• Home
• GOVERNMENT CXO
• E-GOVERNMENT
• CITIZEN ENGAGEMENT
• TAX AND REVENUE MANAGEMENT
• GOVERNMENT PROCUREMENT
• GREEN GOVERNMENT
• GOVERNMENT DATA MANAGEMENT
• GOVERNMENT SECURITY
• GOVERNMENT CLOUD
• GOVERNMENT ANALYTICS
• GOVERNMENT GIS
• CONNECTED GOVERNMENT
• DIGITAL INCLUSION
• EDUCATION IT
• HEALTHCARE IT

Connected Government, Government Data Management

Building a network that never jams

By Jianggan Li | 20 April 2011

Governments are putting more platforms, processes and services online. In order to cope with the increasing processing and storage needs, lots of public sector organisations, looks at virtualisation.

Raghu Subramanian, CTO of Juniper Networks Asia Pacific, believes virtualisation itself could cause a problem – traffic congestion within the government information network.

“Information is being pulled towards a centralised location, while users are increasingly scattered and mobile,” Subramanian says. “The pressure on government data centres is only going to increase exponentially.”

He bets that the trend is only going to keep accelerating. “The cost of storage and servers keep falling, and consolidated data centres will become bigger and bigger.”

Will cloud be of any help? “Ultimately cloud is still made up of data centres,” Subramanian says. “And governments often prefer private cloud, in their own data centres.”

He adds that if organisations do not review their overall network architecture, more server virtualisation will sooner or later introduce headaches for the network team; and storage virtualisation will further compound the problem.

“Traffic congestions can be prevented, but very difficult to solve once your data centre is in that situation,” he says. “Instead of an array of servers, now we have a pool. And traffic going between different virtual servers is very complex – can you tell how much east west traffic you have versus north south flow of data?”

Subramanian thinks a completely different mindset is needed: instead of building with brick and cement, now we should look at interconnecting with steel and glass.

He draws attention to Clos Network which, first devised by Charles Clos in 1953, is a theoretical idealisation of practical multi-stage telephone switching systems that would never jam. The industry didn’t go that way because the availability of VLSI (Very Large Scale Integration) circuits made switching no longer a bottleneck.

As data centres grow larger, and traffic, which was not a problem in the past, becomes an issue, people start looking at designs such as Clos Network again.

While new data centres could be modelled completely in an optimised manner, according to Subramanian, older ones could grow into the mould gradually. “Once you need to expand, you design the expanded part as an optimised module,” he says. “And you replace out-of-date modules with new designs step by step – you will eventually have an overhauled data centre.”

As many new paradigms which cater to large scale networks, universities’ research networks would become the guinea pigs. “The capacity and workload of these experimental networks often represent the trend for other industries tomorrow,” explains Subramanian.

And municipal planners could take the same approach in planning roads such that congestion will become less a problem for future cities, Subramanian envisions, only that redesigning and rebuilding the infrastructure are much more complicated than what we could do for data centres.

---