The International Center for Advanced Internet Research (iCAIR) and iGRID2002 The iGrid ("the International Grid") is a biennial international showcase demonstrating multiple leading-edge next generation applications are demonstrated every two years as part of - showcase. These iGRID demonstrations are organized by the Electronic Visualization Lab at the University of Illinois at Chicago, with other partner institutions. iCAIR is one of the organizing institutions for iGRID2002, which takes place in Amsterdam in September. (ref: www.startap.net/igrid2002) The goal of iGRID is to showcase the evolution and importance of global research community networking. iGrid highlights achievements in grid architecture development and the advancements enabled in science, engineering, cultural heritage, distance education, media communications, and art and architecture. During the last iGRID (iGRID 2000) in Yokahama, Japan, 14 countries participated: Canada, CERN, Germany, Greece, Japan, Korea, Mexico, The Netherlands, Singapore, Spain, Sweden, Taiwan, United Kingdom, USA. At iGRID2000, 24 demonstrations were shown, featuring technical innovations and application advancements requiring teleimmersion, large datasets, distributed computing, remote instrumentation, collaboration, human/computer interfaces, streaming media, digital video and high-definition television.
Hosted by WTCW (Amsterdam Science & Technology Centre), the iGrid2002 conference (23-26 September 2002) is focusing on e-science, Grid and Virtual Laboratory applications enabled by high-performance global networks. The themes of the iGrid2002 symposia (25 & 26 September 2002) are 'International Virtual Laboratory' and 'Visualization and Virtual Reality'. iGrid presents the latest developments in these areas. The iGrid2002 conference will present in-depth workshops, demo presentations, tutorials and symposia. iGrid, however, specializes in presenting what these advanced technologies can actually do as opposed to only discussion them. More than 28 scientific groups from around the world are expected in Amsterdam, where they will showcase many different application advancements and middleware innovations enabled by globally connected high-performance networks. The iGrid2002 conference is also a meeting place for people from the scientific and business community, to meet and discuss future opportunities.
Enabling Next Generation Applications Through Intelligent Optical Networking
The International Center for Advanced Internet Research (iCAIR) and its research partners have undertaken a number of projects directed at creating new advanced digital communication services based on intelligent next generation optical networks. These new services have a potential to significantly enable a wide range of innovative, powerful advanced applications. Almost all of these applications are data and computationally intensive, and most are based on large-scale, even world-wide, computational Grids, e.g., the "Global Lambda Grid," which is a concept being formulated by the StarLight community. The applications that will be supported by the Global Grid include those that are related to high performance computational scientific research, engineering, bioinformatics, materials sciences, data streaming, digital media, data mining, and astrophysics. Computational Grids are part of a next generation "cyberinfrastructure" used for extremely large scale, resource intensive applications. Some emerging Grid infrastructures, such as the TeraGrid, use networks not for standard communications support but as backplanes for high performance computational clusters, comprised of hundreds or thousands of individual compute nodes within widely distributed clusters.
At iGRID2002, iCAIR and its partner research organizations are demonstrating Photonic-Empowered Applications, based on next generation intelligent optical networking technology. These innovative networking technologies are being developed to allow for multiple, high performance global applications. The majority of these technologies are being demonstrated as research works-in-progress - not commercial products.
Traditionally, the design and implementation of network based applications, especially large-scale, high performance applications, have had to be compromised across multiple dimensions - interfaces, services, performance, flexibility, protocols, architecture, technology, etc. These restrictions exist primarily because the most widely deployed communications infrastructure was designed to optimize analog, not data, communications. The iGRID2002 demonstrations being presented by iCAIR and its research partners illustrate the types of applications - Photonic-Empowered Applications - that are possible when such restrictions are removed - applications based on a next generation communications infrastructure optimized for data.
