Optimizing and accelerating traffic across the WAN no longer requires searching for exotic solutions -- there are a variety of appliances available from which admins may choose. Simply install an appliance on both ends of the WAN circuit and let them reduce network chatter, help mitigate the effects of latency, and make better overall use of the link.
This optimization lovefest breaks down, however, with mobile users and small offices. For obvious reasons, an appliance isn't appropriate for users on the go, and branch offices with a small number of users typically can't justify the cost of WAN acceleration hardware. The solution is to deploy a software-based optimization client to each user, whether mobile or not, and allow the soft client to help provide better overall performance and improve the end-user experience.
One such solution is Blue Coat's SG Client, currently based on version 5.2 of the SGOS. When paired with an SG appliance (I used an SG 200 during my tests), the SG Client provided good overall performance improvements over nonoptimized traffic. SG Client allows admins to define specific optimization policies for specific user needs but is available for Windows XP users only. Sorry, Windows 2000 and Vista not supported.
Of bits and bytes
I tested SG Client using the same test bed I used when testing Riverbed's Steelhead Mobile software client. I simulated two different WAN links in my Shunra VE WAN emulator -- one a cable modem connection (3Mbps/512Kbps, 65ms latency), and the other a DSL connection (1.5Mbps/386Kbps, 50ms latency) serving a 802.11b/g Wi-Fi hotspot. I used a Dell Optiplex running Windows XP Pro as my client PC, and I used Macro Scheduler from MJT Net to execute and time each test.
I found that SG Client reduced WAN traffic and transfer time in all test scenarios, with CIFS traffic showing the most improvement. A test case of copying many small files took 20 minutes to complete with no optimization, whereas a cold pass with SG Client copied the same files in just over 11 minutes -- roughly a 1.7X improvement. Another test involving opening, copying, and updating a file in Microsoft Excel showed the most improvement, reducing a nonoptimized time of nearly 20 minutes to a manageable 2 minutes on a first pass (9.6X increase). Although definitely an improvement over nonoptimized connections, SG Client didn't match Riverbed's Steelhead Mobile in overall performance in the same tests (3.9X and 16.2X respectively).
FTP traffic showed the least amount of improvement of all traffic types I tested, barely registering any improvement at all, regardless of whether the download was a first pass or subsequent "hot" pass. SG Client provides no specific optimization for FTP, and no object caching. This means that no matter how many times a file is copied using FTP, it is always treated as a first pass, with only generic TCP optimizations taking place.
Blue Coat is addressing this limitation in the next release of SG Client, due this spring. (Watch for our follow-up review when the new release ships.) GZip compression is applied across all traffic types to help squeeze data down to further reduce traffic, but in the case of my FTP test (an ISO disk image) there wasn't much room for compression.