like MCI that had scale and a knack for staying on top of the evolving technologies shuffling bits around.


Outsourcing Software Development

Posted: February 6th, 2010 | Author: julio.luis.miyares | Filed under: Technology, Technology Outsourcing |

Bangalore June 14-22,2006 If we outsource our Software Development , we can save considerable labor costs. For each USA based employee we can have 4 or 5 based on current ratios of India development resources.  True, if the disparity between where the resources are currently located and where they will be outsourced to is large, you will save some personnel costs but is that the sole determinant whether it is a good idea or not?
Dispersing technology development/support (or operations) throughout the world has benefits besides cost and based on labor rate differentials. If you have a large enough inventory of sites to support that operate 24/7, having a team of System Administrators that overlap various timezones should improve the responsiveness when things inevitably go wrong. Also allows for system maintenance during the slow hours in whatever market the particular site has it’s lull.  Also, you can smooth out the demand and supply imbalances  in actual engineering resources that make acquiring engineering resources problematic regardless of what the pay scale is in certain markets.
On the other hand, if you are working in a domain that is subject to large amounts of uncertainty and market pressures in leap frogging features, what does a lower labor rate coupled with the latency inherent in time differences and maintaining context get you except 2nd , 3rd or further down in timely meeting the market need?
I have worked in situations where design is in the Eastern Time Zone and Web development on India time. A change in design would easily take 48 hour turnaround just to get the correct understanding of what the feature represented and how it fit into the rest of the product/design flow before anything could possibly be engineered.  Given that Frontend Web development is relatively low technology this minimum latency is dysfunctional and has an attendant cost in frustration on the part of the involved product team including executive management as well as missed opportunities in the marketplace.

The aforementioned should not be cast as a criticism of engineering in India as those engineering teams are  thrust into a position where success is not achievable.