The term "Photonic-Empowered Applications" refers to multiple, global, next-generation applications that are being designed and developed as a) highly distributed (including those based on resources at sites world-wide) b) resource intensive - e.g., computationally intensive, bandwidth intensive, storage system intensive, et al. (many are based on grid computing infrastructures and c) highly asymmetric. However, in addition, they are distinguished also by their d) utilization of advanced data communications - based on dynamic multi-wavelength lightpath provisioning, i.e., supported by more flexible DWDM-based networking technology than is implemented in today's point-to-point optical networks. They are also optical network "aware" - that is, they have a capability for directly discovering and signaling for the networking resources that they require, including signaling for the provisioning of lightpaths. In addition, some of these types of applications may be highly periodic and transient (e.g., they may exist only for a few moments at different times throughout a month or throughout a day). Consequently, they may transition instantaneously from little or no network utilization to requiring enormous network resources for days, hours, minutes, or moments, or even milliseconds. Also, many of these types of applications require a much closer integration of such resources than are currently available through existing information technology infrastructure, which tends to distinctly segment system components. Within the emerging new infrastructure environments, the boundaries between applications, computers, and networks truly dissolve. This architectural direction is inherent within leading-edge information technology development projects, such as those focused on Grid computing, Globus middleware, and new experimental (non-production) research networks.
Although these types of application characteristics may seem unusual, they actually reflect real world activities. If they seem extraordinary, it may be because applications are so frequently severely compromised by the various restrictions supporting digital infrastructure. Currently, the development of many large-scale global applications is restricted by the design of the most widely implemented communication infrastructure. One simple specific example (of many that could be cited) of this type of restriction is that visualizing dynamic high definition 3D models and simulations is a critically important tool in many organizations. Yet, very few public, or even private venues exist that provide capabilities for displaying these visualizations. Consequently, addressing the resource needs of these next generation applications requires a revolutionary new approach to data communications.
Photonic-Empowered Applications at iGRID2002
At iGRID2002, Photonic-Empowered Applications are being demonstrated to illustrate the enabling power of supporting high performance global applications with next generation wave-length-based networking, which includes integrating those applications with the optical network control plane. These developments have been undertaken primarily in response to the needs of multiple large-scale applications. These applications are "empowered" because they a) are network aware; b) are unrestricted by traditional legacy network architecture that limit optimizing for data flows; c) have access to new types of intelligent application signaling methods, i.e., provided with "intelligence" through control capabilities provided by a series of optical networking service layers; d) can use those service layers to access next generation control plane and transport plane techniques, including technologies that allowing for application-controlled dynamic lambda provisioning and switching; e) are based on a foundation of next generation multiwavelength network technologies; and, f) are provided with multiple sophisticated L1 process monitors and automatic adjustment mechanisms.
The Photonic TeraStream - A Prototype Photonic-Empowered Application
The prototype Photonic TeraStream demonstration application at iGRID2002 was developed by iCAIR to illustrate the capabilities and technologies required by next generation large-scale global applications, e.g., high performance data streaming for e-Science High Performance Data Streaming (e.g., Computational Genomics, APS experimentation, medical imaging, etc), global enterprise applications, new types of composite applications. Such new applications could include "Global Services-on-Demand." For example, iCAIR, in partnership with the Materials Sciences Research Center at Northwestern University is developing an International Virtual Institute for Materials Science, which is being funded by the National Science Foundation.
The IVI requires a global capability for a type of high performance application that instantaneously discovers, gathers, integrates, and presents for a global audience information technology resources from throughout the world - including those comprised of multiple data types.
These resources include large-scale data streams from experimental repositories at remote locations, scientific visualizations and digital media, and computational processes. Consequently, the Photonic TeraStream application at iGRID2002 is a "composite application" - integrating several component applications. The Photonic TeraStream demonstrates the potential for creating global applications that do not have to be compromised because of the data communication infrastructure restrictions. Photonic TeraStream is a composite demonstration application linking, high performance data transfer (based on Grid FTP), digital media streaming (270 Mbps encoding), high performance remote data access methods (based on iSCSI), and dynamic provisioning. The Photonic TeraStream is supported by an innovative architectural method that envisions a) closer integration between applications and advanced networks and b) liberating applications from multiple restrictions inherent in today's communication infrastructure. Key technologies that comprise the demonstration are a) intelligent application signaling, b) dynamic lambda provisioning on the OMNInet testbed, c) extensions to lightpaths through dynamically provisioned L2 and L3 configurations, in part, to allow for access to multiple types of edge resources. The iCAIR iGRID2002 demonstration shows how such applications can use intelligent signaling to provision their own lightpaths to optimize network based resource discovery and performance, for example, to enable access to and dynamic interaction with very large amounts of distributed data. Lightwave-based applications supported by dynamic lambda-switching provide for significantly more powerful capabilities than those based on today's communication infrastructure.
