Multi Service Port (MSP)

Transcription

1 Technical Annex Moreelsepark EP Utrecht PO Box DA Utrecht ING Bank: NL54 INGB COC Utrecht VAT NL B01

2 Table of contents 1. What is an MSP? Technical properties of the MSP Interface MSP Network services separated by VLAN tags Differences between an MSP and an SSP Bandwidth management Bandwidth setting on the SURFnet side Bandwidth setting on the institution side Bandwidth on Demand, virtual port on MSP Network service reports Abbreviations This publication is licenced under the Creative Commons Attribution 3.0 Licence. 2

3 1 What is an MSP? The Multi Service Port (MSP) is a type of port that SURFnet offers to institutions. This port provides the capability to support multiple network services, such as light paths, simultaneously. Every network service on an MSP uses some of the available bandwidth on the physical port. The various network services on an MSP are separated from each other by VLAN tags. Up to 10 different network services can be supported per MSP. The desired bandwidth is set when the network service is activated, but it can be adjusted after activation. The ports are used efficiently because multiple network services can be combined on a single MSP. For example, several light paths to other institutions and a light path to a cloud provider can be connected to a 10 Gbit/s MSP. Figure 1 illustrates a 10 Gbit/s MSP, a 1 Gbit/s and a 2 Gbit/s light path and a 3 Gbit/s on demand light path. A new light path can easily be added to an MSP without having to install hardware. If you use on demand light paths you can even activate the light path fully independently. This means light paths can be used highly flexibly for long or short periods. Figure 1 Multiple network services per MSP 3

4 2 Technical properties of the MSP 2.1 Interface MSP The physical connection to the SURFnet network is known as the interface. The interface options for the MSP are identical to those for the existing Single Service Port (SSP). If an SSP is available on SURFnet equipment, it can easily be converted into an MSP with a simple modification to the configuration. Typical interfaces are: Gigabit Ethernet: 1000Base-SX or 1000Base-LX 10 Gigabit Ethernet: 10GBase-SR or 10GBase-LR 2.2 Network services separated by VLAN tags The principle behind the MSP is based on the use of a VLAN ID (or VLAN tag). The MSP distinguishes network services based on the VLAN tag 1 which is unique for each network service (Figure 2). Every network service is connected to a virtual switch (VS) at the start and end points in the SURFnet network. This virtual switch provides the logical separation between the various network services within a single Carrier Ethernet switch, before the data traffic is encapsulated. This way, the data traffic can be transported by the SURFnet network separated end-to-end. The virtual switches at the start and end points are connected by a PBB-TE tunnel. The physical path of PBB-TE tunnels through the network does not change so jitter is minimal and latency is constant. When two tunnels are configured in protected mode, PBB-TE tunnels can perform a protection switchover from the primary to the secondary tunnel within 50ms. The VLAN tags that are used to separate the network services can be different at each end of the light path. The SURFnet equipment can, after all, re-tag the VLAN tags which provides a high degree of flexibility. When the VLAN ID is re-tagged the VLAN ID of an incoming data packet is changed to a different VLAN ID. For example, a light path starting on an MSP can use VLAN tag 15 on one side and VLAN tag 31 on the other side. 1 The VLAN tag is used for single tagged packets (802.1Q). The outer tag, the so-called S-VLAN tag, is used for double tagged packets (QinQ 802.1ad). The C-VLAN tags in QinQ packets are not visible and are forwarded transparently. 4

5 Figure 1 Overview of data flow through the SURFnet network between MSP and MSP/SSP Depending on the equipment that is used on the location a light path can also be built between an MSP and an SSP. When doing so the untagged (and any tagged) traffic arriving on the SSP will be given an additional VLAN tag on the outgoing MSP side. In the opposite direction the VLAN tag on the incoming tagged traffic on the MSP will be stripped. Please note: tagged traffic on the SSP will be double tagged on the MSP (802.1ad), because an additional VLAN tag is added. 2.3 Differences between an MSP and an SSP Because the MSP supports a number of network services with different possible end points, certain port-specific properties that were possible on point-to-point connections between SSPs can no longer be used. Firstly, the MSP will not offer transparency for a number of Layer 2 control protocol packets. CDP, LLDP, LACP and PAGP for example. As a result, MSPs cannot be included in a link bundle. In addition, Layer 2 protection mechanisms, such as the Spanning Tree protocol are not supported by a light path on an MSP. 5

6 Nor is Remote port shutdown 2 supported on an MSP. This is to prevent the entire MSP being brought down if other network services could still be active. This is particularly relevant for redundant light paths, where the institution's network determines which path is the active path. The detection of a network fault will have to come from the next higher network protocol such as BGP/BFD/CFM before a switchover to the backup path is performed. 2 If the SURFnet device is no longer receiving light from the client device on one side of a light path, then the SURFnet device will ensure that no light is transmitted to the other side of the light path. Should the network fault lie in the SURFnet domain then no light will be transmitted to the connected client device on either side of the light path. 6

