"I would keep hearing about how wonderful their virtualization projects were and how cool the hardware was," Pousson told a roomful of users at the 2012 Yokogawa User Group meeting this week in New Orleans. "So when management came and told me that they wanted a complete parallel Exaquantum system that would store all of our flow, level, temperature and pressure variables at one-second intervals, I told them that it would be very expensive unless we went to virtualization of the servers. They agreed, so I began to investigate."
Exaquantum is Yokogawa's plant information historian application, and the plant had five physical Exaquantum servers at the time, each installed at a different time with different standard hardware. "With five servers of all different ages," he said, "it's always something. You need five separate backups. Disk or memory expansion requires hardware. All these issues are simplified in a virtual environment."
"You also improve reliability," Pousson added. "If a physical server fails, virtual machines roll over, and a storage area network (SAN) has more redundancy. It was clear that the virtualization path would mean easier upgrades and expansions. Keeping five machines at the same level is a chore, and we have at least two more Exaquantum systems planned."
Outside Help Needed
But Pousson did find that he didn't have all the internal expertise and time to fully complete his virtualization vision. So he outsourced some of the virtualization work to an outside consulting group, BCI. "Decisions may require help from outside," he said. "We needed help with defining our goals, with selecting which virtualization software to go with, what kind of servers we needed and who would do the work. Selecting BCI to help us made the project go well."
When it comes to choosing a virtualization platform, "follow the lead of your IT department and use what they use," Pousson recommended. "This gives you a complete army of virtualization technical support from within your own company."
Even with BCI's help, plant personnel spent a lot of time working on the project. "We also needed to work closely with our enterprise IT department, including help with DNS, firewall and security policies," Pousson said.
Post-virtualization, the logical architecture is pretty much the same as before: five Exaquantum servers, each connected to its own OPC server. But the physical architecture is very different. Four servers share a SAN, and if any server box goes down, any virtual machine that was on that server automatically reboots on a different box.
"You really can lose a server and keep going," Pousson related. "But this is complex stuff--virtual disks need to be sized thoughtfully, and it's easy to get confused setting up multiple machines. Disk performance is critical," Pousson continued. "Historians create massive disk traffic. Multiple historians will hit the SAN very hard. We found we should not have bought 7500 rpm hard disks, but should have spent the money on the 15,000 rpm disks to begin with."
The virtualization effort has gone well so far, and CF has now deployed quad-redundancy for some virtual servers using the VMWare VSphere virtualization solution. "We have two machines moved over, and we're working on the third," Pousson said. "By February they all will be moved over, and our physical servers will be turned off. Our goal is to have at least 12 Exaquantum servers running virtually, and I hope to get five to 10 years' service on this without major changes."