Many of the technologies related to the Photonic-Empowered Application demonstrations at iGRID2002 have been developed on OMNInet, a metro area photonic networking testbed in Chicago. OMNInet has been established, in part, to create a reference model for next generation optical metro networks (www.icair.org/omninet). At iGRID2002, this testbed is being extended to Amsterdam through StarLight and NetherLight in order to demonstrate that Photonic Enabled Applications are possible not only on next generation optical metro area networks (such as OMNInet), but can also be extended to global networks, e.g., through StarLight and NetherLight, as a basis for the global lambda Grid. StarLight is a next-generation global optical networking exchange facility in Chicago (www.startap.net/starlight), and NetherLight is a trans-Atlantic high performance link connecting StarLight to SURFnet in Amsterdam. NetherLight is a facility devoted to research, that is, experimental, not production, networking. (Ref: Photonic TeraStream Diagram). The foundation infrastructure for these next-generation applications consists of advanced optical networks which are currently emerging as experimental networks that are by design not production networks. They are leading-edge, innovative networks designed as a means to create a wide range of advanced networking technology. (Photonic TeraStream Collaborators: Joe Mambretti, Jeremy Weinberger, Jim Chen, Elizabeth Bacon, Fei Yeh, David Lillethun, iCAIR (www.icair.org), NU, USA, Cees de Laat, Leon Gommans, Bas v.Oudenaarde, Bert Andree, University of Amsterdam, The Netherlands (www.uva.nl). Robert Chang, Materials Sciences Research Center, NU USA, Jason Leigh, Oliver Yu, Electronic Visualization Laboratory, Bob Grossman, National Center for Data Mining, University of Illinois at Chicago, USA, Cees de Laat, Leon Gommans, Bas v. Oudenaarde, Bert Andree,University of Amsterdam, The Netherlands, Linda Winkler, Bill Nickless, Caren Litvanyi, Argonne National Laboratory
Acknowledgements: Hal Edwards, Paul Daspit (Nortel, www.nortel.com), Teresa Elliott, Rachel Alarcon (SBC/Ameritech, www.sbc.com), National Science Foundation (www.nsf.gov, #ANI-0123399 et al)
Also, a number of the research projects noted here are being supported by the National Science Foundation, including those related to the Intelligent Application Signaling project, established as a joint research partnership with the Electronic Visualization Laboratory of the University of Illinois, described in one of the final sections of this iGRID2002 overview.
Intelligent Application Signaling and TeraAPI - a UNI for Photonic-Empowered Applications
Before applications can become "empowered," they must be "network intelligent" - they must have an understanding of available network resources, how to access and use those resources and how to adjust to dynamic changes in those resources. This type of mechanism is required because these applications must be much more "network aware" especially about changing network and edge resource dynamics. Consequently, iCAIR is involved in several projects that are developing specialized high performance (including e-science) application signaling methods that will allow very-large-scale distributed applications to directly manipulate a wide range of optical networking functions.
A simple application protocol for making requests of those service layers has been developed. At iGRID2002, Photonic TeraStream application will interact with a client (odin_cli) that is serving as the signaling method for requests from lower-layer network service layers. The odin_cli can also serve for demonstration purposes as an application proxy. In addition, a preliminary architectural model, the TeraAPI has designed, and has been developed (in C) as a prototype API iCAIR. This API is being designed as a mechanism for network resource discovery and high level (i.e., application to network service layer) intelligent application signaling for core resources, e.g., dynamic lightpath management, and for adaptive application responses to dynamic network conditions. The term "Tera" is used to indicate that the architecture is being designed to scale to terascale infrastructure - e.g., in a networking context, terabits per second.
The TeraAPI will serve as the interface between the application, e.g., Photonic TeraStream, and low level optical networking resources, through additional service intermediaries currently in development at iCAIR, such as a) the Optical Dynamic Intelligent Network (ODIN) service layer, b) the TeraScale High Performance Optical Resource-Regulator (THOR), which manages the optical network control plane and resource provisioning, including dynamic provisioning, deletion, and attribute setting of lightpaths, and c) the Dynamic Ethernet Intelligent Transit Interface (DEITI), which has a capability extending lightpaths beyond wavelengths to other L2 links, currently GE links, for example, to allow the Photonic TeraStream to access edge resources such as compute clusters and data storage repositories. These links can be provisioned as GE vLANs.
