The present invention relates to content-aware dynamic network resource allocation in a network such as the Internet and in one embodiment to content-aware dynamic optical bandwidth allocation.
The Internet is a data transport network including legacy networks and optical networks. The legacy networks exist at the outer edge of the Internet and use copper wires to connect client systems and electronic routers/switches. The optical networks form the backbone of the Internet and use fibers to connect optical cross connects or switches. Between the legacy networks and optical networks are edge devices, which are electro-optic.
Well-developed optical transport technology brings an ever-increasing amount of bandwidth to the Internet. However, with conventional network technology, end clients have little ability to exploit optical network resources for their own purposes. As a result, there is abundant bandwidth available in the backbone of the Internet. In fact, a number of emerging Internet applications such as storage area network and streaming media intend to take greater advantage of the abundant bandwidth. These applications are dynamically initiated and provide a variety of content data over the IP protocol.
More specifically, conventional network technology has the following drawbacks.
1. Static Provisioning of Optical Links
Conventional network resource provisioning establishes fixed links to connect customer networks. For example, the optical provisioning is static, fixed bandwidth, and usually takes a long time, e.g., on the order of a month, because it involves inter-network-provider service negotiations and is accomplished manually.
2. Cost of the Optical Link Bandwidth is not Divided Among Many Users and Provisioning Bandwidth is not Flexible
Conventionally, provisioning an optical link (e.g., OC-48) means that a customer owns the full link bandwidth (i.e., 2.5 Gbps). This customer is fully responsible for the cost of the optical link. The cost is significant and the optical link bandwidth use is often inefficient. On the other hand, a number of users can share an optical link but they are not guaranteed a specified portion of the bandwidth.
3. Existing Signaling Protocols not Supported by Optical Gear
Applications can send their bandwidth requirements to the network using existing Internet signaling protocols such as Resource Reservation Protocol (RSVP). These protocols are at the Internet protocol (IP) layer, i.e., the layer 3 (L3), of the ISO Open System Interconnection reference model and require hardware support at network devices. However, optical devices perform data transport at the physical (L1) or the link (L2) layer and thus application signals are not processed in optical networks.
4. Optical Bandwidth Provisioning is not Aware of the Content of Application Traffic
Conventional bandwidth provisioning is based on the TCP/IP characteristics of traffic flow, which include IP protocol types, source and destination IP addresses, and TCP/UDP source and destination port numbers of traffic packets. Such provisioning is limited for L4 or higher-layer content differentiation because a client may use applications that deal with multiple content traffic streams at the same time. For example, audio and video applications employ different types of content and have different bandwidth requirements. On the other hand, optical networks do not support content differentiation because optical devices do not process IP packets.
Prior attempts to solve the problems described above include the following.
1. Intserv/RSVP
RSVP is the Intserv ReSerVation Protocol under the Internet Engineering Task Force (IETF) and is used by applications to signal the network for bandwidth reservations for IP traffic. However, RSVP is thought to be not scalable because backbone routers cannot maintain the flow status for all reserved traffic passing by. In addition, optical gear does not accept RSVP signals from end applications because RSVP is an L3 IP protocol.
2. ATM
Within an asynchronous transfer mode (ATM) network, applications can invoke the ATM user-network interface (UNI) to establish virtual circuits with particular bandwidth assignments. However, the ATM UNI is not applicable for non-ATM applications.
3. GMPLS
GMPLS (Generalized MultiProtocol Label Switching) is a known protocol of traffic path establishment for next-generation optical network. GMPLS is applied with the emerging ASTN (Automatic Switch Transport Network) technology. However, GMPLS does not support granular bandwidth requests from individual clients nor does it allocate bandwidth based on the content of application traffic.
Thus, there remains a need to more fully and effectively exploit the abundant bandwidth existing at optical networks.
The present invention relates to content-aware dynamic network resource allocation. In one embodiment the network resource of interest is optical bandwidth and this embodiment is termed content-aware dynamic optical bandwidth allocation (CADOBA). Although the following often refers to the CADOBA embodiment, those of skill in the art will appreciate that one can use the present invention to allocate other network resources in addition to bandwidth. CADOBA enables clients at end legacy networks to manipulate the network resources of backbone optical networks for their own purposes. One embodiment of the invention provides three intelligent mechanisms: 1) a gateway mechanism operative to transform application requests in an application transparent way to intelligent network control; 2) a provisioning mechanism operative to perform optical link or lightpath setup and bandwidth allocation; and 3) a classification mechanism operative to differentiate application traffic based on L4+ content characteristics and to forward content streams to the provisioned routes (optical links or lightpaths).