7 3 Bandwidth management An MSP in the SURFnet network provides multiple light paths on the same physical port. Light paths provide guaranteed bandwidth with minimum jitter and delay. To achieve these guarantees the bandwidth allocation is set in such a way that the various network services on an MSP cannot affect each other. As a result, each network service can use the entire agreed bandwidth. The mechanisms used for this on the SURFnet side and possible mechanisms on the institution side to optimise the end-to-end performance are described below. 3.2 Bandwidth setting on the SURFnet side SURFnet uses traffic policing to ensure that no network service transmits more traffic to the network than has been agreed. As a result, guarantees can be provided for all network services. Packets that exceed the agreed data limit are not forwarded. This data limit is determined by the Committed Information Rate (CIR). The CIR is the maximum average traffic that may be transmitted and received without packets being lost. But because data traffic displays inherently bursty behaviour a limit is also set on the maximum burst of packets that member institutions can transmit. So during a burst the bandwidth for the network service can be higher than the CIR temporarily. If a burst is too long, the policing mechanism will ensure that the average traffic speed remains in accordance with the CIR value. 3.3 Bandwidth setting on the institution side In order to optimise the end-to-end performance of a network, service it is advisable in most cases that the institution takes the agreed distribution of the MSP capacity across the various network services into account to ensure that no more traffic than agreed is transmitted across a light path so that no data are lost. One way that this can be achieved is to use traffic policing on the institution side also. Traffic policing is most efficiently and fairly applied at the source of a network service. As result of this, the traffic is not transported over the institution s network unnecessarily (where it may have a negative effect on other traffic) and the traffic meets the network service s CIR value. Another method is to adapt the traffic (shaping) to the agreed bandwidth for the network service. This method requires a combination of queuing and scheduling mechanisms. The various options are explained in figure 3. 7

8 Figure 2 Bandwidth setting on the campus side Option 1: Policing traffic at the point where the light path traffic is received on a switch. Most access switches have this type of ingress traffic policing functionality. Option 2: Same as option 1 but with a switch deeper in the campus network. Option 3: Use a QoS mechanism on the core switch that connects the institution to SURFnet. Core switches typically have more options for this. On the other hand, it can be more complicated to use these various capabilities correctly. If this option is chosen, the QoS configuration on the core switch should ensure that the distribution of available bandwidth across the various light paths matches the agreements made with SURFnet. The options above can differ per type of device. SURFnet is prepared to contribute ideas on which option best suits the design of your campus network. If you are interested in this, please contact your SURFnet adviser. 8

9 4 Bandwidth on Demand, virtual port on MSP Multiple virtual Bandwidth on Demand (BoD) ports can be created on each MSP. The administrator of the institution concerned creates these virtual BoD ports via the BoD portal Then an on demand light path can be set up between two virtual BoD ports. These virtual ports can be on any MSP in the SURFnet network. Each virtual BoD port has a fixed VLAN tag where the on demand light path offloads. An on demand light path can only be activated after a virtual BoD port has been created on the MSP. As long as the light path is active, an on demand light path has the same specifications as a fixed light path. 9

10 5 Network service reports A monthly report is available on the SURFdashboard for all network services that use an MSP. These reports show the availability and the traffic volume for the last month. A specific monthly report showing the availability and traffic volume will also be available for network services that terminate on an MSP. Furthermore, real-time traffic statistics for an MSP and an overview of the configured services on it are available via the SURFnet Network Dashboard. Read more about the SURFnet Network Dashboard. 10

11 Abbreviations 8bQ BoD CDP CIR EIR LACP LLDP MSP PBB-TE PAGP SSP IEEE standard describing single VLAN tagging Bandwidth on Demand: term for on demand light paths Cisco Discovery Protocol Committed Information Rate: data speed for packets on a network service where the bandwidth is guaranteed. Ingress traffic policing will label packets as green (CIR) or yellow (EIR) or drop them Excess Information Rate: data speed of packets in a network service where the bandwidth is above the CIR value but which are not yet dropped. This EIR bandwidth is not guaranteed end-2-end. Ingress traffic policing will label packets as green (CIR) or yellow (EIR) or drop them Link Aggregation Control Protocol Link Layer Discovery Protocol (IEEE) Multi Service Port: connection to the SURFnet network that enables you to set up multiple light paths simultaneously Provider Backbone Bridging with Traffic Engineering: Carrier Ethernet protocol that uses static configured tunnels to build predictable network services (latency, jitter, bandwidth) Port Aggregation Protocol, a Cisco proprietary network protocol, also known as EtherChannel Single Service Port: connection to the SURFnet network that enables you to set up a light path 11