Results (
Vietnamese) 2:
[Copy]Copied!
2. Technologies
The concept of multiple coexisting networks appeared in the networking literature in different capacities. In this section, we discuss four such incarnations: Virtual Local Area Networks (VLAN), Virtual Private Networks (VPN), active and programmable networks, and overlay networks.
2.1. Virtual local area network
A virtual local area network (VLAN) [5] is a group of hosts with a common interest that are logically brought together under a single broadcast domain regardless of their physical connectivity. Since VLANs are logical entities, i.e., configured in software, they are flexible in terms of net- work administration, management, and reconfiguration. Moreover, VLANs provide elevated levels of trust, security, and isolation, and they are cost-effective.
Classical VLANs are essentially Layer 2 constructs, even though implementations in different layers do exist. All frames in a VLAN bear a common VLAN ID in their MAC headers, and VLAN-enabled switches use both the destination MAC address and the VLAN ID to forward frames. This process is known as frame coloring. Multiple VLANs on multiple switches can be connected together using trunking, which allows information from multiple VLANs to be carried over a single link between switches.
2.2. Virtual private network
A virtual private network (VPN) [6–8] is a dedicated communications network of one or more enterprises that are distributed over multiple sites and connected through tunnels over public communication networks (e.g., the Internet).
Each VPN site contains one or more Customer Edge (CE) devices (e.g., hosts or routers), which are attached to one or more Provider Edge (PE) routers. Normally a VPN is man- aged and provisioned by a VPN service provider (SP) and known as Provider-provisioned VPN (PPVPN) [9]. While VPN implementations exist in several layers of the network stack, the following three are the most prominent ones.
2.2.1. Layer 3 VPN
Layer 3 VPNs (L3VPN) [10,11] are distinguished by their use of layer 3 protocols (e.g., IP or MPLS) in the VPN back- bone to carry data between the distributed CEs. L3VPNs can again be classified into two categories: CE-based andPE-based VPNs.
In the CE-based VPN approach, CE devices create, man- age, and tear up the tunnels without the knowledge of the SP network. Tunneling requires three different protocols:
(1)Carrier protocol (e.g., IP), used by the SP network to carry the VPN packets.
(2)Encapsulating protocol, used to wrap the original data. It can range from very simple wrapper proto- cols (e.g., GRE [12], PPTP [13], L2TP [14]) to secure protocols (e.g., IPSec [15]).
(3)Passenger protocol, which is the original data in cus- tomer networks.
Sender CE devices encapsulate the passenger packets and route them into carrier networks. When the encapsu- lated packets reach the receiver CE devices at the end of the tunnels, they are extracted and actual packets are in- jected into receiver networks.
In PE-based L3VPNs, the SP knows that certain traffic is VPN traffic and process them accordingly. The VPN states are stored in PE devices, and a connected CE device be- haves as if it were connected to a private network.
2.2.2. Layer 2 VPN
Layer 2 VPNs (L2VPNs) [16,17] provide end-to-end layer 2 connection between distributed cites by transporting Layer 2 (typically Ethernet but also ATM and Frame Relay) frames between participating sites. The primary advantage of L2VPN is its support of heterogeneous higher-level pro- tocols. But its lack of a control plane takes away its capabil- ity of managing reachability across the VPN
There are two fundamentally different kinds of Layer 2 VPN services that an SP could offer to a customer: point-to-point Virtual Private Wire Service (VPWS) and point-to-multipoint Virtual Private LAN Service (VPLS). There is also the possibility of an IP-only LAN-like Service (IPLS), which is similar to VPLS except that CE devices are hosts or rou- ters instead of switches and only IP packets are carried (either IPv4 or IPv6).
2.2.3. Layer 1 VPN
Accompanied by the rapid advances in next-generationSONET/SDH and optical switching along with GMPLS [18]control, the Layer 1 VPN (L1VPN) [19,20] framework emerged from the need to extend L2/L3 packet-switchingVPN concepts to advanced circuit-switching domains. It enables multiple virtual client-provisioned transport net- works over a common Layer 1 core infrastructure. The fun- damental difference between L1VPNs and L2 or L3 VPNs is that in L1VPNs data plane connectivity does not guarantee control plane connectivity (and vice versa).
The main characteristic of L1VPN is its multi-servicebackbone where customers can offer their own services with payloads of any layer (e.g., ATM, IP, TDM). This allows each service networks to have independent address space, independent Layer 1 resource view, independent policies, and complete isolation.
L1VPN can be of two types: Virtual Private Wire Ser- vices (VPWS) and Virtual Private Line Services (VPLS). VPWS services are point-to-point, while VPLS can bepoint-to-multipoint.
2.3. Active and programmable networks
While active and programmable networks may not be considered as direct instances of network virtualization, most of the projects in this area pushed forward the con- cept of coexisting networks through programmability. In order to allow multiple external parties to run possibly conflicting code on the same network elements, active and programmable networks also provide isolated envi- ronments to avoid conflicts and network instability.
The programmable networks community discusses how communications hardware can be separated from control software. Two separate schools of thought emerged on how to actually implement such concepts: one from tele- communications community and the other from IP net- works community [21].
2.3.1. Open signaling approach
Open signaling takes a telecommunication approach to the problem with a clear distinction between transport, control, and management planes that constitute program- mable networks and emphasize QoS guarantees for created services [21]. It argues for modeling communication hard- ware using a set of open programmable network interfaces to enable controlled access to switches, routers, and even- tually network states by external parties.
2.3.2. Active networks approach
The active networks [22] community allow routers and switches to perform customized computations based on packet contents, and they also allow network elements to modify packets. The active networks approach allows cus- tomization of network services at packet transport granu- larity instead of doing so through a programmable control plane. The result is increased flexibility through a more complex programming model with higher security risks.
Different suggestions on levels of programmability exist in active networks literature. At the one end, ANTS [23] of- fers a Turing-complete machine model at the active router enabling each user to execute any new code. At the other end of the spectrum, DAN [24] only allows the user to call functions already installed at a particular node. However, due to lack of interest from network operators to open up their networks to external parties, none of the proposals are in use.
2.4. Overlay networks
An overlay network is a virtual network that creates a virtual topology on top of the physical topology of another network. Nodes in an overlay network are connected through virtual links which correspond to paths in the underlying network. Overlays are typically implemented in the application layer, though various implementations at lower layers of the network stack do exist.
Overlays are not geographically restricted, and they are flexible and adaptable to changes and easily deploy- able in comparison to any other network. As a result, overlay networks have long been used to deploy new fea- tures and fixes in the Internet. A multitude of overlay de- signs have been proposed in recent years to address diverse issues, which include: ensuring performance [25] and availability [26]of Internet routing, enabling multicasting [27–29], providing QoS guarantees [30], pro- tecting from denial of service attacks [31,32], and for content distribution [33], file sharing[34] and even in storage systems [35]. Overlays have also been used as testbeds (e.g., PlanetLab [36]) to design and evaluate new architectures. In addition, highly popular and widely used peer-to-peer [34] networks are also overlays in the application layer.
However, in their seminal paper on network virtualiza- tion, Anderson et al. [1] point out that existing overlay technologies cannot be considered as a deployment path for disruptive technologies because of two main reasons. First, they are mostly used to deploy narrow fixes to specific problems without any holistic view of the interac- tions between coexisting overlays. Second, most overlays, being designed and deployed in the application layer on top of IP, are not capable of supporting radically different architectures.
Being translated, please wait..