Another embodiment of the invention provides a method for performing content-aware optical bandwidth allocation over a network including an edge device and an optical control unit. Typically networks include client systems, legacy networks, optical networks and edge devices connecting the legacy networks with the optical networks. Both legacy and optical networks have network control units. A control unit can be a part of a network device, or it can be an associated device. Both legacy and optical networks have network domains. An ISP network is an example of a network domain. All network devices in a network domain are managed in a similar way.
The method includes: receiving signal traffic of applications from the edge device; determining a network resource request for a particular application data traffic with content specification from a received signal; communicating with an optical control unit to allocate a network resource to provide a network resource provisioned route, and creating rule data for the edge device to forward content-specific application traffic onto the network resource provisioned route.
Still another embodiment of the invention provides a system for performing content aware optical bandwidth allocation over a network including an edge device and an optical control unit. The system includes a gateway module, a provisioning module in communication with the gateway module and a classification module in communication with the gateway module.
The gateway module, in communication with the edge device, receives signal traffic of applications from the edge device and determines, from the signal traffic, a bandwidth request associated with a content-specified traffic.
The provisioning module receives a provisioning request from the gateway module regarding the bandwidth request for specified content traffic, and communicates with an optical control unit to allocate optical bandwidth to provide a provisioned optical route. The provisioning module also communicates with the edge device to allocate the appropriate bandwidth for the content-specified traffic forwarding towards a provisioned optical route;
The classification module receives a content classification request from the gateway module, and, based on the classification request, creates a content-aware traffic routing rule for the edge device to forward specified content traffic to the provisioned bandwidth route.
These and other features of the invention are more fully set forth with reference to the following detailed description and the accompanying drawings.
The foregoing and other objects, features, and advantages of the invention will be apparent from the following description and apparent from the accompanying drawings, in which like reference characters refer to the same parts.
The present invention relates to content-aware dynamic network resource allocation. In one embodiment the network resource of interest is optical bandwidth and this embodiment is termed content-aware dynamic optical bandwidth allocation (CADOBA). With reference to
Thus, the CADOBA system can reside on control units. That is, in one embodiment each control unit runs a CADOBA system respectively and they work together through intra-CADOBA communications.
Traffic refers to the data of an application traveling across a network, and signal traffic refers to the data of network resource requests of applications signaling to a network. Content refers to the L4 or high-layer (L4+) data carrying through the IP protocol, i.e., the payload of an IP packet. Typical content specifications include TCP/UDP headers, message identifiers such as HTTP (Hypertext Transfer Protocol) headers and RTSP (Real Time Streaming Protocol) method tags, and other L4+ characteristics.
With reference to
With reference to
With reference to
In operation, and with reference to
With reference to
With reference to
With reference to
With reference to
In order to appreciate user needs and to allocate optical resources in real time, CADOBA is focused at the optical edge (devices) even though it is distributed in the network. Applications at end networks signal their content specifications and network resource requirements to the network using a well-known signaling protocol such as RSVP.
The CADOBA gateway system residing at a control unit of an edge device, obtains signal packets by causing the transport unit of the edge device to filter application traffic and to forward signal traffic filtered from the application traffic to the gateway module. The edge device can filter application traffic using L3+ packet filters. The signal traffic includes network-related parameters, e.g., bandwidth, requests/requirements, and content tags/identifiers.
The CADOBA provisioning system at a control unit sets up a new lightpath in an optical network domain for the requested application traffic or manipulates an existing lightpath with dynamic bandwidth utilization. The provisioning system also sets up bandwidth configuration of an edge device so that application traffic can pass through without a bottleneck to the optical network domain.
The CADOBA classification system at a control unit causes an edge device, i.e., an electronic switch, to filter application traffic and to differentiate packets with specified content characteristics to form content streams. The classification system further causes the electronic switch to route the content streams to respective provisioned optical routes.
In one embodiment, the CADOBA processing is dynamic as application content and requirements may vary at runtime.
In summary, embodiments of the invention can include the following elements.
1. Dynamic Provisioning of Optical Links
Embodiments of the invention perform flexible optical route or lightpath establishment in optical networks according to application requests. For example, with certain optical products, one embodiment of the invention dynamically sets up an RPR (Resilient Packet Ring) over an existing optical SONET or DWDM ring, and increases or decreases the RPR use of the whole optical ring bandwidth.
2. Cost of the Optical Link Bandwidth is Divided Among Many Users
Embodiments of the invention enable a number of applications to share one optical link or lightpath by identifying the content of the applications and routing the applications traffic accordingly. For example, by forwarding applications traffic to respectively optical Ethernet ports in an RPR and adjusting the port rates, embodiments of the invention provide each application the optical bandwidth it requests. As a result, these embodiments of the invention increase the return on the service provider investment by utilizing the optical transport capability efficiently.
3. Provisioning Bandwidth is Flexible and Economic
Embodiments of the invention enable applications to request bandwidth in small increments, e.g., as small as 1 Mbps, which is a small portion of the whole link bandwidth. For example, with certain optical Ethernet support, an application can request a bandwidth from 1 Mbps to 1 Gbps.
