Differentiate services from MPLS

One reason for the huge success of the Internet is the simplicity of the IP protocol it uses. The Internet provides a best-effort service: the IP network tries its best to forward data packets from the source to the target, but does not make any commitment to the quality of service QoS that can be provided for packet forwarding. Therefore, the quality of services provided by the IP network is unpredictable.
Many new Internet businesses are multimedia applications that require either huge bandwidth, strict latency assurance, or one-to-many or multi-point-to-multi-point communication capabilities. These new services require the IP network not only to provide simple and best-effort services, but also to provide new services. The current "dumb" IP network must have certain intelligence.
IETF is optimistic about two IP QoS standards: DiffServ and MPLS. DiffServ replaces the IP service type TOS) field to the DS field, and uses it to carry the information required by the IP packet service. It is a strict layer-3 technology and does not involve low-layer transmission technology. On the other hand, MPLS defines a method to map layer-3 traffic to connection-oriented layer-2 transmission technologies such as ATM and frame relay. It adds a specific routing information to each IP packet, allows routers to assign different explicit routes for different types of traffic to provide routing capabilities that are not directly related to QoS, such as traffic engineering TE), to improve the efficiency of IP routing.
I. Differentiated Services 
DiffServ originated from IntServ. DiffServ aims to provide different service levels for traffic on the Internet. Compared with IntServ, DiffServ defines a relatively simple and coarse-grained control system. In addition, DiffServ is used for each type of QoS control after stream aggregation, rather than for each stream as IntServ does. Therefore, DiffServ is scalable and can provide QoS Services on large networks.
DiffServ classifies incoming streams at the edge of its domain, and specifies a type flag for each type (DiffServ Code Point DSCP ). The core router in the domain can view the DSCP value and forward the scheduling package according to the specific hop-by-hop behavior of each class. DiffServ calls a group of packets forwarded based on the same PHB as behavior aggregation BA ).
The DiffServ workgroup has defined the Ding relationship between DSCP and PHB table 1), but it also allows the ISP to customize the ing relationship with local significance.
Table 1 Recommended DSCP values
Dscp phb description
101110 EF absolute QoS
001XXX AF1 QoS is between EF and BE. Each AF can be divided into three priorities: 12 in total
010XXX AF2
011XXX AF3
100 AF4
000000 BE best effort for business
1. Acceleration of forward EF)
Ef phb or EF) traffic is not affected by other PHB traffic, ensure that the package exit rate is higher than the specified value. Similar to traditional lease lines, ef phb can provide low packet loss rate, low latency, low jitter, and guaranteed bandwidth services. Ef phb forwarding only provides a minimum queue for accepted fixed traffic and convection, and discards any flow that exceeds the specified number of EF streams on the edge router.
2. Ensure the forwarding of AF)
AF provides different levels of forwarding features for IP packets, and assigns a specific number of forwarding resources, such as the buffer zone and bandwidth, to each of the four levels of AF ), and assign one of three different discard priorities to each package. Af phb allows a higher probability of packet forwarding when the overall traffic does not exceed the preset rate.
Ii. MPLS 
MPLS originated from IP switching and label switching. MPLS defines the architecture and protocol for encapsulating IP traffic with the new routing protocol, while DiffServ only focuses on the IP packet fields independent from the existing routing protocol, therefore, it has significantly changed traditional IP networks than DiffServ.
An IP address is a connectionless network. Each vro searches for a matched Next Hop Based on the address of each packet received and forwards the packet accordingly. However, the vro uses the longest prefix matching address search, that is, to search for the entry with the longest matched prefix. MPLS adds a fixed-length label to each packet to the edge router of the network entry. The core router forwards the label value to the egress edge router and then restores it to the original IP packet. Because the destination address is searched based on a fixed-length label, MPLS can achieve high-speed forwarding. The forwarding path determined by the tag is called the tag switching path LSP ).
