Monday, 23 July 2012

What my 6 year old using an IPAD got me thinking about…….


 

As I watched my 6 year old learn how to use the new IPAD in matter of hours and how comfortable she was I noticed something “she managed to download most of the applications of her liking from ITunes and % of operation which she did using a browser where down to 10%”. More interestingly virtual keyboard was used even lesser. This actually got me thinking and specifically in the larger context in terms of Information Technology and what are most likely things going to happen in the coming years.

1. Smart & Rich Application find there way back into main stream: With most new devices be it phone or tablet they provide much richer interface and the browser will soon be out fashioned by the RIA applications installed on the phone or tablets. Interesting advancement in GPU and glass taking the center stage for display allows the potential for rich application development. I see a lot more new frameworks come up in this area. The new User Interface is not limited to the browser. Interactive User Interface Design is really the need of the hour. A closer look at Windows 8 tablet and most of your daily use applications will be replaced by live tiles. The User Interface Design plays a big role in IPAD too “intuitive interactive user interface design and essentially easy to use” that’s the reason major % of IPAD users in US are senior citizens.

2.Live Data: With Rich Applications comes the need of having lot more sensible data on the display. Data is alive essentially its push. I don’t see the devices in a position to either process that kind of data or cache that kind of data, at the same time I see the responsive to be the state of art. This is where the cloud comes into play. The MVC frameworks will have to revisited here to a large extent. The views on the devices are data rich and bank on lot of business intelligence and analytics. This part is essentially cloud.

3.No Keyboard please: The Applications running of device will be intelligent would guess the next expected request of the user well before time.  The touch plays a major a role in graphical and non-graphical views and data fetches on these operations will be very many. The chances of user going to the Keyboard will be very limited.

4. Majority Applications in the Cloud:  Cloud is the way to go. The bank or the mall next door all of it will be on the cloud in no time. I see majority of the application move to the cloud by 2020. The Cloud geographies have to evolve , at present we see none , the geographies will be defined based on security needs. This may sound slightly stupid. I see a reason for cloud geographies defining based on business needs of data segregation.

5.Internet of things: Most the devices which are likely to come in the mainstream in the coming years will have the capability of connectivity to internet (ANT, ANT+). These devices will be transmitting data to applications on the cloud & can get to the point of taking instructions back from the applications. These instruction could be basic or could be more advanced in terms running a diagnostic check and sending data back to the applications on the cloud. The application on cloud may decide later if a technician needs to be sent on premise or send a set of instruction to solve the problem. With an estimated number of devices been close to 30 billion by 2020 the cloud for devices will have to really huge… & intelligent.

6. Connected all the Time do you really need a Phone: It may feel a bit out of sci-fi movie , an individual with multiple devices is connected all time then why would you need mobile phone. The mobile phone is miniaturized device attached to hear, intelligent one which tells you a person wants to chat/talk to you. Do you really need an expensive mobile to do that job.

 

I know a lot more is likely to change in our daily lives the way we do things will change. Above are few of things which I think are likely to change. They can be many more. This is just light reading.

 

 