4. Application Signal can Achieve Optical Control
The gateway module allows an application to achieve optical control by intercepting application signals at the edge device and transforming the application signal to a corresponding network control. For example, the optical control signal in turn effects a reconfiguration of optical devices, including, in one embodiment, a reallocation of bandwidth and establishment of an optical link or light-path.
5. Network Resource, E.G., Bandwidth, Allocation is Aware of Content
The classification module possesses strong content processing capability that can differentiate application streams according to their content (L4 or higher layer). For example, with certain optical gear, application traffic is classified and routed to a provisioned optical lightpath by forwarding packets to specific optical Ethernet ports. Consequently, content applications can achieve higher quality. Higher quality applications in turn lead to an increase in demand for bandwidth and generate more revenue from specialized network services.
In a demonstration of the operation of one embodiment of the present invention, an optical network included fibers and optical devices, a legacy electronic network included PC host systems, and content switches used as the edge devices bridging the two network domains: electronic (legacy) and optical. Portions of one embodiment of a CADOBA software system ran at the control unit of an edge device. Portions of the embodiment of the CADOBA system also ran at other network control devices. This embodiment acted as an intelligent gateway, transferring requests from content applications to network control units. The result in Table 1 shows that embodiments of the CADOBA system in
The two programs Slideclient and Slideserver transmit continuously 2˜3 MB JPEG images, running on two Pentium-2 Linux PCs. Slideserver is the image source and sends a requested image to Slideclient. Slideclient displays an image it receives.
“Required BW (Bandwidth)” is the bandwidth parameter requested by an application signal for the Slideclient-Slideserver content traffic. Slideserver signals a changing bandwidth every 30 seconds or longer.
“Provisioned BW (Bandwidth)” is the real bandwidth allocated by CADOBA for the requested content traffic. It is not always equal to the “Required BW (Bandwidth)” as network bandwidth becomes less available.
“Throughput” is the maximal rate that Slideclient receives pure image data from Slideserver via TCP at each signal case. It excludes overheads of L1-L4 protocol headers in a packet. The maximal throughput reached by the application programs is 47.51 Mbps.
An initial demonstration of the CADOBA system (see
With reference to
SlideServer 106 could signal again later at any time that the bandwidth requirement is changed. When this embodiment of the CADOBA system (at control unit master) processed the signal, it would perform a new network setup/configuration over the content switch and optical device when necessary. The new network setup/configuration can include reallocation of bandwidth, reconfiguration of forwarding priority and reconfiguration of optical content routes. As a result, at runtime, the illustrated embodiment provided dynamic bandwidth allocation and forwarding conditions, i.e., it provided different service levels based on signals from the requesting application.
Embodiments of the present invention advantageously encourage application developers to develop bandwidth intensive applications such as video teleconferencing, streaming media and storage area networking.
The present invention is not limited to bandwidth allocation. Rather, one can use embodiments of the invention in other applications. For example, one can use embodiments of the invention to create an optical virtual private network (VPN), to develop accounting packages that meter network usage by an application, and to build a many-to-many videoconference channel.
Furthermore, since certain changes may be made in the above constructions without departing from the scope of the invention, it is intended that all matter contained in the above description or shown in the accompanying drawings be interpreted as illustrative and not in a limiting sense.
This application claims priority to U.S. provisional application Ser. No. 60/336,469 filed Nov. 2, 2001 and entitled “Content-Aware Dynamic Bandwidth Allocation.
Number | Name | Date | Kind |
---|---|---|---|
5893911 | Piskiel et al. | Apr 1999 | A |
5995503 | Crawley et al. | Nov 1999 | A |
6594260 | Aviani et al. | Jul 2003 | B1 |
6631134 | Zadikian et al. | Oct 2003 | B1 |
6850980 | Gourlay | Feb 2005 | B1 |
6856627 | Saleh et al. | Feb 2005 | B2 |
6876668 | Chawla et al. | Apr 2005 | B1 |
6950391 | Zadikian et al. | Sep 2005 | B1 |
20020067726 | Ganesh et al. | Jun 2002 | A1 |
20020087699 | Karagiannis et al. | Jul 2002 | A1 |
20020136204 | Chen et al. | Sep 2002 | A1 |
20020150093 | Ott et al. | Oct 2002 | A1 |
20030165139 | Chen et al. | Sep 2003 | A1 |
20030169749 | Huang et al. | Sep 2003 | A1 |
20030210694 | Jayaraman et al. | Nov 2003 | A1 |
20050185654 | Zadikian et al. | Aug 2005 | A1 |
Number | Date | Country | |
---|---|---|---|
60336469 | Nov 2001 | US |