When he went looking for a new backup application, Waterbury Hospital Virtual Systems Administrator George Adamo...
needed something that could handle virtual machine protection and still support old-fashioned tape.
That led the 400-bed Connecticut hospital to PHD Virtual Backup & Replication for data stored on virtual servers, which includes the hospital's Cloverleaf system that provides real-time access to patient records. Waterbury backs up the Cloverleaf database with 12 VMware hosts -- six in production and six at a colocation site -- through storage area network replication.
Adamo said he began searching for a virtual machine backup product in late 2009 because he was unhappy with the way IBM Tivoli Storage Manager (TSM) supported those backups. The hospital's servers were about 70% virtualized at the time.
He first checked out Veeam Software, but Veeam did not support tape until this year. He also looked at Quest vRanger Pro (now part of Dell), but it cost a lot more than PHD Virtual.
PHD was focused on Citrix XenServer then, but Adamo said the vendor never gave up on VMware backups. "PHD had cash from Citrix to develop a product to back up the Citrix hypervisor, but they did not abandon the rest of the market."
Since then, PHD has expanded its offerings. Through a partnership with Kroll Ontrack, PHD revamped its Microsoft Exchange and SharePoint recover capabilities. PHD also added a CloudHook module to send data to the cloud, and moved into disaster recovery this year by acquiring disaster recovery testing vendor VirtualSharp. PHD is also preparing to preview its Microsoft Hyper-V support.
Adamo said he purchased PHD's ReliableDR based on the VirtualSharp technology, and is preparing to implement it to replicate to a colocation instead of using VMware Site Recovery Manager. He also said he is looking forward to PHD's Hyper-V support.
"The original reason PHD was attractive was its link to tape," Adamo said. "As time has gone by and the old-time tape guys like me have gotten more comfortable with disk-based backup, there's less of a need for that."
Most of Waterbury Hospital's backups go to disk now -- either an IBM DS4800 on-site or a Dell Compellent array in a colocation site. Adamo said he is impressed with the deduplication ratios and speed of recoveries since he switched over to PHD and disk.
"Cloverleaf and payroll are the most critical systems that I had to back up with PHD, and recovery is rapid," he said. "The most common thing is recovery of a database backup for a DBA [database administrator]. It [has] happened three or four times in the last couple of months. I go into the PHD console, mount the backup as a CIFS [Common Internet File System] share and email the link to the DBA. I say, 'Email me back when you're finished so I can take it down.' Ten minutes later, they have their stuff back."
PHD Virtual can't protect all of the hospital's data. Adamo said he still uses TSM for physical servers that host financial and material management applications. Medical imaging systems are also still physical and get archived to tape. He is starting to look at Linear Tape File Systems (LTFS) to facilitate tape archiving.
"We'd like to get rid of tape, but I think it will continue to be an archive medium for us," he said. "We have onerous archive needs with our laboratory data, picture archiving and communication system (PACS) storage and cardiology images."
Adamo said he uses Hyper-V a lot more now because of its attractive licensing and the way it works tightly with Microsoft Active Directory. He said he will likely eventually use PHD there, too.
Adamo said his one criticism of PHD is its lack of support materials for backing up Exchange. When he started protecting Exchange 2010 with PHD software, he actually found more useful information on the Veeam website. "The approach is similar; you have the appliance take a snapshot, and then you mount it and do the backup. The architecture is the same
"One of the issues when working with Exchange 2010 is the time out between database availability group servers. The servers think the cluster is compromised because there's a ping delay between the two nodes, so they'll fail over and you have two databases failing over. You need to adjust the timeout to 10 seconds.
"If you're doing a backup of Exchange, no matter which node you go to, you can back up the databases and restore them whether they're active or not because you're replicating data. If I want both nodes backed up, should I be staggering backups so I don't have an issue? That involves some trial and error. PHD should be doing that research and presenting it on its website."