New enterprise applications are viewed as strategic and competitive advantages for businesses competing in today's ultra-competitive markets. Teams of personnel and millions of dollars are spent on internal facing and client facing applications such as CRM, ERP, etc. Financial institutions and other industries are investigating and deploying new technologies such as grid computing for the purpose of giving their business stronger and faster processing power as well as lowering costs through shared computer processing. Whether applications are internally developed or customized from existing commercial packages, the need to thoroughly test for proper operation of features and user volume usage has been embraced by the Enterprise IT community. Application rollout projects proceed through the following steps: initial development, unit test, integration test, system test, pilot program, and full production. The stakes and costs to fix defects get increasingly higher, as the project progresses towards full production.
Ultimately, the application is to be deployed to the production environment where a crucial piece is often overlooked: the impact of the network infrastructure characteristics on the application. Typical IT test environments are captive labs, usually relatively close in geography. Real networks, especially with the current trend of more work being sent off-shore and distributed networking, are global in their reach. Networking departments are well accustomed to monitoring and trouble shooting network conditions, yet applications that pass all lab tests are still at risk when deployed into the production network. Applications are at risk even when the network is healthy and passing traffic, simply because the network characteristics are different from the lab connectivity characteristics. PacketStorm Communications' network emulators reduce this deployment risk.
The figure above illustrates a simplified enterprise network. The headquarters, branch offices, and data centers all share the private corporate network. Telecommuters can access the corporate network through a VPN that can be accessed via dial-up, DSL, cable modem, or other access method. Enterprise applications can work flawlessly on the LAN, but can act very different when distant users must access them over a WAN.
Pre-Deployment Testing Using Actual Network Characteristics:
PacketStorm units have the capability to capture and save actual network impairment characteristics. This saved network profile can then be reused in your test lab. Using this procedure, application feature and load tests can be run over an accurate representation of real network characteristics. Application time out values, and other network affected parameters can be tuned prior to production rollout, saving time, money, and frustration.
Establishing Application Breaking Points:
Real networks are prone to impairments; delay, dropped packets, jitter, packet reordering, bit errors, etc. Higher level IP protocols such as TCP try to account for these conditions though the use of CRC checks and packet retransmission procedures. In many cases, these procedures for the recovery of corrupted data allow for acceptable use of the application. However, as conditions continue to degrade the network will have to increase the occurrences of retransmissions and users will experience delays. Other higher level protocols, such as UDP, lack error correction procedures. PacketStorm network emulators allow the user to create ideal network conditions and then incrementally and predictably worsen them. As the PacketStorm precisely controls these worsening conditions, a network conditions threshold can be determined for the application's acceptable operation. This information can be used to set alarm thresholds in network monitoring systems and determine appropriate SLA's with service providers.
Bandwidth Engineering and Network Topology Simulation:
Typical lab networks not only fail to incorporate real network characteristics and impairments, but also fail to emulate actual connection speeds and topologies. Production networks usually have relatively high core bandwidths, but remote branches, telecommuting employees, off site storage locations, etc can connect in a wide variety of ways such as T-1, FT-1, ISDN, 56 kb, or even lower rates. It is important to gauge application performance not only within the high speed core, but also to observe performance in real network topologies. PacketStorm network emulators allow for this type of testing. PacketStorm systems support the capability to emulate these edge access rates while still using standard Ethernet 10/100 or Gig-E interfaces. PacketStorm users can set transmission rates to match the access rates and observe how applications perform. These rates can be easily changed to allow the user to clearly understand what edge access rates are required. Therefore ensuring that minimum bandwidth requirements are met, yet not oversubscribing bandwidth and wasting money. In addition, hundreds of endpoint connections can be independently configured precisely recreating the make up of the traffic that transverses the core network. For example, a high speed link comprised of aggregated traffic coming from various endpoints traveling to different locations, each with its unique bandwidth availability and end to end characteristics and impairments. In addition to standard Ethernet interfaces, PacketStorm supports native WAN interfaces of T-1/E-1, T-3, OC-3, and OC-12. These interfaces allow the characteristics of the actual physical interface to be incorporated into a pre-production test. PacketStorm emulators accurately emulate the real network topology which enables application tuning to these conditions and ensuring a more confident deployment of these services.