Equity Office is headquartered in Chicago and has about 40 remote offices. Its primary data center is located at headquarters, with an active disaster-recovery center in Madison, WI. Each remote office is connected as a spoke to both headquarters and the backup data center as hubs. The data centers also act as hubs for the Cisco VoIP system.
Until the change to the Oracle (formerly Talari) SD-WAN, each office was connected to a Verizon Business MPLS service by at least one T1, with a few larger offices having two or three T1s.
The MPLS service was put in about a decade ago, and shortly afterward, the company underwent server consolidation, pulling virtually all servers out of remote offices and centralizing them in the data center.
The only way Equity Office could do that was by installing WAN optimization gear from Riverbed at all sites, effectively reducing WAN traffic enough so branch workers could get seemingly local performance from central data center servers. “They were the missing piece that made it possible,” says Momchev.
Each office has DSL connections that offered Internet backup to the MPLS network, but where there were MPLS outages, the performance of the backup was poor, he says.
But in the mornings when employees were logging on for the day, traffic spiked so much that even with the Riverbed gear, performance dropped. Workers sometimes needed to wait a minute for a file to download. Since he was already optimizing the bandwidth, his only recourse was to buy more bandwidth, but the costs of bigger MPLS links were too high, he says.
So he looked at Elfiq, Ipanema and Oracle (formerly Talari), ruling out Elfiq because at the time it couldn’t support the particular MPLS/DSL hub-and-spoke configuration Equity Office had implemented. He trialed Ipanema and Oracle (formerly Talari) gear for six or seven months each. He says Ipanema’s gear couldn’t make per-packet decisions needed to keep up VoIP quality. “I have VoIP to take care of, so that was a deal breaker,” he says.