Data centre migrations can involve a lot of pain. And a transnational migration – let alone several – in order to comply with data sovereignty regulations is almost an unthinkable nightmare. But for Jean-Manuel Becker, the director of IT services at Melbourne-based educational services company Pearson Research and Assessment, it's becoming a routine part of his job.
Pearson delivers NAPLAN testing to primary and high school students in all Australian states bar Queensland. But the Melbourne-based company also delivers testing services internationally, in Asia and a number of Middle Eastern companies, including the United Arab Emirates and Oman.
These countries frequently have regulations about the storage and processing of examination data which often means it can't leave the country, ruling out any cloud-based processing of results.
Pearson creates exam papers with barcodes that identify which student the paper is intended for. Answers are scanned in and processed with a combination of OMR (optical mark recognition – which recognises answers for 'bubble'-style multiple choice questions) and clipping of images. The clippings are fed into a Web-based environment that lets teachers mark the exams. After exams are marked the results are compiled and sent to educational authorities.
At the height of NAPLAN, which is held for students in grades 3, 5, 7 and 9, the company processes about 3.1 million exams in the space of three weeks. The online marking environment is used by some 1100 Australian teachers.
In all of the countries that Pearson operates in, it's a very seasonal business, Becker said. "It will happen for two months during the exam or even one month and then nothing will happen there," he said.
In addition to dealing with a significant number of users and IO-intensive batch processing, the nature of the business means that it has to deal with sensitive personal data. In some countries, Becker said, this means exam data can't be stored outside the country, or even, in some cases, outside the ministry of education.
Becker's solution: If the data can't go to Pearson's data centre, Pearson's data centre will go to the data.
"What I've got in Melbourne to provide the national test environment [for NAPLAN] is a standard SAN system with virtualization, VMware and all that," Becker said.
"But what I had to develop for the Middle East project is a portable data centre."
Becker's data-centre-in-a-box is a single 2U unit running an array of virtual machines to provide Oracle 11g and other databases, a firewall and load balancers, 20 virtual Linux boxes running the Apache Web server and JBoss to deliver Pearson's online marking application, and some .NET applications running on Windows virtual machines that process exam paper images delivered by high-speed scanners.
This virtual data centre is literally portable: Pearson moves the box housing the appliance around countries in the Middle East depending on where it's providing its testing services. "I just need one power plug, one Ethernet plug and I'm working," Becker said.
Previously, Becker was using NetApp storage systems and Dell servers. "I was moving racks around, which is not very easy to do in the Middle East by plane," Becker said. "I had to move large UPSes and well and even portable air conditioning."
Now instead he just has to cart round his 2U unit – a NX-3000 series appliance from Nutanix – and a single UPS, and he no longer worries about aircon, instead just relying on whatever cooling is already available at different sites.
Pearson purchased its appliance in January, although Nutanix only this week opened its first Australian office in Sydney. The company, which was founded in 2009, has also announced a Melbourne-based technical support operation for the Asia Pacific region. The decision to open an office here was been driven in part by the high level of virtualization in Australia and New Zealand, said Vignesh Shashidhar, the northern region territory account manager for Nutanix.
"I became aware of [Nutanix] in November," Becker said. He conducted extensive testing on an NX-2000 series appliance the company loaned him to ensure it could cope with the IO-intensive workload Becker had in mind for it.
"We purchased the [NX-]3400 which is a bigger box than we were testing with so we were confident there would be enough processing power to hold my Oracle environment, my SQL environment and Linux and Windows, all in the 2U environment."
Storage and compute are integrated in a single box, and Becker's data-centre-in-a-box offers up to 128 virtual CPU cores and 1TB of memory.
"It's a very condensed and packed data centre," Becker said. Becker's largest VM – 32 CPUs and 120GB of RAM – runs his Oracle database, and flash memory in the Nutanix appliance is used for caching.
Nutanix appliances are better known for delivering virtual desktop infrastructure, Becker said. "I'm really pushing [it] to the edge but it's working well," he said.
Becker has housed his mini data centre in a rugged enclosure that's also employed by the US Army, which is also a Nutanix customer. The size and weight means that it can be moved by two people, and Pearson can now fly it between countries in the region as required.
Pearson is currently examining the possibility of expanding to Turkey and Lebanon, Becker said. If this goes ahead, he is likely to expand to a 'fleet' of portable data centres, the IT director added. "It's small, it's effective – you've got fast IO – and it's portable," he said.