194 Chapter 10: Grid Networks and Layer 3 Services Some experimental results indicate that the insecurity of Cipher Block Chaining (CBC) – a popular technique to extend ciphers beyond block size – increases as Os 2 /2 n , where n is the block size in bits, and s is the number of blocks encrypted [20]. Intuitively, this insecurity results from the dependency of the ciphertext of one block upon both the plain text for that block and the ciphertext of the preceding block. Should two blocks yield the same ciphertext, and the plaintexts of the next block are also identical, then the ciphertexts of the next block will be identical. This situation creates a vulnerability to certain types of malicious attacks. As a matter of common practice, a rekeying event should occur any time B bytes have been sent through an encrypted security association whose crypto-transform uses a CBC mode of operation (e.g., 3DES) [21]. This event sets B to n/8 ∗ 2 n/2 , wherein n is the block size in bits. A security association that uses 3DES n = 64 at 1 Gbps requires a keying event every 274.9 seconds. At 10 Gbps, it requires a keying event every 27.5 seconds. As speeds increase, designers will need to take into account the challenge of executing a rekeying event at shorter intervals. It should be noted that a keying event typically requires a very large integer exponentiation, which is a very demanding challenge when compared with ordinary message crypto-processing. 10.10 IP MULTICAST The IP multicast extensions [22] were introduced to relieve the network from forwarding as many copies of a set of data as there are receivers. Furthermore, its receiver-driven style of operation is meant to relieve the sender (or publisher) from tracking the subscribers to the data. In general, the promise to mitigate traffic volumes and complexity is likely to appeal to Grid communities, especially when multicast aligns with a “push” style of data diffusion in the Grid. In practice, however, few networks have IP multicast enabled on the scale that would be significant to Grids. The difficulties in policy and security enforcement, in scaling reliable services above IP multicast [23], in cross-provider support, and providers’ additional operating costs in running extra control plane software have greatly hindered a widespread roll-out of IP multicast services. An informational document by the GGF surveys the IP multicast landscape in greater detail [24]. On a confined footprint, IP multicast has proved to be an important enabler of group communication for the increased reliability of systems [25,26]. As Grids mature and the expectations of their being dependable grow, it is expected that IP multicast will continue to play an important role at the inter-component level (and replicas thereof). Chapter 3 has substantiated the case for profound architectural flexibility in the Grid infrastructure (whether it is general or network specific). More than ever, Grid designers are empowered to specialize into realizations of multicast semantics at the layer 7 (e.g., application-level multicast) and/or layer 1 (e.g., optical multicast). 10.11 INTERNET LAYER 3 SERVICES This chapter presents only a small number of topics related to Internet layer 3 proto- cols and services. One point emphasized in this chapter is that these protocols were References 195 designed specifically to support services in dynamic environments. Consequently, they have inherent features and functions that can be integrated with Grid processes in order to customize specific types of network services. ACKNOWLEDGMENTS Joe Mambretti developed, Sections 10.1–10.5 and Franco Travostino for Sections 10.6–10.1. REFERENCES [1] B. Carpenter (1996) “Architectural Principles of the Internet,” RFC 1958, June 1996. [2] S. Blake, D. Black, M. Carlson, E. Davies, Z. Wang, and W. Weiss (1998) “An Architecture for Differentiated Services,” RFC 2475, December 1998. [3] K. Nichols, V. Jacobson, and L. Zhang (1999) “A Two-bit Differentiated Services Architec- ture for the Internet, RFC 2638, July 1999. [4] K. Nichols, S. Blake, F. Baker, and D. Black (1998) “Definition of the Differentiated Services Field (DS Field) in the IPv4 and IPv6 Headers, RFC 2474, December 1998. [5] D. Black, S. Brim, B. Carpenter, and F. Le Faucheur (2001) “Per Hop Behavior Identifi- cation Codes,” RFC 3140, June 2001. [6] B. Davie, A. Charny, J.C.R. Bennet, K. Benson, J.Y. Le Boudec, W. Courtney, S. Davari, V. Firoiu, and D. Stiliadis (2002) “An Expedited Forwarding PHB (Per-Hop Behavior),” RFC 3246, March 2002. [7] A. Charny, J. Bennet, K. Benson, J. Boudec, A. Chiu, W. Courtney, S. Davari, V. Firoiu, C. Kalmanek, and K. Ramakrishnan (2002) “Supplemental Information for the New Defini- tion of the EF PHB (Expedited Forwarding Per-Hop Behavior,)” RFC 