In addition, the TeraAPI must interact with a range of network middleware functions, described below. There are multiple ways that such an interface can be utilized. For example, two general models are a) the application requirement parameters can be signaled into the network at each instantiation requiring recourses, setting off a sequence of resource discovery and utilization processes or b) the application can signal a specific call including a set of known parameters that would evoke the utilization of a pre-set "package" of resources, including all required resources and linkpaths that allow access to those resources. The iCAIR presentations at iGRID2002 demonstrate the utility of this latter approach.
Optical Dynamic Intelligent Network (ODIN) Services
The demonstrations being shown at iGRID2002 envision large-scale, powerful, high performance Photonic-Empowered applications completely provisioning and controlling global, dynamically provisioned lightpaths, which could be provisioned as global dynamic VPNs (GDVPNs) - for example, as a basis for the Global Grid. The Photonic TeraStream prototype application is based on new signaling methods for not only for dynamic lightpath provisioning, which could be used to create optical VPNs (OVPNs), but also for extending those lightpaths to edge resources through other types of dynamically provisioned L2 GE links, including complete Ethernet vLANs.
The Photonic TeraStream utilizes the TeraAPI to access the Optical Dynamic Intelligent Network (ODIN) services module. The ODIN service layer is software being designed and developed at iCAIR as an intermediary between high performance distributed global applications and lower level network service layers. Collectively, these service layers allow for dynamic, integrated coordination between the Photonic-Empowered Application that may reside on a client network with various processes and resources at the optical network layer. The other service layers (control processes) include the TeraScale High Performance Optical Resource-Regulator (THOR) and the Dynamic Ethernet Intelligent Transit Interface (DEITI). The ODIN services layer enables high performance applications while providing for network transparency, allowing the applications to utilize the full power of photonic networks without having to deal with their complexity. ODIN is a powerful mechanism that can be used to establish "services on demand" and, as noted, not only dynamically allocated lightpaths, but also dynamically allocated transient (or permanent) Optical Virtual Private Network (OVPNs). In part, ODIN accomplishes its functions through interactions with lower layer optical services such as the TeraScale High-Performance Optical Resource-Regulator (THOR).
ODIN provides a single point of control for a defined set of network service requests within a single administrative domain. This point of control is incorporated within a process that is on a control server. The process has total understanding of the topology and current resource allocations within the administrative domain. ODIN accepts requests for resource allocations from applications over the network. The process listens on a TCP socket for requests from applications, and responds to those requests over a connected session linked to that application. When resources are allocated to fulfill those requests it communications with the requisite network switches to configure them to meet those application requests. These switches can be optical-domain DWDM switches, Ethernet switches and/or IP routers.
In sum, ODIN is server software that is comprised of components that a) accept requests from clients for resources (the client request a resource, i.e., implying a request for a path to the resource - the specific path need not be known to the client), b) determines an available path - possibly an optimal path if there are multiple available paths), c) creates the mechanisms required to route the data traffic over the defined optimal path (virtual network) d) notifies the client and the target resource to configure themselves for the configured virtual network (ODIN returns a new IP and subnet mask in response to a resource request). In order to create virtual networks, ODIN uses two other services, or control processes, THOR, the Lambda god, which creates the Optical vLAN (OVPN) and DEITI, which provides for L2 extensions across Ethernet links at the network edge.
TeraScale High-Performance Optical Resource-Regulator (THOR - the "Lambda God")
The TeraScale High-Performance Optical Resource-Regulator (THOR) does the primary work of the dynamic lambda provisioning required by the Photonic TeraStream. Just as Thor (the Norse god of lightning and thunder) cast dramatic bolts of light, the Lambda god is an all powerful controller of the lightpaths in the optical network. The Lambda god is an optical route allocation and management system. THOR establishes and deletes lambda-switched paths, i.e., lightpaths, based on an understanding of a) the application requirements (e.g., the Photonic TeraStream); b) physical optical network topology; c) potential capabilities for resource allocations within that topology and; d) performance optimization. THOR components include a) a mechanism for receiving requests; b) a mechanism for fulfilling requests, such as allocating and managing network resources (e.g., routes); c) a mechanism for monitoring state information such as route configuration data. After receiving a client (e.g., application) request(s) through ODIN, THOR a) determines the state of the lambda-based lightpaths (lambda-switched paths) in the optical network; b) determines the most optimal lightpath(s) for a particular request c) creates a lightpath; or a an OVPN, by configuring the photonic node switches d) notifies the client and the target resource to configure themselves for use of the OPVN; and e) reallocates optical resources when they are no longer being used.
