Lecture 15: Addressing and Routing Architecture

Transcription

1 Lecture 15: Addressing and Routing Architecture Prof. Shervin Shirmohammadi SITE, University of Ottawa Prof. Shervin Shirmohammadi CEG Addressing & Routing Addressing is assigning identifiers to devices. These identifiers can be local or global, private or public, temporary or persistent. Routing consists of learning about the reachability within and between networks and applying this reachability to forward packets in the network. Together, they form a complete picture of network connectivity. Prof. Shervin Shirmohammadi CEG

2 Addressing Fundamentals IP Addressing uses a combination of Address Identifier and Mask The mask is used to separate the address into a network and host function. This is very important in the distinction between local and remote parts of the network. E.g.: Address Identifier XOR Mask == Subnet Which means is on subnet Prof. Shervin Shirmohammadi CEG Local or Global Type of Addresses Local communication addresses like link-layer (MAC Address). Not advertised outside of the local network: there's no point since there is no link-layer connectivity between non-local devices. Global addresses are required for devices outside of the local broadcast region like IP addresses. Private or Public Both are global addresses, but private addresses are not advertised and forwarded (on purpose) while public addresses are. Temporary of Persistent Temporary are usually assigned using DHCP while persistent addresses are assigned either manually or are hardcoded (like Ethernet address carved into an Ethernet network card. Prof. Shervin Shirmohammadi CEG

3 Local vs. Remote Network /16 Network /16 Other Network Devices on the same subnet are directly connected and therefore, for IP, address resolution is done at different layer (MAC) than that done at the routing layer (IP) In communicating to devices on other networks there must be a router connecting the networks Prof. Shervin Shirmohammadi CEG Explicit Routing Company A Packets to are routed here. Company B ISP X Route to ISP Z Internet ISP Y Route to Routing Table / / Prof. Shervin Shirmohammadi CEG

4 Addressing Mechanisms Classful addressing older style of addressing Subnetting A better way to distribute addresses Variable-length subnetting Even more refined than subnetting Supernetting and Classless interdomain routing (CIDR). An efficient way to advertise addresses, and currently used on the Internet. Private addressing and Network Address Translation (NAT). A way to re-use certain IP addresses without collision with the rest of Internet Prof. Shervin Shirmohammadi CEG Classful Addressing Outdated form of addressing offers a simplistic solution for addressing schemes. Based on pre-determined mask lengths where: Class A = Mask (127 Networks & over 16M Addresses/Network), First Octet Range Class B = Mask (16K Networks & 64K Addresses/Network), First Octet Range Class C = Mask (2M Networks & 254 Addresses/Network), First Octet Range Class D = Multicast address Class E is reserved. Prof. Shervin Shirmohammadi CEG

5 Limits to Classful Addressing Very few Class A and B addresses, and all have already been allocated. That leaves class C to allocate new addresses Many networks require more addresses than class C but fewer addresses than B offers. On the other hand many organizations with A or B cannot use all of the networks offered by class A or B. That has led to variable-length subnets Prof. Shervin Shirmohammadi CEG Subnetting Allows a classful network address to be segmented into smaller sections by using part of the device address to create another level of hierarchy. Basically it takes address space away from the devices and gives it to the network. Useful for Internal addressing and routing Allows you to assign subnets to specific buildings, or specific groups, hence localizing traffic and simplifying routing. Has no effect on external routing The hierarchy is not revealed to the outside world. Address Identifier XOR Mask == Subnet Prof. Shervin Shirmohammadi CEG

6 Subnetting for Class B Network 2-Bit Mask Subnets Devices/Subnet 3-Bit Mask Subnets 8190 Devices/Subnet Class B Network 64K Devices 4-Bit Mask Bit Mask Bit Mask Bit Mask Subnets 4094 Devices/Subnet 31 Subnets 2046 Devices/Subnet 63 Subnets 1022 Devices/Subnet 127 Subnets 510 Devices/Subnet 8-Bit Mask Subnets 254 Devices/Subnet Notice that all zeros are not allowed as either subnet part or as host part. Also, all ones are not allowed for the host part. Prof. Shervin Shirmohammadi CEG Example A company has bought IP class address It has 14 departments and it wants to give each its own subnet. What will be each of the subnets, and their subnet mask? This is a class B address so we subnet into the 3 rd octet. To have 14 subnets, we require 2^4 = 16 subnet divisions, so we play with the first 4 bits in the third octet. The mask will be , and subnets are: Prof. Shervin Shirmohammadi CEG