3247, March 2002. [8] T. DeFanti and M. Brown, “EMERGE: ESnet/MREN Regional Science grid Experimental NGI Testbed – Final Report,” DOE Office of Scientific and Technical Information. [9] V. Sander (2003) Design and Evaluation of a Bandwidth Broker that Provides Network Quality of Service for grid Applications, NIC Series Vol. 16, John von Neumann-Institut für Computing (NIC). [10] G. Malkin (1998) “RIP Version 2,” RFC 2453, November 1998. [11] J. Moy (1998) “OSPF Version 2,” RFC 2328, April 1998. [12] D. Oran (1990) “OSI IS-IS Intra-domain Routing Protocol,” RFC 1142, February 1990. [13] Y. Rekhter and T. Li (1995) “A Border Gateway Protocol 4 (BGP-4),” RFC 1771, March 1995. [14] M. Gouda and M. Schneider (2003) “Maximizing Router Metrics,” IEEE/ACM Transactions on Networking, 11, 663–675. [15] S. Deering and R. Hinden (1998) “Internet Protocol, Version 6 (IPv6) Specification,” RFC 2460,” December 1998. [16] C. Alaettinoglu, V. Jacobson, and H. Yu (2000) “Toward Millisecond IGP Convergence”, NANOG 20, October 22–24, 2000, Washington, DC. [17] S. Rai, B. Mukherjee, and O. Deshpande (2005) “IP Resilience within an Autonomous System: Current Approaches, Challenges, and Future Directions,” IEEE Communications, 43(10), 142–149. [18] B. Fortz and M. Thorup (2002) “Optimizing OSPF/IS-IS Weights in a Changing World,” IEEE JSAC, 20, 756–767. [19] S. Kent and K. Seo (2005) “Security Architecture for the Internet Protocol,” RFC 4301, December 2005. 196 Chapter 10: Grid Networks and Layer 3 Services [20] S. Kent (2005) “IP Encapsulating Security Payload (ESP),” RFC 4303,”December 2005. [21] M. Bellare, A. Desai, E. Jokippi, and P. Rogaway (1997) “A Concrete Treatment of Symmetric Encryption: Analysis of the DES Modes of Operation”, http://www- cse.ucsd.edu/users/mihir/papers/sym-enc.html. [22] S.E. Deering and D.R. Cheriton (1990) “Multicast Routing in Datagram Internetworks and Extended LANs”, ACM Transactions on Computer Systems, 8(2), 85–110. [23] Reliable Multicast Transport, Internet Engineering Task Force (2004) http://www.ietf. org/html.charters/rmt-charter.html. [24] V. Sander, W. Allcock, P. CongDuc, I. Monga, P. Padala, M. Tana, F. Travostino, J. Crowcroft, M. Gaynor, D. Hoang, P. Primet, and M. Welzl (2004) “Networking Issues for grid Infrastructure,” Grid Forum Document, No. 37, November 2004. [25] K. Birman (2005) Reliable Distributed Systems: Technologies, Web Services, and Applica- tions. Springer Verlag. [26] F. Travostino, L. Feeney, P. Bernadat, and F. Reynolds (1998) “Building Middleware for Real-Time Dependable Distributed Services,” the First IEEE International Symposium on Object-Oriented Real-Time Distributed Computing, p. 162. Chapter 11 Layer 2 Technologies and Grid Networks John Strand, Angela Chiu, David Martin, and Franco Travostino 11.1 INTRODUCTION A consistent theme of this book is architectural flexibility of the infrastructure. Grid environments are designed and implemented to provide a suite of directly usable services and to present options for customization. Grids allow infrastructure resources, including network services, to be configured and reconfigured so that they can meet the exact requirements of applications and higher layer services. This chapter focuses on how Grids can benefit from flexible, reconfigurable layer 2 network services. The Grid community has been exploring options for supple- menting traditional network services with those at layer 2 that can provide exact service qualities, while also providing flexibility through dynamic allocations. One capability inherent in most layer 2 service implementations is switching, a capability for establishing a nonpermanent connection between two or more points. Tradi- tionally, layer 2 switching capabilities have been provisioned with an understanding that such connections would be semipermanent. However, Grids are motivating the development of dynamic switching capabilities for layer 2 services. 11.2 LAYER 2 TECHNOLOGIES AND GRID REQUIREMENTS Grids have typically used layer 2 services as a statically provisioned resource. More recently, Grid environments are being created that utilize dynamic layer 2 resources Grid Networks: Enabling Grids with Advanced Communication Technology Franco Travostino, Joe Mambretti, Gigi Karmous-Edwards © 2006 John Wiley & Sons, Ltd 198 Chapter 11: Layer 2 Technologies and Grid Networks to meet requirements that cannot be easily met by layer 3–4 services. Primarily, this trend is motivated by application requirements. However, it is also being motivated by the technology advances in layer 2 environments that are allowing for enhanced ad-hoc layer 2 service provisioning and reconfiguration. This chapter presents a number of layer 2 technologies that are important to the Grid community. Layer 2 of the OSI model