Monday, 16 July 2012

Cloud–Application Migration Experiences


Just finished a public facing capital markets portal migration to cloud.  A quick background the application provides advisory services in capital markets to about 100k users and architecturally built on the Microsoft stack of ASP.NET MVC 3 , Entity FW , WCF , Sql Server and involved the complexity of a service bus built on ESB toolkit for integrating with third party application for pure data pull & BI and dash boarding capability used were third party. Additionally a lot of third party controls used. Source Code Repository and Build Management was Team Foundation Server. The application evolved over a period of 3 years and deployment standpoint a DC and fully functional DR. The DC had roughly 4 Web Front End Server(inclusive of Reporting) , 2 Application Server, 2 Database Server clustered in an Active / Passive Model.
Agile was followed and overall estimation initially had been at 18 weeks with 4 releases. The actual completion took 22 weeks 3 major releases and 3 minor releases.
Cloud another disruptive technology, my heartfelt sympathies for developers. I too am a part of the same brigade. Cloud is indeed the biggest disruptive technology which has come in. The developers are at yet another unlearn and learn curve. This one is strangely different from the last one been Web where the developer had to move out of the client server era into a 3 tier one which pretty much got the developers to learn html, client side and server programming and whole variety of incremental education coming in. Cloud presents a new challenge following are some of my thoughts. What I’m penning down is what has journey been through this the pain areas and blind curves.
The application that we have written is about 3 years in functioning and an n-tiered applications. These application run in the companies data centers controlled by the infrastructure folks, now the applications are hosted in an unknown location depending on which deployment model of cloud we are targeting the re-architecting can vary from a complete rewrite to simple migration.  The complete application had to be re architected piece meal wise to suit the customer needs based on feature priority.
The First choice ideally would be to move the application to PAAS (ex: Azure or Amazon), some key considerations which need to be taken into account are
    • Decomposing the architecture into web and worker role. The web and worker role are stateless and loosely coupled which means rewriting the communication across the web and worker some example are rudimentary queues.
    • State Management is an area which needs special attention. Thumb Rule: Everything in cloud is stateless.There are multiple strategies on how to manage state in cloud common cache etc..
    • Third party controls are not welcome , Have addressed the work around’s this.  Some applications which really required the third party controls not web these are business specific were hosted on VM i.e Persistent VM offering by Azure, which allowed us to install whatever was required and keep the state. These VM were easily accessible to the other web and worker role VM.
    • MVC Revamp: If one is using MVC frameworks which by the way is the defacto standard for web based application today, the wiring between the controller and view may need to be looked back into. The model which is existential surface for objects has to be stateless. The database is no more relational not a compulsion one can still end up using relational but the motto is go No Sql. The datastore can vary anything from blobs, table (not relational store), hadoop or a relational store. This means a new Object Cloud Store Mapper framework.  Written an Object to Cloud Store Mapper framework will soon be sharing it as open source. The framework can work across multiple types of storage provided by Amazon, Azure , on Premise Database Server. CQRS Pattern had been implemented for faster reads.I think it helped.
    • Services Layer: If the application has a well defined services layer which serves the integration need there is good news all PAAS offering have a good enough Service Bus.
    • Identity Management:  Most PAAS have well defined authentication and authorization policies with federated choices like  authenticating with an on premise Active Directory, LDAP, Facebook, Google… The application needs small effort to support the cloud identity management.
    • Storage:  The storage options in the cloud are many ranging from non relational blobs, tables, hadoop to relational Sql Azure.
    • Data Migration: From on premise sql to sql azure was straight forward.
    • Resources Accountability:  A close watch on the code written is needed in terms how much Compute, Memory, Network & Storage(CMNS) static code analysis can assist in code clean up. Every useless instruction which gets execute is compute. The costing model on the cloud is based on CMNS.
    • User Interface: This is a big area, Although not cloud a lot has changed in this area in last 2 years, with introduction of various device types the tablet, the smart phone and browser innovations to HTML 5, Metro style UI the list is endless. The new mantra in UI is interactive designs this can change the programming model largely. HTML 5 alone has brought such innovative stuff discussed later.  The User Interface has been redesigned with more interactive design to suit multiple form factors. UI needed special consideration.
    • Microsoft Application Blocks for Azure: Auto Scaling ,Transient Fault Handling Block
    • Source Code Repository:  Have used GIT and seem to work exceptionally well does the job, There was no need to go in for some expensive tool like Team Foundation Server. The GIT server was tweaked to support some IDE experience. The source code of the same can be shared.
    • Test and Staging Environment- To cut risk both these environment were setup on Azure. Had to tweak GIT to work along these environment. Performance Testing for the application was kind of crazy I still haven’t got a handle on this. I’m assuming everything will work ok its cloud after all infinite compute & storage. Still working on it will update when we close this.
I’m trying to see if I can share the artifacts and source code for framework as open source. Lets hope.