Network admins have spent many a late night trying to figure out how to improve application response or file replication across the WAN. Faster performance is all about bigger pipes, right?
Not so fast, Mr. Bandwidth. More capacity may help, but the real solution is to reduce the effects of latency, minimize application chattiness, and move data as close to the end-user as possible.
Packeteer has been a longtime player in the WAN optimization space, and the company's newest release combines two of its most popular technologies: WAN optimization and traffic shaping. The iShaper blends the traffic shaping capability of the PacketShaper appliance with the WAN optimization and acceleration features of the iShared appliance. Rolled into a 3U appliance behind a unified user interface, iShaper gives network admins a single point of management for managing and accelerating WAN traffic. My test appliances ran version 8.25g1std of the PacketShaper software, and version 220.127.116.11 of the iShared software.
Two for one
I installed a pair of iShapers on my tried and true WAN test bench and ran a lot of traffic through them during the course of the evaluation. Unlike most other WAN acceleration appliances, iShaper does not drop transparently into the network but instead routes between LAN and WAN interfaces. This is merely a minor disruption in the network scheme, but can cause some reconfiguration issues when deploying, such as changes to existing IP addressing. Also, because of the combination technologies, setup and configuration are slightly more involved than what I experienced with a Riverbed or Blue Coat installation. All in all, it took less than an hour to have both core and branch office appliances configured and optimizing traffic.
Performance increases over nonoptimized traffic were among the best ever recorded for CIFS traffic, edging out Riverbed in both my "many small" and "single large" file transfer tests. iShaper also scored best in my Excel four-step test, which opens an Excel spreadsheet, saves it, reopens it, and saves it with a different name back to the core. Many of the performance gains are a direct result of the underlying architecture of iShaper. Built on Windows Storage Server 2003 R2, iShaper makes full use of Windows' DFS (Distributed File System) replication and takes full advantage of being part of a native CIFS filer.
FTP traffic is better in this release but still not up to the results posted by Riverbed. Compared to the results of last year's iShared test, iShaper has cut Packeteer's FTP times by more than half for both 128Kbps and T1 links. What is odd is that FTP performance is slower than CIFS performance using the very same test files and link speeds. A user would be better off copying the file via CIFS than FTP. The MAPI save attachment test also cut its time from last release in half, a very nice increase in speed. But as with FTP, Steelhead just edges out iShaper in MAPI performance.
Unlike Riverbed's byte- or segment-level caching, which is shared across all traffic, Packeteer's Wide Dictionary Compression (WDC) is isolated by traffic type. For example, if I copy a file using CIFS or NFS across the WAN, and then another user FTPs the same file via the same WAN circuit, the FTP does not receive any benefit of my previous file copy. Only on subsequent FTP copies will WDC come into play. Packeteer says this will change in the iShaper 4.0 release, due in Q3 2008, which will share byte-level caching across all traffic types.
Recently, both Cisco and Riverbed announced plans to host third-party applications on their appliances. Because of iShaper's Windows Storage Server foundation, it is already poised to provide branch offices with additional services without additional overhead. Services such as local file storage and application and printer sharing are all available as they would be on a Windows server.