is defined as the layer at which data is prepared for transmission required by aspecificphysicalprotocol. The data is “packaged” so that it can be sent and received across an infrastructure reliably. Functions include those for link control, such as general flow control, error correction, and for Media Access Control (MAC), which governs placement of data on specific physical infrastructure. This last point should be understood in the context of the current network trend, discussed in earlier chapter, toward network virtualization. There are many different implementations of layer 2 services, including virtual services. For example, it is possible to create from higher level processes emulations of network services that resemble physical links with regard to functionality, but actually provide those capa- bilities at higher layers. Virtual Private Networks (VPNs) created using the TCP/IP protocols have been popular for many years. Although many of these virtual services make use of higher layers, they simulate a layer 2 network and are thus included in this chapter. The IETF is developing several architectural approaches that provide for virtual- ization functions for layer 2-type services, including “virtual line services.” One of the most popular protocols developed by the IETF is Multiprotocol Label Switching (MPLS) [1], which is discussed in the next section. It is notable that MPLS and related technologies have been described as layer 2.5 technologies because they do not strictly adhere to the classic definition set forth by the OSI model. 11.3 MULTIPROTOCOL LABEL SWITCHING (MPLS) As explained in Chapter 8, in a typical IP network, traffic is sent from source to desti- nation through a series of routers. Each router relies on the information provided by IP routing protocols (e.g., Open Shortest Path First (OSPF) or Border Gateway Protocol (BGP)), or static routing, to make an independent forwarding decision at each hop within the network. The forwarding decision is based solely on the desti- nation IP address, and this decision is stored in a routing table. When a packet arrives at a router, the router examines the destination address of the incoming packets, looks this address up in the routing table, and then sends the packet to the specified next-hop router. If incoming traffic to the router is heavy, packets are stored until they can be processed. If outbound links are busy, packets are queued until they can be sent. This simple process was the technique used in earliest IP routers and, as other chapters have noted, is still in widespread use today. All packets are treated equally and each packet is analyzed independently. However, from the earliest days of the Internet, network engineers noticed that many Internet applications tend to generate streams of packets with identical source and destination information. These point-to-point streams are called “flows.” Having a router make the same routing decision over and over for the packets in a flow 11.3 Multiprotocol Label Switching (MPLS) 199 is an inefficient use of resources. Besides this inefficiency, the simple and robust destination-based hop-by-hop routing has the following limitations: • It is insensitive to traffic conditions in the network and is therefore inefficient in the way it utilizes network capacity. • Destination-based routing does not allow packets belonging to different classes of services to be routed and protected separately to meet their QoS needs. • Destination-based routing is vulnerable to security attacks by anyone who learns the destinations of other users. The common approach today of filtering incoming packets for disallowed destinations is prone to provisioning errors and costly to implement. To address these limitations, the IETF formed the MPLS Working Group to combine various approaches from the industry into one common standard [1] that can run over any media (e.g., peer-to-peer, frame relay, Asynchronous Transfer Mode (ATM)). With MPLS, each packet is associated with a label that identifies its Forwarding Equivalence Class (FEC). FEC can be defined by a common egress router and option- ally by other characteristics (e.g., a common VPN or class of service). The label is encoded in the connection identifier when available (e.g., ATM Virtual Path Identifier (VPI)/Virtual Channel Identifier (VCI) and Frame Relay Data-Link Connection Iden- tifier (FR DLCI)), otherwise a “shim” is inserted between IP header and link layer header. Based on the label of an incoming packet, a router determines the packet’s next hop as well as the label to use on outgoing interface and performs the label swapping. Label Switched Paths (LSPs) can be established based on the IP address of a router that is downstream on the hop-by-hop route. On the other