Just Dial Pvt Ltd, which opted for implementation of open source, is a local search engine that caters to more than 25 million users spread across 240 Indian cities. The search service
Explaining the implementation of open source, Sandipan Chattopadhyaya, the CTO of Just Dial says, "Understanding the business model, it was a strategic decision to move to open source infrastructure. This was not an evolutionary process, but a choice. The challenge was about how to make a suitable solution for a dynamic business model." The entire solution was developed in-house. Implementation of open source entailed rebooting of the entire business infrastructure from MSDOS to Linux.
One of the biggest considerations in the implementation of open source is life expectancy of the project. That is because there are many come and go situations in the open source world, which makes stability of the entire solution of utmost importance. Chattopadhyaya avers, "RHEL has been one of the most stable things so far. For us, that was the best bet." Just Dial uses the LAMP (Linux, Apache, MySQL and Python) architecture in its implementation of open source.
The second factor to consider in Just Dial's implementation of open source was standardization, which arose because of too many alternatives. With implementation of open source - RHEL, Just Dial could waive off such worries. If one was to take a snapshot of the kernel installation, it would look completely different across various functions. There would be no difference between similar machines. However, the operating systems of Web servers and database servers would hardly match. Hence, vestigial parts have been removed, and the I/O systems have been very differently tuned—towards RAM affinity in the database server, and for disk affinity in the Web servers—while going in for the implementation of open source.
Data was the third key
factor with regard to implementation of open source. The company's strategy for the server-side
and client-side is completely different. It has a distributed data center model residing in nine
locations across India, and disaster recovery (DR) sites in three locations. Chattopadhyaya
informs, "We are talking about humungous data volumes. It will be extremely foolhardy to make it
reside on one location."
The data growth is around half a GB per day. The US site is also fully hosted in India for the time being. Just Dial's implementation of open source resides on around 200 servers (of Dell PowerEdge make). The company has not opted for server virtualization at all. It has designed the application in such a way that there is no peak load capacity to be concerned about, since these always operate at a steady pace. However, Just Dial does have many parallel structures and hot standbys. Every server has two backups. According to Chattopadhya, a fail-proof mathematical model is being used for load balancing as part of the project for implementation of open source.
With regard to other factors pertaining to the implementation of open source, when it comes to storage, SAN is being used only for backups (in some cases). Otherwise, the organization uses the machines themselves for its own storage. Chattopadhyaya states, "It is basically a game between RAM, anticipated load, and how you balance, which gives us scalability." The company is using Microsoft SQL database and Sphinx. Just Dial also has its data structures for optimized searches. On the client-side, 98 per cent of its desktops are Linux boxes using Ubuntu.
Interestingly, the implementation of open source has come with such benefits—capability-wise, trending-wise, and business intelligence-wise—which could not even be thought of in the older version. Earlier, the company used to take around 80,000 calls per day. After the implementation of open source, Just Dial has been taking a staggering 2,70,000 calls per day. "We are talking about value engineering. People may think of it as a project, but we think of it as a product including our service," states Chattopadhyaya while highlighting the benefits of Just Dial's implementation of open source.
It took around nine months from end-to-end for the implementation of open source project to go live. It was done in three phases. Firstly, the call-taking part and the website were done, followed by the marketing and customer service. No training was given to users because of the simplicity with which the implementation of open source could be handled.
On a concluding note, Chattopadhyaya points out what the implementation of open source is ultimately going to mean, "The vision is to make it absolutely high-tech. It has to be so high-tech that when someone uses it, he feels that no technology is being used."