Event Information:

  • Fri

    Array-based replication, software replication keys in data protection

    Whether it is done in hardware or software, replication is becoming a key piece of data protection projects.

    U.K.-based marketing services firm Brightsource and Collierville, Tenn.-based protective equipment company MCR Safety were among the organizations last year that added replication capabilities to aid backup and disaster recovery.

    Brightsource installed a Nimble Storage SAN for array-based replication, and MCR Safety installed InMage physical and virtual appliances that combine continuous data protection (CDP) and host-based replication.

    Brightsource gains RPO flexibility with array-based replication

    If you had a 500 TB database and you modified one minor field, the whole lot would need replicating.

    Ed Rollinson,
    head of IT for corporate marketing, Brightsource

    Brightsource's Nimble CS2000 system is dedicated to one customer with storage needs requiring a great deal of flexibility, according to Brightsource head of IT for corporate marketing, Ed Rollinson.

    Brightsource also uses an EMC VNXe 3100, but Rollinson said he was unhappy with the way that system handles replication. He said the VNXe replicates any change in a database, meaning, "If you had a 500 TB database and you modified one minor field, the whole lot would need replicating. We were looking at a data set of about 4.2 TB [for a customer project], and it would never complete. There is little control over that replication."

    To get data from the VNXe off-site, Brightsource puts data on a portable hard drive, and an employee takes it by taxicab to an off-site data center. "That's the position having an inefficient replication process has left us in," Rollinson said. "It's cost us considerable amount of time, money and energy doing it that way. That's why we realized we needed something that did data replication more efficiently."

    Nimble's replication is built into its Cache Accelerated Sequential Layout operating system and managed through VMware vCenter, rather than a separate application that EMC requires. Rollinson said Nimble's replication lets him set recovery point objectives (RPOs) for each data set. For instance, he replicates application data every hour, VMs every 12 hours and archive data every 48 hours, instead of replicating 4.2 TB at once.

    "We can replicate large chunks of data over time, or we can run smaller replication jobs more frequently," Rollinson said. "We can change those schedules as quickly as we want. Before, our backup and recovery process was focused on what the technology could give us. I was deeply uncomfortable about that. I always felt the business should take the lead on where we want to be with our DR scenario and how our recovery point objectives and recovery point times met."

    He said his challenge now is whether there is enough bandwidth to handle replication, "but that's easier to manage."

    Brightsource added a Nimble array to store time-sensitive data from a financial services customer that sends out customized information to distribution lists. The data and the emails sent often change due to factors such as varying interest rates and changes in services, resulting in frequent updates.

    "This project takes a large amount of data that would represent a particular marketing campaign," Rollinson said. "They would give us a clean data set. We take that data set and determine what assets needed to go into that marketing collateral.

    "What is incredibly important is not the general throughput of the system, but what we termed our burst capabilities. When we started evaluating products, we said if the system is faster and better than what we do now, that's a good start, but also, at what point does this system fall on its knees? And can we adequately plan for burst work and the challenges that might bring?"

    MCR receives block-level protection with host-based replication

    Replication was part of a two-part data protection transition for MCR Safety, which installed an InMage hardware appliance and a virtual appliance in August 2013. The first step was to replace its backup process of using IBM Tivoli Storage Manager to back up to tape. That was slow and also left MCR with most of its data without disaster recovery, according to MCR systems engineer Michael Cantkier. The next step will be to add a DR site and replicate to it with InMage.

    Cantkier worked as a consultant for MCR and sold the company the InMage products before joining MCR full-time. He said the company only backed up "things considered most important" before switching over.

    Now, MCR uses InMage to protect its data on Microsoft Windows (Exchange, SQL and file servers) and IBM AIX servers (DB2). The virtual appliance runs on VMware virtual machines and the hardware appliance protects the AIX systems running DB2 databases for MCR's enterprise resource planning system. The appliances back up to an IBM DS4700 array, which used to be MCR's primary SAN, but is now dedicated to InMage for data protection.

    Cantkier said he looked at several options before recommending InMage. He considered doing a software-hardware refresh around TSM, as well as bringing in software from Actifio, Veeam Software or Visions Solutions' Double-Take. He said he picked InMage because of its CDP, solid AIX support and replication.

    "We needed to support Windows and AIX," he said. "Everybody does Windows, but not everybody does AIX."

    Although MCR is not using replication yet, Cantkier said that was a requirement.

    "We'll turn InMage into a replication and DR tool," Cantkier said. "We'll replicate to our own cloud and fail over to a different site. Today, InMage protects our entire environment on a block level up to the moment. Any time a block is written to a disk or LUN, that block is split and written to another LUN on another SAN. InMage keeps track of those changes."

    He said InMage's CDP feature makes restores "like a DVR, where you slide back in time to where you want. The most recent stuff is more granular. You can recover from two to five minutes go. The older you get is less granular. The last week we have is a daily point in time that you can recover to. You choose that point in time, and it spins up a virtual LUN that you mount; it grabs files. You have multiple options for recovery. You can spin up an entire virtual machine that's in an alternate location, or you spin up a physical machine as a virtual machine in VMware."