hand, LSPs can also be established using explicit routing determined offline or by a separate routing protocol. Explicit routing can take the bandwidth availability and traffic conditions into account. MPLS has many potential advantages compared with the traditional hop-by-hop destination-based routing. They are as follows: • MPLS increases routers’ forwarding capacity by switching based on short labels. Although this was one of the main motivations of introducing MPLS, this advantage is diminishing due to many new ASIC designs for fast and scalable prefix-matching algorithms. • MPLS provides a way to perform Traffic Engineering (TE) [2] by overlaying logical connectivity on top of the physical topology to distribute loads more efficiently and in a more scalable way than the traditional IP over ATM overlay model, which was used for traffic engineering for large IP backbones during the early years. • MPLS is connection oriented, and thus is capable of providing faster and more reliable protection and restoration than what current Interior Gateway Protocol (IGP) (e.g., OSPF and IS-IS) rerouting can provide [3]. • MPLS can map traffic with different class of service requirements onto different LSPs, which are routed and protected independently [4]. • MPLS provides added security if packets sent on outgoing interfaces are limited to those from particular connections. • MPLS labels may be stacked to implement tunnels, which is useful in constructing VPNs [5]. 200 Chapter 11: Layer 2 Technologies and Grid Networks The last two in particular enable service providers to offer IP services with enhanced security over a shared network infrastructure, as described below. 11.3.1 MPLS AND SHARED NETWORK INFRASTRUCTURE As other chapters have explained, many applications are not well served by a tech- nique that treats all packets equally as in standard “best effort” Internet services. Also, many communities, including Grid communities, want private networks on which they can determine many key service parameters. One response to this requirement is to build separate infrastructures for each community, but building and maintaining private physical networks for multiple communities is cost prohibitive. The real goal is to design, implement, and maintain a physical network that can be shared, but which is able to treat traffic differently based on individual communities. MPLS has become an ideal tool for meeting this goal. Each community of users sees a VPN, while the network service provider sees a shared network. Using MPLS, traffic from a specific community can be tagged and treated uniquely throughout the network in accordance with that community’s requirements. Each router uses the MPLS labels to make custom routing decisions. For example, a router may send traffic over a moderately loaded path or give preferential treatment together with some key QoS mechanisms, e.g., buffer management and queue scheduling, to those communities that have paid for a premium service. 11.3.2 MPLS AND VIRTUAL PRIVATE NETWORKS Given the cost and complexity of deploying and managing private networks, more and more enterprise customers are implementing IP services to connect multiple sites over a service provider’s shared infrastructure with the requirement that the service provides the same access or security policies as a private network. This require- ment motivated the creation of VPNs. Two implementation models have gained widespread use: • the overlay model, in which the service provider provides emulated leased lines to the customer; • the peer-to-peer model, in which the service provider and the customer exchange layer 3 routing information and the provider relays the data between the customer sites on the optimum path between the sites and without customer’s involvement. The overlay model is easy to understand and implement since the service provider simply offers a set of Virtual Connections (VCs) connecting customers sites either by using some layer 2 WAN technologies such as frame relay and ATM, or with some IP-over-IP tunneling such as Generic Route Encapsulation (GRE) tunneling and IP security (IPsec) encryption. However, it has a few major drawbacks: • It is well suited for enterprise customers with a few central sites and many remote sites. However, it becomes exceedingly hard to manage in a more meshed configuration due to scalability limitation. 