French car parts maker Bontaz used to find IT a stop-start process. Every time it suffered the slightest loss of electrical power, it had to restart its entire IT infrastructure. That was then, however, and seven years on, it’s pleased with its move to software-defined storage from DataCore, which it has just upgraded to run Dell NVMe flash storage.
The challenge for Bontaz was to achieve IT that never breaks down. The company is a world leader in hydraulic sub-assemblies built for almost all of the world’s carmakers. One car in two globally uses its components, which come from 24 factories in 11 countries.
“All IT systems in our fleet, including R&D servers and those that work with manufacturers’ design departments, are in France,” said Benoit Belleraud, chief of systems and networks at Bontaz.
“That totals 130 virtual servers that run applications such as PLM [product lifecycle management], CAO [corporate admin and ops] and simulation, but also virtual desktops, file sharing and email.
“All of it runs on two VMware ESXi servers,” he said. “Up to 2015, we mirrored the contents of these two servers to two others in a remote location. Synching was automated and in real time at the level of storage so that if we had an outage in one place, we could restart activities from the other. But in reality, it didn’t work.”
The key problem? The company’s site, near Annecy – in the foothills of the Alps – often suffered power cuts and aircon breakdowns. And on every occasion, secondary servers needed to be started and up-to-date data injected.
“That operation would take several hours,” Belleraud bitterly recalled. “You couldn’t call that a business continuity plan.”
By 2015, Bontaz decided it couldn’t go on losing hours of critical activity in this way because of its direct impact to employees and collaboration with carmakers.
Belleraud met with the company’s integrator partner, Resilience, also based in Annecy. Its diagnosis was that with synchronisation between sites happening only at the level of the Hitachi storage, it lacked a system that automatically restarted volumes on working servers wherever they were.
Resilience recommended storage virtualisation – or software-defined storage – namely SANsymphony from DataCore.
The benefits were immediately visible, with initial configuration only taking a short time. That involved defining “datastores”, which meant segmenting storage into logical modules of a defined size and launching it on specified servers according to rules of priority and bandwidth.
Physical servers ran at two sites on four ESXi instances in total, with VMs shared between them. If an outage occurred at either site, SANsymphony reconfigured systems by switching activity from one to the other, automatically and immediately. SANsymphony also provided backup via continuous data protection (CDP).
“We backup all our servers daily to a third location using Veeam,” said Belleraud. “This backup allows us to roll back to previous versions more easily, such as when a file is deleted by accident but we can’t go back to the backup of the night before. CDP is complementary in the sense that it backs up locally and in real time all data created and modified in the last 24 hours.”
Belleraud illustrates this with an example. “CDP has saved us quite a few times,” he said. “You can find some important documents have been lost, such as those related to subsidiaries or documents that are just absolutely critical. And, in a minute, you can get them back.”
Bontaz was deployed seven years ago, with Dell disk arrays installed at the same time. These recently reached end-of-life, and Bontaz underwent a hardware refresh.
“A pleasant surprise was that the upgrade would only cost us €165,000, which is the same as in 2015 because our capacity needs are no more than they were then,” said Belleraud. “We have stayed at the need for 50TB in production.”
He explained that the onset of more greedy forms of data was compensated for by the departure of some applications to the cloud. In 2023, email will migrate to Microsoft 365. “But we will always retain locally any data that is too critical to go to the cloud,” said Belleraud.
Hardware configurations have been upgraded also, with SANsymphony now running on two Dell servers with NVMe SSDs of 8TB each, where previously capacity had been on spinning disk. Another difference is that the critical CDP cache has been extended to 3.2TB compared with 1TB previously.
“The transition to the new system was very easy,” said Belleraud. “In a week, we switched everything over to the new infrastructure, without suspending any production services.”
The only snag noticed was that the two SANsymphony instances – the old and the new – contended to present their console to administrators, but the problem was resolved with DataCore’s support.
“This migration was simple, just as SANsymphony admin is every day,” he said. “We only have to do anything about once a month on the system console, such as when we need to modify the size of a datastore.”
Read more on software-defined storage
- Software-defined storage vs NAS/SAN: What are the options? We look at the pros and cons of software-defined storage and weigh up when it’s a better option than buying NAS and SAN pre-built hardware shared storage arrays.
- Container-native storage: A definition, and what to ask suppliers. We look at container-native storage and its benefits as a way of bringing persistent Kubernetes storage to containerised applications.