THOR has an understanding of the network configuration to such a degree that it can allocate lambda switched paths (resources at the level of multiple Gbps), but also provide for lambda resource sharing (i.e., multiple paths on a single lambda).
THOR is a mechanism for interfacing to an optical network control plane, in this case, the OMNInet control plane, which is using a signaled overlay architecture. DWDM is migrating from a technology used primarily for point-to-point provisioning to one that can be used for dynamic wavelength (lambda) switching and service provisioning. THOR is a mechanism to allow for enhanced utilization of these capabilities. For example, THOR interlinks the resource requests of the Photonic TeraStream with lower level API software that is linked to controller-based intelligent photonic protocols, which are used for configuration and auto discovery. THOR controls the DWDM layer by direct calls to a UNI API that Nortel Research Labs developed for OMNInet. These control tools are based on GMPLS, Generalized Multiprotocol Label Switching, an emerging IETF standard. However, this particular impementation has additional optimization components.
THOR and Control Plane Architecture for Next Generation Optical Networks
There are four primary architectural models for optical network control planes (with many hybrids and variations), overlay, signaled overlay, peering, and integrated. The overlay model has a separate control planes, for example, for both a client network, e.g., an IP network, and for its underlying supporting optical network, which is connected to the client network via a UNI. OMNInet is based on signaled overlay architecture, which allows for client networks to signal into the optical network to allow for dynamic services and resources provisioning. On the OMNInet testbed, this type of provisioning is being accomplished directly by applications. In other words, network providers need not be involved in this type of instantaneous provisioning, allowing for much faster services deployment. THOR is the interface between the ODIN service layer and the OMNInet control plane.
OMNInet is provisioned with a GMPLS-based intelligent optical control plane, with optical UNI and other interfaces (CLI, GUI, general API), a data plane control interface and a routing API. GMPLS, an IETF standard, is being used for resource discovery, link provisioning, label switched path creation, deletion, and property definition, traffic engineering, routing, channel signaling, and path protection and recovery. The GMPLS implementation being used was designed and developed by Nortel Research Labs.
OMNI and OMNInet
The Optical Metro Network Initiative (OMNI) OMNInet is a metropolitan optical network testbed currently being deployed in Chicago and Evanston, Illinois. This project is a joint partnership among SBC, Nortel, iCAIR, the Electronic Visualization Lab of UIC, the Math and Computer Science Division at Argonne National lab, and CANARIE, the Canadian national research network. The OMNInet testbed is being designed to optimize for metro area data services, which has several implications. One of these implications of this optimization approach is that there are no SONET components. (SONET-based networks are optimized for traditional communication services not data communications.) Another is that the types of networking technologies being designed and implemented for this testbed are those that specifically address metro area vs long haul network requirements. Also, this testbed has been designed with a focus on optimizing for highly asymmetric, high performance data communication services.
OMNInet is an integrated network system with three primary functions: 1) Unified IP routing, using an implementation of an IETF standard - Generalized Multi-Protocol Label Switching (GMPLS) 2) An optical networking system, based on Micro-Electro-Mechanical Systems (MEMS), that supports lambda switching and 3) A cross-layer network intelligence, including at the Dense Wave Division Multiplexing (DWDM) level. The physical core of OMNInet is a metro-Area dark fiber fabric that has been qualified for multiwavelength-based services.
OMNInet's initial testbed is a metropolitan-area, four-node, optical network linking a core node on Northwestern University's (NU) Chicago campus (just northeast of Chicago's downtown "Loop" area), with nodes at the University of Illinois at Chicago (UIC) (west of the Loop), at the CANARIE CA*net3 node at its Chicago Point of Presence (PoP) (just south of the Loop), and at a node on NU's Evanston campus (20 miles north of the Loop). This deployment constitute a very wide area testbed sufficient to conduct advanced experimentation in next generation phonic based networking.