11.4 Ethernet Architecture and Services 201 • Proper provisioning of the VC capacities requires detailed knowledge of site-to- site traffic profiles, which are usually not readily available. The peer-to-peer VPN model was introduced to alleviate the above drawbacks. A main solution in this space is BGP/MPLS IP VPNs standardized by the IETF [5]. It utilizes MPLS’s connection-oriented nature and its ability to stack labels, together with Multiprotocol BGP (MP-BGP). MP-BGP offers rich routing policies and the capa- bility to exchange VPN routing information between routers that are not directly connected, such that core provider routers need not to be VPN aware. Note that this network-based VPN solution does not preclude customers’ or the provider’s desire to use IPsec for added security through encryption. 11.3.3 GRID NETWORK SERVICES AND MPLS As noted, Grid environments place unique stresses on network services because of the dynamic nature of Grid systems. The Grid demands that services including data flows be dynamically configured and quickly reconfigured. However, many network implementations are based on statically configured systems with known patterns of behavior. For example, in a standard MPLS implementation, a label switched path is established manually unless Label Distribution Protocol (LDP) [6] is used to set up hop-by-hop LSPs based on shortest path routing. Grid networks demand capabilities for temporary capacity requirements and dynamic topologies and routing. In addition, network services must have QoS guarantees and security assurances. One approach would be to establish MPLS paths for all possible configurations of a Grid. This method is being assisted by equipment vendors, who are improving the performance of routers with large numbers of MPLS paths both by increasing memory and processing power and by new methods for the intelligent management of unused paths. The approach assumes that all Grid nodes are known in advance. Although, fortunately, this is often the case, it does not provide the techniques for rapid reconfiguration. Nothing in the MPLS standard precludes rapid reconfiguration, but, at this time, the tools to accomplish this function are still being researched. Some of the basic tools required are emerging from initiatives established by Grid middleware architecture. 11.4 ETHERNET ARCHITECTURE AND SERVICES Since its invention in 1973 at Xerox Corporation in Palo Alto, California [7], Ethernet has become the most popular Local Area Network (LAN) technology, connecting hundreds of millions of computers, printers, servers, telephone switches, pieces of laboratory equipment, etc., worldwide. Most of today’s network data traffic is gener- ated from Ethernet interfaces. In fact, it is not exaggerating to say that nearly all the IP traffic consists of Ethernet frames. The inventor of Ethernet, Dr. Robert Metcalfe, credited the success to several main factors, including packet switching, distributed nature, speed-up in data rate, and a business model that includes commitment to standardization, fierce competition, preserving install base, etc. 202 Chapter 11: Layer 2 Technologies and Grid Networks The IEEE 802 project has developed a highly defined standard specification for Ethernet in accordance with layer 2, and to some degree layer 1. These specifications describe the required attributes and measures for service quality, based on service availability, out-of-order frames, frame duplication, frame loss, priority, and other parameters. An important part of the specification is the IEEE 802.1D standard, which defines attributes of MAC bridges, such as traffic classification. One measure of success for this standard is its ubiquitous deployment. Another is that switches based on this standard are being implemented on chips, including 10-Gbps rate switches. Owing to fierce competitions based on an open standard, the Ethernet community has adopted the price and performance model of the PC industry. For each new generation, the speed of Ethernet has been improved 10-fold with a targeted price increase of only three to four times. As with the TCP/IP architecture, a key strength of Ethernet’s architecture has been its simplicity and distributed nature. There is no central controller required for Ethernet. The auto-negotiation feature allows seamless interconnection of Ethernet interfaces at different rates and eliminates human errors. Over the last 30 years, the Ethernet medium changed from the original shared coaxial bus in the first generation to dedicated point-to-point fiber-optic links. The medium access has also changed from half-duplex to full-duplex. This change removed the distance limitation imposed by the MAC protocol and enabled Ethernet packets to travel extended distances in their native format. The changes in band- width and MAC opened the path for Ethernet to penetrate from LANs into backbones and WANs. Despite all the changes, the format of Ethernet frames has been kept invariant in all the Ethernet formulations. An end-to-end Ethernet solution eliminates much of the format conversion inefficiencies when different technologies such as SONET/SDH and ATM are used for backbone