7 Variable-length Subnetting Subnetting divides the network into a number of equal-sized subnets which is often inefficient. Variable-length subnetting is subnetting in which non-equal or variable length subnets are used. E.g., an organization with Class B address has a number of workgroups divided as shown below: Workgroup Engineering Marketing Administration Sales R&D Support Total Groups Size/Group (Devices) 400 (1200) (1350) Prof. Shervin Shirmohammadi CEG Classful and Subnetting Solution Classful solution gives us 65,534 devices. That s enough devices, but putting all departments (i.e., 5730 devices) in the same subnet is neither scalable nor manageable. Subnetting solution: let s have 1 subnet per group. We have 43 groups, so we need 2^6 1 = 63 subnets, which means we can use 6 bits for the subnet part, and the remaining 10 bits for hosts. But 10 bits for hosts part gives us 2^10 2 = 1022 devices, and marketing has 1950 devices. What to do? Use variable length subnetting. Prof. Shervin Shirmohammadi CEG

8 Variable-length Subnetting Solution We can resolve this by using a combination of 4-bit and 8-bit masks. 4-bit mask gives us 15 subnets each with 4096 devices. That s enough for Engineering and Marketing. We use the first five subnets as follows: Engineering (3 subnets) -> , , Marketing (1subnet) -> Administration (1 subnet) -> Why Admin? 8-bit mask gives us 255 subnets and 254 devices. That s fine for Sales, R&D, & Support. We use as many as the remaining 10 subnets as needed, breaking them into sub-subnets. For example, the next subnet, can be broken into another 15 subnets from to All these 15 will go to Sales. We still need another 23 subnets: 1 for R&D and 22 for Support. For these, we break the next two subnets, and Prof. Shervin Shirmohammadi CEG Supernetting Supernetting is the concept of aggregating network addresses by changing the network mask to decrease the number of bits recognized as the network part. Millions of Class C addresses can be allocated in lieu of Class A & B. The result is that too many Class C address groups need to be allocated to an organization and advertised among all the Internet routers. The number of routes would grow exponentially such that some experts had predicted that the Internet would collapse by Obviously this did not happen, since supernetting was invented. Say a company needs to support 10,000 devices. A class C address supports up to 254 devices, so 40 class C networks are needed. How are we to advertise these 40 class C addresses? Prof. Shervin Shirmohammadi CEG

9 Supernetting Technique If we take a set of 16 contiguous addresses from a Class C address like we can see that the first 4 digits of the subnet octet do not change. This range of values can be represented as with a subnet mask of where the last 4 bits in the third octet are ignored. This then can be used to advertise a group of addresses as /20 which means addresses from > Prof. Shervin Shirmohammadi CEG Classless InterDomain Routing (CIDR) The concept of supernetting suggested that indeed we do not need class boundaries, since each group can advertise its own subnet mask too. This in effect lead to classless Classless InterDomain Routing (CIDR). Addresses must be assigned in contiguous blocks following logical topology. The number of addresses in a CIDR block are powers of 2. Network Prefix can be anything, and need not be a power of 2. It is transmitted along with address Used in conjunction with classless routing protocols (e. g. EIGRP, OSPF) E.g.: /22 advertises 4 networks: 240, 241, 242, and /23 advertises 2 networks: 240 and /24 (this is the natural mask for class C) advertises 1 network: /21 advertises 8 networks: 240 to /17 is equivalent to a range of 2 7, or 128, networks from > Prof. Shervin Shirmohammadi CEG

10 Classful vs. CIDR Classful router must advertise all 4 nets Rtr 1 Classless router only advertises one I m router 1 and I know how to get to networks: Rtr 1 I m router 1 and I know how to get to networks: / Prof. Shervin Shirmohammadi CEG bits Private Addresses and NATs Private IP Addresses are reserved addresses that can t be forwarded to the Internet > (10/8 prefix) > (172.16/12 prefix) > ( /16 prefix) Pros: Makes changing ISP easier Increases security Cons: Outsourcing management may be difficult Mergers may require renumbering Network Address Translation: translates private addresses <-> public addresses A binding is created between the addresses that lasts a period of time. Can be implemented in Router, Firewall, or Specialized device. Prof. Shervin Shirmohammadi CEG

