Logistics provider Transfreight's switch from array-based replication to hypervisor-based replication gave the company more flexibility in picking storage systems, cut costs and maintained its short recovery
Transfreight, which helps manufacturers optimize their supply chains, is based in Erlanger, Ky., but has its main and secondary data centers in Kitchener and Cambridge, Ontario.
It changed data replication products from EMC RecoverPoint to startup Zerto Virtual Replication (ZVR) this year when it switched from EMC Clariion CX3-20 arrays at its two data centers to Dell Compellent in its main data center. The company kept one Clariion at its secondary site, and uses ZVR to replicate between sites over the WAN. That was not possible with RecoverPoint, which requires EMC storage at both ends.
"Our data centers are asymmetric," said Tamas Nemeth, Transfreight's manager of IT systems. "We don't have the same setup at the second location. It's not really cost effective for us to buy the same storage when we need to upgrade technology. We were looking to save money and still provide high availability."
ZVR and RecoverPoint have common roots. Zerto's founder and CEO Ziv Kedem also started Kashya, which developed RecoverPoint. EMC acquired Kashya in 2006. The data replication products differ, however, in that RecoverPoint is limited to EMC arrays, while ZVR works inside VMware hypervisors and supports any storage.
"We needed to find an application that could do the same thing we had before, but could replicate between different storage brands," Nemeth said.
Nemeth said he bought the hypervisor-based replication product in February and began using it for production data in June after several months of testing. Transfreight protects Microsoft SQL Server and Exchange and its mission-critical financial applications with ZVR. The company runs 300 SQL databases on VMware and virtualized Exchange in 2008, Nemeth said.
Instead of taking snapshots, ZVR uses a virtual replication appliance software module that continuously replicates data from virtual machines to the remote site over the WAN. Every write command is captured, cloned and sent to the disaster recovery site.
Nemeth said Transfreight's recovery point objective (RPO) is less than two minutes for its mission-critical applications.
"Our goal is to have a very short RPO," he said. "We had a very short RPO with RecoverPoint, and we certainly didn't want to give up on that. That was a critical part of our decision."
Zerto lets customers create Virtual Protection Groups (VPGs) that protect an application across virtual machines with one automated failover and failback process. Nemeth said he has 15 VPGs set up, and each can fail over independently. He said he hasn't needed to fail over yet, but the process has worked in testing.
"The big advantage is we can have any kind of storage now," he said. "And because it's on the hypervisor level, we don't have to use VMware Site Recovery Manager or other third-party tools to failover. ZVR doesn't just failover storage, it fails over the virtual machine(VM) too. RecoverPoint took care of storage only. We had to failover VMs manually or with other tools."
Nemeth said another potential benefit is ZVR allows replication among three sites. Transfreight doesn't do that yet, but he said it may add another site.
He also said ZVR's continuous replication causes no noticeable performance drag. "We sized our VMware environment accordingly so there was no negative impact on production."