transport. Unlike SONET/SDH and ATM technologies, which tend to be vendor dependent, Ethernet products from various vendors are highly compatible. In order to support time-sensitive traffic such as voice and video as well as many grades of data traffic, Ethernet now requires additional attributes to continue its development, especially in the areas of QoS, reliability, and management. Many industry groups, including the IEEE, the ITU, the Metro Ethernet Forum (MEF), and the IETF are working diligently on relevant standards to address challenges in each of these areas. Over the next decade, Ethernet will continue to augment or even replace traditional WAN technologies. The merits of using large-scale metro and regional layer 2 networks to suit Grid applications have been demonstrated in research initiatives on several testbeds, described in the appendix. These experiments and demonstrations have shown that dynamic layer 2 path provisioning over Dense Wavelength-Division Multiplexing (DWDM) channels can bring capabilities to distributed infrastructure that cannot be easily provided through traditional routed network services. 11.4.1 ETHERNET ARCHITECTURE FEATURES AND CHALLENGES In recent years, many network operators have started migrating their transport network away from SONET/SDH-based infrastructure to a design optimized around 11.4 Ethernet Architecture and Services 203 IP and Ethernet. The goal is to create a “flatter” network with overall reduced costs of equipment and maintenance, permitting more rapid, flexible service introduction, provisioning, and control. This trend has moved from LAN to MAN, and yet need to happen in the WAN. Key deficiencies of Ethernet that are restricting its use in the wide area include its use of broadcast signaling for resource discovery, its flat address space, its lack of fast protection and restoration capability as well as sophisticated management tools for remote operations and fault detection and resolution, the limi- tations of its spanning tree capabilities, its minimal security attributes, and its limited capabilities for virtualization. All of these areas are important to implementing flex- ible Ethernet services within Grid environments and are all being addressed through IEEE standardization initiatives. One important activity is the IEEE 802.1ah [8], which is defining an architec- ture and bridge protocols that will allow for the interconnection of multiple bridge networks (defined in 802.1ad [9]) to ensure VLAN scalability and associated manage- ment capabilities, e.g., through SNMP. The 802.1ad effort addresses issues such as layer 2 control protocol tunneling, VLAN stacking (“UNI QinQ”), and spanning tree segmentation. The 802.1d and 802.1w define the standards and protocols for the MAC spanning tree specification. Spanning tree prevents loops and assists in fault isolation. The 802.1s (“virtual bridge local area networks”) extends the standard to include support for multiple spanning trees for metro areas. Unlike SONET/SDH which provides fast protection and restoration, existing Ethernet technologies still lack such recovery capabilities from failures. The Resilient Packet Ring (RPR), a ring-based protocol, which was standardized by the IEEE 802.17 work group [10], has been designed particularly to address this need for Ethernet transport. See a brief overview of RPR in Section 11.7. RPR’s ability to guarantee SONET-level protection and restoration time, and its efficient statistical multiplexing capability, together with its support for multiple classes of services with strict perfor- mance guarantees, make it an ideal traffic management layer for Ethernet-based services. Unlike LAN/enterprise networks, in which Ethernet dominates, carrier networks are known for sophisticated and effective management capabilities, and set a high standard for reliability, availability, fast failover, and recovery to ensure that Service Level Agreements (SLAs) made by providers can be met. One of the greatest chal- lenges facing service providers as they deploy Ethernet-based solutions in WAN envi- ronments lies in achieving the same level of Operations, Administration, and Mainte- nance (OAM) support that users are accustomed to receiving with traditional carrier networks. Specifically, SLAs attached to specific services need to be met without regard to the underlying technologies used to provide them. Ethernet OAM is one of the capabilities required to meet SLAs. It includes link performance monitoring, fault detection and fault signaling, and loopback testing. Many standard bodies and forum are addressing this challenge: • ITU-T Y.1730 – Ethernet-based networks and services – provides the Ethernet OAM objectives, requirements, and a quick view of what types of functions need to be implemented, as well as some of the underlying reasons for implementing the function. [...]