National Library Board innovates with private, hybrid and public clouds

Little innovation will come about if there are no constraints in life, said Kia Siang Hock, deputy director of technology and innovation at Singapore’s National Library Board (NLB).

 

“If everyone could buy and spend on technology freely, companies would see little need to innovate,” shared Kia at the Open Cloud Conference 2014 held in Singapore. Kia added his organization’s involvement with the cloud stems from limited budgets from the government coupled with the need to optimize the digital engagement experience for NLB’s users and staff. These tasks include making increasing amounts of resources, such as the National Archives, available for research by NLB users.

 

The NLB’s foray into the cloud includes experimenting with private, hybrid and private clouds.

 

Going private

The organization made sure to start off with a private cloud due to security concerns stemming from its position as a government statutory board. “Having a government organization use the private cloud would require more consideration,” shared Kia. “We could experiment and innovate with cloud applications more freely within the confines and safety of the private cloud.”

 

The board’s current IT infrastructure includes some 500 odd servers, with 340 virtualized over the years, for both production and staging environments. 340 servers run on 29 physical hosts, an average of 15 virtual servers per physical machine, which allowed for significant cost savings in power usage and hardware expenditure.

 

Kia’s team faced  a number of issues when its foray into virtualization started 6 years ago, including tension between apps and systems and the lack of competency to run a virtualized environment. “We faced performance issues, including complaints that virtualized servers were running slower than desktops,” shared Kia, adding the IT team likely started off too aggressively when they aimed to have 25 virtualized to 1 physical server. “Never underestimate the complexity of virtualization,” Kia quipped.

 

NLB’s tender specs now specify that it is mandatory for staging servers to run on a virtualized environment.

 

First to be implemented on NLB’s private cloud were internal platforms to run a variety of service applications, a Platform-as-a-service (PaaS) deployment. “We implemented an adaptive service architecture portfolio and took the approach of building architecture on the platform only as we needed it.”

 

NLB’s primary purpose for the service applications on its private cloud was getting users digitally engaged with its media and content. Such services for users were set up as components of its private cloud-based PaaS 5 years ago, and most have not changed since. Over the years, Kia’s team built up components that can be used or any project.

 

Next to be implemented in NLB’s private cloud was a Hadoop cluster for text analytics, made up of 13 virtual servers, for the ease of fuss-free expansion. “We wanted to be able to push related documents or resources from all our collections, no matter what the format, whenever a user searched for a particular term or looked at a particular resource,” said Kia of the NLB’s analytics service, which uses established open source algorithms for text analytics.

 

Going hybrid

A hybrid cloud model was chosen for the NLB’s enquiry management system, a SaaS-based system deployed to manage patron enquiries across all the board’s enquiry channels. Non-sensitive data is hosted externally, with sensitive data kept encrypted internally.

 

Kia revealed that the product’s SaaS nature meant the application could be deployed quickly and cost savings for such a model were cut by about a third versus hosting the application wholly in-house.

 

Public case

The NLB’s experiments with the public cloud dealt with the archiving of numerous websites (in the hundreds and thousands) that the board deemed important to Singapore’s history. These websites needed to be crawled regularly, a bandwidth-intensive task. Post-crawling, indexing needed to be performed in order for the archives to be searchable. “We utilized the public cloud for crawling and indexing due to the large amount of bandwidth required. The content would subsequently be stored in our archives after the crawling and indexing is complete,” said Kia. “We only perform such tasks once a year so there was little point, cost or scalability wise, in engaging our own servers to perform these tasks.”

 

“The impetus for cloud is clear – customer expectations are getting higher and users expect everything to work in a library. Combine these with intense cost pressures and complex ecosystems and you have a clear case for the cloud,” said Kia.