In an effort to improve application response time and become more productive, SNL Financial LC, which collects, analyzes and disseminates publicly available financial information,
The arrangement worked so well that SNL now has 60% of its approximately 625 employees in the U.S., and 40% in India and Pakistan. However, as the firm ramped up its overseas operations, employees there began experiencing lags in application response time lags, according to CTO Galen Warren.
The sites in India and Pakistan were linked to SNL's data center via virtual private network connections on the public Internet. This arrangement proved adequate at first, Warren says. "Our internal entry and data collection systems are custom-built, optimized to work efficiently over a WAN." But as employees gained expertise, they were given additional tasks that involved working with commercial applications like Microsoft's Excel. Many of these applications use chatty, wide area network (WAN)-inefficient protocols like Common Internet File System (CIFS).
Employees at the remote sites often ran applications on hosts at headquarters in remote desktop mode. When an application wasn't WAN-friendly, there was often significant lag time and slow response time between when the user input commands on the keyboard and when the results were seen on the screen.
Last spring, Warren's team began looking for a product that would optimize and accelerate performance and improve application response time over the WAN. The move was especially timely, given that SNL was moving its Indian operations into a new, wholly owned building.
After evaluating several offerings, SNL picked Blue Coat SG, a combination application accelerator and security appliance from Blue Coat Systems Inc. in Sunnyvale, Calif. One factor in that decision: The financial firm was already using SG boxes to scan Web browser traffic for viruses and block end users from certain websites.
Going with Blue Coat's accelerator did not save SNL on hardware costs. The company still needed to install an SG box at headquarters and at the site in India, because its existing SG boxes lacked the horsepower to run both security and application accelerator modules concurrently, Warren says. Still, the two companies had an existing, and satisfactory, relationship, he says.
More importantly, Blue Coat SG's accelerator gave good results. During tests on a simulated, high-latency network, end users were able to open files as much as 48 times faster. FTP file transfers were completed up to 18 times faster.
SNL agreed to purchase Blue Coat's accelerator in July. The office in India got it up and running in a matter of days, Warren reports.
Once an SG accelerator box was installed at each end of the overseas link, end users experienced a significant performance boost and improved response times for both Web-based and desktop applications. The lag for applications accessed on the corporate host via Remote Desktop "is almost imperceptible," Warren notes.
Blue Coat SG utilizes several techniques to make applications run more efficiently over a WAN connection:
Prioritization, queuing and bandwidth allocation: SG directs network devices to give precedence to high-priority and latency-sensitive applications.
Optimization: SG optimizes chatty protocols like CIFS and HTTPS, reducing the number of back-and-forth interactions over the WAN.
Object caching: When someone requests a discrete object, such as a Web page or CIFS file, it is stored locally so the next request does not have to travel over the WAN.
Byte caching: Byte patterns, such as icons that keep showing up in a block of data, get cached on the receiving end, so that the sending device need only transmit a 12-byte reference each time they occur.
Compression: Reduce the size of a data transmission by any of several methods, including removing all extra space characters, inserting a single character to indicate a string of repeated characters, and substituting smaller bit strings for frequently occurring characters.
The above techniques work in concert to boost overall application response time, according to Blue Coat spokesman Chris King: "So if a PowerPoint file is being sent, the SG box first checks the object cache to see if it's already been delivered. If it's there, but a currency check says 5% has changed, the box only sends that 5%, optimized and compressed."
Indeed, the most impressive test results occurred when a file or other object was being requested a second or third time, and caching as well as compression came into play, Warren says. Noncached transmissions saw definite but less dramatic improvements, he adds.
Around the same time as the Blue Coat deployment, Warren's team gave another boost to remote application performance by setting up a Multiprotocol Label Switching (MPLS) link between the new site in India and headquarters. Part of a company-wide MPLS rollout, the new connection cut round-trip latency by about a third, from 350 milliseconds over the older Internet connection to about 220 milliseconds, Warren says.
He adds it's still too early to quantify ROI. "We expect the main payback to be people working more efficiently in remote offices," Warren says. Lagging system response was definitely hurting productivity at remote sites, he adds. Furthermore, the accelerator "will allow remote offices to do things they couldn't before," by enabling them to improve response times and run applications that are inherently "not WAN friendly."
SNL has ordered a Blue Coat SG appliance for the Pakistani office. Warren says he's considering using SG to boost application speeds on SNL's domestic network as well. "Latency here is not nearly as bad, WAN traffic volume not as heavy, so we wouldn't get as much bang for the buck," he says. "But if we can save money on bandwidth, we'll do it."
Elisabeth Horwitt is a contributing writer based in Waban, Mass. She can be reached at firstname.lastname@example.org.
This was first published in November 2006