... Networks: Enabling Grids with Advanced Communication Technology Gigi Karmous-Edwards © 2006 John Wiley & Sons, Ltd Franco Travostino, Joe Mambretti, 218 Chapter 12: Grid Networks and Layer 1 Services (QoS) control Section 12.6 is dedicated to a new optical networking technology, Optical Burst Switching (OBS), with a focus on the integration with Grid 12.2 RECENT ADVANCES IN OPTICAL NETWORKING TECHNOLOGY. .. layer 1 Grid network services A general introduction to layer 1 Grid network service is given in Section 12.2.1 Network control and management issues are discussed in Section 12.3 Section 12.4 deals with the current technical challenges facing the layer 1 networks The Grid network service with an all-optical network infrastructure is presented in Section 12.5 with a focus on Quality of Service Grid Networks: ... wavelengths; and (7) near-real-time feedback of network performance measurements to the applications and Grid middleware As recognized by the standards bodies, new challenges involved in the Gridbased optical control plane will need to address concepts associated with (i) application-initiated optical connections, (ii) interaction with higher layer protocols, (iii) interaction with Grid middleware,... for optical paths [26] Chapter 7 describes the components of Grid middleware and interactions with the applications and the network resources and control in more details A key idea is that, once a request is initiated for Grid resources, the Grid resource manager and resource allocation components coordinates and processes the request and integrates seamlessly, with the Grid control plane behaving as... Chapter 5 This method has been used in Grid environments to provide applications with a means of directly provisioning lightpaths dynamically over advanced optical networks 12.2.3.4 Enumeration of Grid requirements and challenges for layer 1 Common to all solutions related to the layer 1 network control plane is the need to accommodate the dynamic aspects of Grid applications These applications frequently... Travostino (2004) “Optical Network Infrastructure for Grid, ” Grid Forum Document No 36, August 2004 [25] J.D Jones, L Ong, and M Lazer (2005) “Interoperability Update: Dynamic Ethernet Services Via Intelligent Optical Networks , IEEE Communications Magazine, 43(11), pp 4– 47 [26] The Infiniband Trade Alliance, http://www.infinibandta.org/home [ 27] The OpenIB Alliance, http://www.openib.org/ [28] Y Haviv... of Layer 1 Networks etc In addition, the Grid middleware has to be aware of the status of the network and successfully provision the necessary on-demand, guaranteed connections To solve these challenges, the optical networking community, in conjunction with the Grid community, has to rethink the role of both the management plane and the optical control planes and their interaction with Grid middleware... optical networks With the advancement of optical networking technology, high-performance, Optical Circuit-Switched (OCS) networks featuring dynamic dedicated optical channels that can support high-end E-science applications are emerging [6–11] On these networks, long-lived wavelength or subwavelength paths between distributed sites could be established – driven by the demands of applications Together with. .. services for their production networks The Grid community is further exploring means of extending recovery mechanisms to provide more adaptive responses within the Grid context 12.4 CURRENT RESEARCH CHALLENGES FOR LAYER 1 SERVICES 12.4.1 APPLICATION-INITIATED CONNECTIONS An increasing number of Grid applications today require accessibility to the globally distributed Grid resources Many of these applications... to compete on either costs or performance with an adequately utilized SONET/SDH alternative Unfortunately for Grid networks, the deployed SONET/SDH networks historically have only provided “pipes”: Static connections are reconfigurable only on a timescale of weeks or longer, and (if a network operator is involved) via a provisioning request and negotiation with the operator Since the durations of even . environments are being created that utilize dynamic layer 2 resources Grid Networks: Enabling Grids with Advanced Communication Technology Franco Travostino, Joe Mambretti, Gigi Karmous-Edwards © 2006. AND GRID REQUIREMENTS Grids have typically used layer 2 services as a statically provisioned resource. More recently, Grid environments are being created that utilize dynamic layer 2 resources Grid. Changing World,” IEEE JSAC, 20, 75 6 76 7. [19] S. Kent and K. Seo (2005) “Security Architecture for the Internet Protocol,” RFC 4301, December 2005. 196 Chapter 10: Grid Networks and Layer 3 Services [20]