11 Routing Static routing The reachability is entered manually to the router. Method we commonly use for our small networking labs. Dynamic routing More typical of a real network. Typical routing protocols are: RIP/RIPv2, OSPF, and BGP4 Destination is determined by looking at the network portion of the packets destination address and choose the best destination (one with the more explicit route. i.e. the more specific). We now consider a routing mechanism that consists of: Establishing routing flows Identifying and classifying routing boundaries Manipulating routing flows. This will be based on the flow analysis process that leverages the flow specification and flow map discussed in lecture 9. Prof. Shervin Shirmohammadi CEG Establishing Routing Flows Segment the network into functional areas and workgroups. Identify boundaries between these areas. Form relationships between boundaries and routing flows. A functional area consists of groups within the system that share a similar function. These may consist of users (workgroups), applications, devices, or combinations of these and they may share similar jobs, locations, functions within the network (backbone routing). Workgroups are groups of users that have common locations, applications, and requirements, or that belong to the same organization. Prof. Shervin Shirmohammadi CEG

12 Example of Workgroups & FAs FA1 Bldg A Routers FA2 Scientists Accounting FAb1 FAb3 FA3 FA4 WG1 Management WG2 Bldg B Scientists Bldg C Prof. Shervin Shirmohammadi CEG Routing Boundaries These are physical or logical separations of a network based on requirements or administration of the network. Physical Boundaries can de identified by isolated LANs, DMZs, physical interfaces on network equipment, physical security. Logical Boundaries can be identified by the FAs, WGs, administrative domains (Autonomous Systems AS), and routing management domains. Prof. Shervin Shirmohammadi CEG

13 Hard Boundaries These boundaries are routing boundaries in which EGPs are predominantly used: Exterior Gateway Protocols (EGPs) communicate between AS s or AS and external network. DMZs and interfaces to ISP. Interior Gateway Protocols (IGPs) communicate within an AS. Your AS/Administrative Domain DMZ Internet Hard Boundary Prof. Shervin Shirmohammadi CEG Soft Boundaries Typically found within a single AS and are usually placed at the junction of FAs and WGs Your AS / Administrative Domain FA1 WG1 WG1 WG1 WG1 FA3 WG2 WG1 FA2 WG1 FA4 Prof. Shervin Shirmohammadi CEG

14 Internet Routing The Internet uses hierarchical routing The Internet is split into AS s AS corresponds to an administrative domain Assign each AS a 16-bit number Examples: University, company, backbone network Stanford (32), Sprint (1239), MCI Worldcom (17373) Within an AS, the administrator chooses an Interior Gateway Protocol (IGP) Examples of IGPs: RIP (RFC 1058), OSPF (RFC 1247) Between AS s, the Internet uses an Exterior Gateway Protocol AS s today use the Border Gateway Protocol, BGP-4 (RFC 1771) Prof. Shervin Shirmohammadi CEG Why different Intra- and Inter-AS routing? Policy: Inter-AS: admin wants control over how its traffic is routed who routes through its net. Intra-AS: single admin, so no policy decisions needed Scale: hierarchical routing saves table size, update traffic Performance: Intra-AS: can focus on performance Inter-AS: policy may dominate over performance Prof. Shervin Shirmohammadi CEG

15 Boundaries & Routing Flows Routing Flows are flows of routing information passed between FAs and ASs. Hard Boundary FA4 FA5 External Networks FA 1 FAb1 FAb2 These are important for the architecture and design because routing flows can be manipulated at routing boundaries. Prof. Shervin Shirmohammadi CEG FA2 FA3 Routing Flows AS Manipulating Routing Flows Controlling routing flow in a network is vital to the proper operation and performance of the network. This involves determining the proper combination of addressing and routing. Techniques: Default Routing Route Filtering Route Aggregation Policies & Policy Enforcement Points Default Route is the route used when there is no other route. Generally the route with the highest capacity to the network. Route Filtering is a technique to hide networks from the rest of the AS. Implemented as a rule (if IPPacketDest = then DropPacket) Route Aggregation is a technique to exchange routing between AS s Policies allow AS to accept or deny traffic, Prof. Shervin Shirmohammadi CEG

16 Addressing Strategies When addressing, we need to keep in mind the future scaling requirements. Area of Network Addressing Scheme Enterprise Wide Functional Areas Work Groups Networks Hosts Supernetting (CIDR) Natural Class Subnetting Variable-Length Subnetting Prof. Shervin Shirmohammadi CEG Example of Variable-length Subnetting Hub router can interconnect up to 10 networks. WG routers can support 4 networks each with 10 to 20 devices CIDR block /20 AS WG1 ISP Router Hub Router WG2 WG3 ISP WG4 WG5 Prof. Shervin Shirmohammadi CEG

17 Solution ISP Router ISP AS Hub Router /30 63 subnets 2 devices/subnet /27 6 subnets 30 devices/subnet Prof. Shervin Shirmohammadi CEG