 |
 |
 |
 |
| |
|||||
|
|
|
|
|
||||
Creating a Plug and Play Server Prototype Fills Technology Void William Wollman, Harry Jegers, Beth Loftus, and Caleb Wan 
s technology advances to the front line, it is putting increased capabilities into the hands of ordinary soldiers. Armed with wireless networks, powerful lightweight laptop servers, and handheld computers, a soldier can upload tangents for accurate bombing runs or download the latest satellite photos pinpointing the location of enemy units lying in wait. One drawback to this "network-centric warfare" technology is maintaining connections on the battlefield. Soldiers are moving continuously, and this rapid movement in current military operations has created challenges. When they set up computers on the battlefield, how do they find servers to connect to? What do they do if the server to which soldiers are connected goes down or if a shift in the battle line forces the soldiers to move out of range of that server? Reconnecting to the network can require the soldiers to know how to reconfigure the system or a network router. Off the battlefield, a network administrator would be called in to perform these tasks. The military cannot afford to train all its soldiers in network administration. A MITRE team began looking at state-of-the-art solutions to this problem. What if the soldier's computer and the network with which it was attempting to connect could automatically configure themselves? What if soldiers' computers were just like their radios or walkie-talkies: all they had to do was turn them on and they would work, even if the soldiers moved out of range of their servers or if their servers were knocked out of the networking system? The team focused on a capability called "plug and play" that would allow soldiers to easily and securely attach to a network to share communications anywhere and any time. Using plug and play technology, servers and clients can be enabled to securely affiliate with any command post's local wireless network access point with minimum planning. Once affiliated, basic network services can be quickly located by clients. The team saw this as a different way of looking at the connectivity problem, one that could benefit not just the military but other organizations looking for disaster recovery help. To thoroughly assess networking challenges, a prototype was created. We worked with a combination of open source and commercial router systems to demonstrate the benefits and feasibility of protocols. Through prototyping, MITRE is able to assess capabilities, further basic technology, or demonstrate something new to fill voids in the commercial market. This prototype, which has been demonstrated successfully, employs anycast, server load balancing, and global server load balancing. Our work demonstrated a solution that can increase our military's capability to quickly and easily deploy resilient network architectures. This work could be transitioned to any tactical command and control environment. The MITRE research team first looked into the technology available for this project. Options for data delivery included broadcast, multicast, unicast, and anycast. We chose anycast, which permits an address to be shared by a set of computers. Data sent to the anycast group will be delivered to at least one computer within the common set. Usually, the information being transmitted to an anycast address will be delivered to the closest anycast group member. The anycast address can be either a unicast network address or an application-specific address, such as a URL. Anycast techniques are being used by the Army and other organizations to solve challenges such as disaster recovery. Another key element is the server load balancer (SLB), which increases network resiliency by permitting multiple servers to be represented as one destination. Requests received by the SLB will be delivered to at least one of the computers within its server farm. The SLB can also be configured to monitor the "health" of each individual server and automatically remove any particular server that goes down. (A single server failure or server maintenance time period will not affect service availability.) An SLB also enhances network security by monitoring and controlling server access. Throughout the Internet today, SLB technology is used to support many server deployments. A third key element is the global server load balancer (GSLB). This enhances SLB by allowing multiple server farms to be separated and physically distributed throughout a network. Network survivability is enhanced through location diversity. The server farms maintain commonality by using a common domain name (e.g., www.anycast.org). By collecting and maintaining information about the health of the various server farms and network performance parameters, the GSLB can direct a client to a preferred server farm. Our work focused on deployment simplification of a resilient network architecture that uses anycast techniques created with SLB and GSLB. As part of this, we then developed another piece of the puzzle: an SLB Registration Protocol (SLBRP). The SLBRP automates the anycast process to permit servers to locally register their services with an SLB and to allow users to connect to a server without manually configuring a network or complex server networking. Once a server is registered, the SLB monitors the health of the service and provides continued access. When a server detaches from the network, the SLB automatically removes the server information from its configuration. Meanwhile, the GSLB works with the SLB to direct users to the "best" server. Agents measure server response time and network performance parameters and provide this information to the GSLBs. When a server farm is removed from service, the SLB will de-register the farm and associated domain from the GSLB. The MITRE team considered building a detailed simulation to demonstrate to the military how our plug and play system would operate. We decided it would be more effective, however, to build a prototype of the SLBRP and the plug and play anycast environment. With a prototype we could demonstrate that such a system was possible, not only for a proprietary commercial system, but also for an open source system. When considering commercial systems for prototyping, we chose an experimental version of the Cisco IOS that supports Java with an embedded Java Virtual Machine. This version allowed us to enable plug and play SLB networking with Cisco routers. We also leveraged the Cisco Service Assurance Agent technology for our network performance monitoring agents. In addition, we used the Linux Virtual Server and Zebra routing open source software for the project. The prototyping process successfully demonstrated the benefits and feasibility of the SLBRP used with a combination of open-source and commercial router systems. We have demonstrated to the Army and Navy how the prototype provides mobile server support for anycast, plug and play server load balancing, and intelligent server selection via dynamic global server load balancing. The SLBRP is a simple and efficient way to automate anycasting addressing. By leveraging plug and play SLB configuration concepts, we significantly helped to reduce the complexity associated with both server deployment and mobility in the battlefield. The concepts can also be used to enhance network configuration management within any organization. Our work in this area points the way for the evolution of network-centric warfare. In the future, soldiers without any specialized computer training will be able to bring the full force of the military's information system to bear at every point on the battlefield. Research and development work, including prototyping, allows MITRE to maintain our technical excellence while providing capabilities to the industrial community to advance particular technologies (in this case, networking). Our sponsors benefit through having more capable commercial products in the long run and through MITRE's continuing ability to guide industry toward capabilities that will solve sponsors' critical problems. |
|||||
| For more information, please contact William Wollman, Harry Jegers, Beth Loftus, or Caleb Wan using the employee directory. Page last updated: January 7, 2005 | Top of page |
|||||
Solutions That Make a Difference.® |
|
|
Download this issue [1.9 MB]