Tina Ifremer sounds like the latest R&B singer you’ve never heard of. But, actually, it’s the name given to the Atempo backup software that’s been in use for 23 years by the French national institute for ocean science and technology.
That’s quite a lengthy period for a piece of software to be deployed, given that customers are much more likely to change things every few years.
And for good reason. The institute – Ifremer, for short; it’s an abbreviation – needs to be able to access records that date back over that period of near two-and-a-half decades.
Ifremer specialises in oceanographic studies, aims at sustainable use of the sea’s resources and the marine environment, and shares information with a wide range of organisations to those ends.
It employs 1,500 in a workforce that includes scientists, engineers, technicians and admin staff across five centres in France and its overseas territories. Meanwhile, it has around 20 operational sites across the world’s oceans.
It’s a public sector organisation – an EPIC, in the terminology – established so that while most of its funding comes from government ministries, Ifremer has the capacity to respond to service requirements from elsewhere to help with funding.
Centralising data
A third-party support provider, RIC, based in northern France, looks after operational maintenance, infrastructure, networks, backup, storage and the security of research data. It also looks after high-performance computing (HPC) on the Datarmor supercomputer, which carries out oceanic modelling that includes temperatures, salinity, and so on.
Ifremer selected Atempo as a data protection provider back in 2002. That included protection of bare metal servers, databases and the organisation’s network-attached storage (using NDMP). A total of 400TB (terabytes) is protected, with Tina deployed at each site and backups centralised at a site in Brest, France.
That configuration allows for rapid and reliable restores from the central site while also protecting local sites. Ifremer works to a 3-2-1 backup schema and also stores backups off-site on tape. That occurs via a Quantum DXI disk-based backup appliance that stages data off to tape, which optimises performance during backup windows and builds in data reduction.
Four challenges; four purchasing criteria
The key challenges and, as a result, the key criteria in product choice were:
- The need to be multisite, with an IT environment shared across the world’s oceans, which complicated centralisation of backups.
- Large volumes of data, with hundreds of terabytes of data resulting from research that needs to be stored.
- Large numbers of small files, which presented issues of performance, especially in indexing millions of small files.
- Self-service restores, carried out by the researchers without needing to depend on the IT team.
“We want to protect the intellectual heritage of Ifremer,” said Béatrice le Berre, who works on the team that runs Tina at Ifremer. “Ifremer has the challenge of data sovereignty. We are partisans of the principle of open data in science, but at the same time, we have to be aware of espionage and attempts to take critical data.”
Jérôme Le Letty, who also works on the Tina team, added: “Being able to navigate temporally was another key factor in the adoption of the solution. That allows us to visualise the state of a disk in the Tina UI [user interface] and view its deleted files, maybe in error, and recover them, and that’s essential.”
Next: VM backup
At each of the remote sites, backups take place daily and locally, going to tape or disk depending on the size of the site. Once a week, local backups are sent to the central datacentre at Brest.
“The solution does its work and does it well,” said Le Berre. “Tina works on the different platforms that we use, even the Linux servers.”
Next, the plan is to back up virtual servers to Tina. Also, Ifremer wants to create a digital replica of the oceans on the Datarmor supercomputer, which itself is undergoing a huge update.
Ifremer pays for Tina via a three-year renewing licence with no restriction on data volumes stored. The software is updated twice a year.