Each node includes a Wave Division Multiplex (WDM) photonic switch, an Optical Fiber Amplifier (OFA), optical transponders/receivers (OTRs), and high-performance L2/L3 router/switches. These nodes were designed and developed by Nortel Research Labs. (The core nodes are not products but ratherinovative research concepts, custom built by Nortel Research Labs.) The photonic switches are supported by Optera 5200 OFAs to compensate for link and switch dB loss with 8 ports capable of supporting 10Gbps optics. Application cluster and compute node access is being provided at each location by PassPort 8600 L2/L3 switches, which are provisioned with 10/100/1000 Ethernet user ports, and 1GigE 1550XD trunks (i.e., dedicated fiber that supports 1550nm cross-connect trunks. These sites are connected by dedicated technology trial-fiber services provided by SBC/Ameritech.
Out-Of-Band Management Layer
Currently, a separate OMNInet control plane has been provisioned out-of-band using completely separate fiber, provided through the Metropolitan Research and Education network (MREN). At some point in the future, this control plane will be migrated to a supervisory wavelength. This control plane enables User-to-Network Interface (UNI) control signaling via a UNI interface to the optical transport network and bi-directional signaling to the connection control plane. 10GigE trunk interfaces, using true 1550nm 10GigE, have been implemented, with a specialized set of protocols that allows for enhanced optical network intelligence, including a wavelength signaling protocol, a wavelength routing protocol, and an optical link management protocol. Optical wavelengths are being used to implement virtual lightpaths to transport data streams (relevant to scientific applications) with specific service levels. In part, control plane software is being used to segment aggregated and individual virtual paths in accordance with the characteristics and requirements of specific applications, including those related to traffic classes, security requirements, latency sensitivity, etc.
Physical Layer Monitoring and Adjustment
OMNInet is supported at L1 by state-of-the-art tunable lasers, amplifiers, and filters, which are part os subsystems and systems developed by Nortel Research Labs. However, to provide for reliability and optimal L1 performance, OMNInet is provisioned with a wide-range of sophisticated pre-fault detection mechanisms, also developed by Nortel Research Labs, which monitor network conditions and adjust resources in response to specific detected characteristics. These types of critical components are key elements to next generation photonic-based networks. The photonic core of OMNInet is interconnected by multiple fiber strands, dedicated to the testbed, fully qualified for multiwavelength based services. The fiber provisioning for OMNInet is being undertaken by SBC, as part of the technology trial.
Photonic-Empowered Data Mining, National Center for Data Mining at UIC and OMNInet
The National Center for Data Mining at UIC (ref: www.ncdm.uic.edu) is working with iCAIR on linking various applications to methods for interlinking next-generation data-mining techniques with the methods being developed on OMNInet for dynamic lightwave provisioning. For the last several months, in part, to prepare for iGRID2002, researchers at iCAIR and NCDM have also been running a number of tests to ensure optimal performance of TCP and UDP, including testing using stripped TCP. Researchers at the NCDM have been using OMNInet to test protocols that they developed to allow for the design of network based applications with reliable end-to-end performance and speeds that scales to multiple-Gbps. These protocols include PSockets and SABUL, which are open source libraries to build network applications with advanced functionality. NCDM's SABUL is an innovative protocol that using UDP as a transit protocol but provides for reliability by using TCP as a control protocol. At iGRID2002, the National Center for Data Mining is presenting a number of Photonic-Enabled applications.
Photonic-Empowered Applications and Middleware
iCAIR develops architecture within the framework of standards bodies. iCAIR is also a member of the Global Grid Forum and Globus communities, and iCAIR researchers closely follow (and have contributed to) the efforts of the Internet Engineering Task Force (IETF), the standards organization for the Internet. However, the IETF has rarely addressed the issue of standard mechanisms for application signaling to network fabrics to allow for general provisioning much less large scale dynamic resource provisioning. In 1998, iCAIR and its partner organizations held a National Science Foundation workshop on Middleware to identify existing middleware services that could be leveraged for new capabilities as well as identifying additional middleware services requiring research and development. This workshop resulted in an IETF RFC - 2768, which describes Middleware as a "reusable, expandable set of services and functions that are commonly needed by many applications to function well in a networked environment."