MPLS implements an explicit LSP and selects an optimized edge-to-edge path based on traffic QoS requirements. MPLS can also perform Load Balancing traffic engineering within the network, or Virtual Private Network VPN ). The initial motive of MPLS is to achieve high-speed route forwarding. However, as the performance of the router continues to improve, this reason no longer exists, however, it is becoming increasingly popular to establish connections on the IP network to implement traffic engineering and establish VPN.
In MPLS Traffic Engineering, you can use the following label distribution process to create LSP.
1. constraint routing label distribution protocol
Constrained routing label distribution protocol CR-LDP) is an extension of the existing label distribution protocol LDP, can support constrained routing. The CR-LDP can create an LSP in the tag request message through lsr B and LSR C ). In addition, the CR-LDP allows you to set traffic parameters in a tag request message, such as peak rate, commitment rate, and burstable. However, CR-LDP does not really support QoS Assurance for LSP traffic.
2. RSVP-TE
A RSVP-TE is a tag Distribution Protocol extended from an existing RSVP protocol. It uses several new RSVP objects, such as mandatory LABLE-REQUEST objects and LABEL objects. RSVP-TE can support additional functionality for creating and maintaining LSP, including on-demand downstream Label Distribution, explicit LSP instances, allocating network resources for explicit LSP, and using make-before-break before interruption) "re-routing the established LSP tunnel, tracking the real routing of LSP tunnel, diagnosing LSP tunnel, node digest ideas, preemptive selection and controllable management.
3. Management personnel establish one by one
Network administrators use network management systems such as Simple Network Management Protocol (SNMP) and command line interface (CLI) to establish MPLS usage information for all routers on each LSP.
Iii. DiffServ supported by MPLS

MPLS uses a label as the cushion shim) to encapsulate an IP packet. The core router cannot see DSCP, and DiffServ is not compatible with MPLS. Therefore, IETF proposes a method for MPLS to support DiffServ.
The DiffServ supported by MPLS can map multiple BA of DiffServ to an LSP of MPLS and forward traffic on the LSP according to the PHB of BA. LSP has two ways to map to BA: E-LSP and L-LSP.
1. E-LSP
The E-LSP assigns multiple BA To an LSP with the EXP field, and uses the EXP field of the MPLS cushion header to represent the PHB of a package. You can map up to eight BA entries to the EXP field.
2. L-LSP
The L-LSP assigns an LSP to an BA to show the discard priority of multiple packets), determines the packet scheduling policy based on the MPLS label, and determines the discard priority according to the cushion header or Layer 2 packet discarding mechanism. The EXP field cannot be used for pure atm mpls.
Because MPLS network devices exchange tag values in each hop, it is difficult to manage the ing between tags and DSCP. The E-LSP is easier to control than the L-LSP, because the E-LSP can determine the EXP ing between the EXP field and DSCP for each packet throughout the network in advance.
Iv. Conclusion
DiffServ provides a service that treats different businesses differently and sets different priorities and forwarding characteristics for different businesses. However, it does not specify how to forward the services, nor does it try to eliminate congestion. When the network is not congested, even the best-effort IP packet can also get good network services, so there is no DiffServ is the same. However, the DiffServ network may also be congested because even if DiffServ adjusts the traffic at the edge of the network, on the core vro of the network, congestion and load imbalance may also occur due to traffic convergence ). In this case, DiffServ may damage services with high priority when QoS congestion of high-priority businesses is serious at the cost of low-priority businesses, but does not eliminate congestion. The Traffic Engineering provided by MPLS can solve the congestion problem caused by load imbalance.
DiffServ does not care about the technologies used by low-Layer Networks to solve network congestion problems, while MPLS Traffic Engineering does not care about the QoS mechanism used by the traffic. But when MPLS carries DiffServ or IntServ, because of label encapsulation, the core router cannot see the IP packet header. Therefore, IETF proposes a technology that MPLS can support DiffServ.
Obviously, QoS for IP networks is not enough because only DiffServ and MPLS technologies are used. Congestion Control technologies such as RED and queuing scheduling technologies such as WFQ are also required), constrained routing, application-layer traffic redirection, and traffic balancing technology.