However, the company struggled with poor quality video calls and sluggish desktop application performance, particularly during peak Internet traffic hours. Latency between the Singapore and Monaco offices was unstable and unpredictable. Lines were getting dropped, and video for video conference calls was freezing up.
“We were facing latency issues and stability of the IPsec VPN between Singapore and Monaco,” said Lucas Foo, manager of IT at Petredec.”This latency was impacting our Monaco office’s ability to report its trading positions back to headquarters in Singapore, and it was also impairing traders’ ability to communicate with each other.”
“The video conference quality was poor, and every call we had from Singapore to Monaco had frozen frames,” said Foo. “We also encountered failed transactions on desktop applications when the connectivity was poor.”
At first, Petredec's IT team had hoped dedicated WAN optimization controller (WOC) appliances would eliminate application performance lags and video conference quality problems. These WOC controllers, from the now defunct Expand Networks, used protocol optimization methodologies to prevent network latency. They also compressed and/or changed data streaming over the network and prioritized traffic according to business priorities. Unfortunately, Petredec continued to have WAN performance issues in both its Singapore and Monaco offices, Foo said.
“While we did not have true network visibility tools in-house, we were able to determine that the performance issues were strictly on the WAN," he said.
The company then tried optimizing the WAN by improving the WAN link via a WAN acceleration appliance.
“The thought was that by accelerating information flows across the WAN by using a combination of data compression, data deduplication caching of duplicate files or parts of files so they didn’t have to be resent over the WAN, that we would be able to attain the quality of service we wanted for our video calls and for application performance on distributed desktop computers in remote offices around the world,” said Foo.
Petredec looked at WAN accelerator products from a number of different vendors but didn't find any that would improve its videoconferencing. “There still didn’t seem to be a way for us to achieve a reduced and stable latency on the WAN between our two sites in Singapore and Monaco," Foo said.
Frustrated with persistent WAN performance problems, Petredec then opted to try a cloud-based WAN optimization service from Aryaka. “Videoconference quality improved with no drops or stills, and performance of remote desktop protocol sessions was much better with good application response times, Foo said.
“Our end user experience is very important to us, and we measure it by logging the time needed to complete an activity and the overall experience in getting the work done," he said. "Productivity increased with the cloud-based WOC. Work got done faster, and there was less idle time.”
[Find out how to evaluate WAN optimization products in WAN Optimization Part 2: Put Performance Second."
Altogether, it took Petredec one month to implement WOC in the cloud, using a dedicated two-person team. This was followed by a two- to three-week testing period.
“Initially, we were concerned with latency for the videoconferencing,” said Foo, “This ended up being a quality of service problem in the switch that was resolved once QoS settings were moved from the WOC Expand box to the firewall to get the end to end QoS required by the cloud.”
With its move to the cloud, Petredec also acquired a new set of reports that enhanced its network monitoring. “The reports are helpful in determining current usage, the applications that are in use, and the stable latency that we are looking for,” said Foo.
“Also, our vendor has a Web portal that gives us greater visibility of the network because it helps us to better understand the last mile connectivity and the network uptime.”
In the future, Petredec wants to break down traffic by protocol and perform traffic shaping. “We have traffic shaping to ensure that the right application gets the bandwidth assigned,” said Foo. “Ultimately, we’d like to separate out the bandwidth consumption on the RDP (remote desktop protocol) from the video traffic, and then allocate the right bandwidth levels.”