Middleware components include APIs, authentication, authorization, and accounting (AAA) issues policy framework, directories, resource management, networked information discovery and retrieval services, quality of service, security, and operational tools. This type of middleware framework can serve as useful context to identify various components, that require integration into an application signaling architecture, including signaling methods, access/admission controls, and a series of defined services and related resources, management of service levels and priority attributes, scheduling, service attribute setting functions, a feedback mechanisms for notifying applications or systems about performance variations, etc. These types of mechanisms imply various capabilities for: 1) an interaction with some type of policy implementation and enforcement, 2) dynamic assessment of available network resources, 3) policy monitoring, 4) service guarantees, 5) conflict resolution, and 6) restitution for lack of performance and/or fulfillment.
Dynamic Ethernet Intelligent Transit Interface (DEITI)
Applications such as the Photonic TeraStream will have to access resources that are not linked to wavelengths. The OMNInet research team recognizes that it is difficult to deploy lightpaths over wavelengths to all areas where resources must be accessed, especially at the edge of the network. However, to optimize performance, it is necessary to provide a means to instantaneously provision L2 transit paths while also avoiding the performance degradation that would result by introducing unnecessary L3 packet inspection processes. One potential solution would be to implement high performance MPLS. However, currently few edge routers are MPLS enabled. Another potential solution is to utilize dynamic L2 provisioning through Ethernet vLANs. Consequently, iCAIR has established a project to design an L2 extension to its ODIN service layer beyond lightpath provisioning. This technology - Dynamic Ethernet Intelligent Transit Interface (DEITI) allows for extending optical resource provisioning to dynamic vLANs. To some degree, this capability can be considered a proxy or substitute for MPLS. However, it may prove useful over the longer term in many areas where MPLS is not implemented.
vLANs and (I)AAA
At iGRID, the University of Netherlands research contingent to this project are demonstrating a technology that can also include specific separate considerations of Identification - thus - AAA).
They are demonstrating an Ethernet based switch setup allowing a 802.1Q VLAN configuration. For this demonstration, modular switches are interconnected using a 1000BaseSX connection. Each switch supports 16 FE connections into the GbE uplink. Four stations (each with dual FE interfaces) are connected in pairs to the two switches. One interface of a station connects to a switch, the other to a common network. A 5th station is used as AAA server that controls both switches using SNMP and the 802.1Q bridge mib extentions. With this implementation, a station can send an XML/SOAP based connection request to the AAA server, that then obtains a policy that determines if the request can be honored or not. If the request is granted, a bypass connection is created via the VLAN switches using the 2nd ethernet interface for a specified amount of time.
The iGrid demonstration shows the setup, the request, the policy and its execution. This principle can be used to create connections on demand that bypasses the regular Internet for high-performance data intensive applications (that would justify this kind of implementation). This demonstration is good example of the application of the principles of Generic AAA (IETF RFC2904).
Application of Intelligent Signaling and Control of Dynamically Switched Networks
Some of the research being conducted on intelligent application signaling has been undertaken by iCAIR as a joint project with the Electronic Visualization Lab of the University of Illinois at Chicago (www.evl.uic.edu). Collaborators on this project include Tom DeFanti, Maxine Brown, Jason Leigh, and Oliver Yu. This project, "Application of Intelligent Signaling and Control of Dynamically Switched Networks" is being funded by the National Science Foundation. The project is using OMNInet to test methods for application level dynamic control of resource discovery, allocation and adjustment to allow more flexibility in service provisioning, infrastructure deployment and service resource management. The research components include: a) research into the behavior of advanced science applications, not just on an extremely high performance optical network, but on one that can be dynamically adjusted on a granular level; b) identification of application-level network requirements; c) investigation of management techniques for optical networks and of new service provisioning models; d) research into new methods for application signaling; e) investigation of interconnections between application signaling and IP-based control planes, including with GMPLS; f) test deployment of these techniques on an advanced optical testbed; g) analysis of results; h) experimentation with multi-service provisioning to ensure gateways to traditional networks and protocols; i) development of a system for performance metrics, monitoring and analysis; j) creation of a testbed for StarLight, the next generation, optically-based Science Technology and Research Transit Access Point (STAR TAP) and for other advanced research networks.
About iCAIR Mission & Projects Announcements
Corporate & Research Partners Advanced Internet Initiatives Other URLs Home