By:

RAKESH RANJAN

An Internet withoutThe Internet Protocol

What is the need to think in this way.

Due to growth in the internet it brought many challenges.

Cache poising attacks.

Growth of routing table at alarming rates.

IPv4 address space exhaustion.

This new scheme will completely remove the IP address instead it we will use here locators and identifiers.

Thus this new models is combination of locator-identifier.

Introduction

The main reason behind this concern- Scalability of routing.-routing table growth at alarming rate-

due to which today’s router may soon unable to meet demand.

IP address exhaustion.

However, there is a solution for address exhaustion that is IPv6 ,but before it come in practical its demerit came.

What are the demerit of IPv6- Although ipv6 came to give solution for ipv4,it simply

increases the address size .it doesn’t address the cause of this increasing such as- multi-homing, load balancing.

Ipv6 requires greater memory and more processing time than IPv4 forwarding, increasing load at the routers.

What would be an internet without internet protocol look likes ?

To lead this answer, here two observation are-

Minimalism in inter-domain routing.

Minimalism in end-host identification.

Minimalism in inter-domain routing…..Due to result of ipv4 address space exhaustion it fuels prefix growth and with this many prefix being added to the routing tables.so we left the use of IP prefixes in inter-domain routing and propose to use ASN in routing announcements and packet forwarding.

Reasons------------ It reduces the forwarding table sizes.

It allows fixed length looks-up during packet forwarding which is faster than the longest prefix match currently performed on IP prefixes.

It is found in many research that if we want fast internet than we have to break the present combination of locator and identifier

Minimalism in end-host identification…. Here in this we will eliminate IP address as identifier and

leaving host names as sole identifier.

With the use of ASN,we will find that this ASN posses properties that can enhance packet forwarding speeds as well as scale internet routing.

Using host names as identifiers reduces the requirements for the DNS, aiding caching and reducing the viability of many cache poising attacks.

When it has been examined, the causes of this growth are……

Routers failure to aggregate prefixes that can be aggregated.

Address fragmentation

Load balancing

Multi-homing.

Growth in inter-domain routing tables.

NIMROD , LISP , eFIT ,ENCAPS, ISLAY

These all aim to scale internet by using locators.

What are the basic idea behind this……. Encapsulating each packets in a special wrapper.

Way of Forwarding of packets in the core of internet.

Locator-identifier split proposals

An example.Suppose a client (host1.isp.com) in ASN 1000,wants to communicate with a server, www.website.com , in ASN 2000. the following sequences of steps will occur:

First, the client which is configured with its own ASN information, contacts its DNS resolver to get the ASN for www.website.com. The client then creates a packet with a header containing host1.isp.com and www.website.com as source and destination addresses and 1000 and 2000 as source and destination ASN locators. The packet is sent to the default router 1.

Router 1 forwards the packets by looking up 2000 in its forwarding table and exchanges ASN reachability rather than prefix reachability. Subsequent core routers, routers 2,3 and 4, repeat the forwarding table look-up.

Architecture details

When the packet arrives at router 5, the router recognizes its own ASN locator and forwards the packet using the destination host name www.website.com .

To reply to the client, the server simply reverses the source and destination host names and ASN locators.

This architecture differ from the modern internet

as:-Here client contact the DNS resolvers only for

the first packet of the connection.

DNS response packets return only ASN instead of IP.

Packet header here contain source and destination host name and ASN locators

Further the routers forward packets based on ASNs.

Header design

In this architecture there are two components which are:

A layer with inter-domain routing locators.

A layer for end-host identification which will be used by intra-domain routing protocols.

In IP intra-domain routing, IP prefixes are used to direct a packet to the appropriate subnet, at which point the packet is forwarded using link layer information. In this approach, we can map the packets from their host names to their subnet and MAC addresses at an organizational name server.

Hence, here we are mapping host name directly to their subnet and MAC addresses.

Thus optimization i.e. to minimize forwarding and look-up time or requirements takes place in this end-host architecture.

Intra-domain routing optimizations

Here DNS look-up is divided into two….. The host name to ASN mapping takes place in a

regular DNS look-up.

Host name to subnet ID and MAC mapping is confined to the destination AS.

This separation of mapping infrastructure allows DNS resolvers to perform extensive, a long-term caching for host name to AS mapping which are fairly static.

This allows an authoritative name server to load balance a host name across systems.

Some changes requires in Intra-domain protocols for this structure.

In DHCP,DHCP server must communicate the network information to the client. The server must provide at least one DNS host name, a subnet ID and the encompassing ASN to the client. The DHCP server must communicate this mapping information along with the client’s MAC to the authoritative DNS server.

Second, the changes to ARP protocol are straight forward.

Intra-domain protocols

At present there are 42 different record types. This include to find mail servers (MX) , authoritative DNS servers for the domain, records to map host name to IP addresses and records to map IP addresses to host names etc.

But in this architecture, DNS requirement is simpler as only one type of records is required to map host names into routing locators.

Example- “ Example.com X 24 “

This simplicity has two outcomes…… Very few records have to maintain.

Since one records is maintained its thwarts statistical and related data attacks to poison the DNS cache.

DNS issues

Growth in ASNsToday, there are fewer ASN than IP prefixes.

A research has been done by Huston and found the main cause of this growth is – multi-homing.

And to avoid this multi-homing, in this architecture, an organization can simply rank each of its providers as primary and secondary etc.

Today almost one-third of total ASNs are completely for multihoming

Thus we can eliminate these unnecessary ASNs .

Examining ASN as locators

Today, IP prefixes of variable length and router perform a longest prefix match whereas in this approach fixed length ASNs is used.

When it examined the impact of this factor on packet forwarding speed. We find that ASNs based packet forwarding makes look-up and update operations faster at the same time requires one third of the memory.

Modern software routers use the trie data structure to perform longest prefix matching on IP prefixes.

A trie must perform O(log(n)) memory references. Where n=no. of bits in IP address.

ASN-based routing yield performance of O(1),since ASNs are fixed length and exactly one match has to be found.

Forwarding table look-up performance

Performance of ASN-based and IPv4 forwarding.

Forwarding approach ASNs IPv4

Number of entries 101,310 288,685

Storage required (MBytes) 2.90 9.73

Lookup times (ns)

Average 155 1129

Standard dev. 40 253

Minimum 133 543

Update times (ns)

Average 157 4018

Packet header growth Using host names as end-host identifier impacts DNS only in

positive

To determine this overhead an observation has been done…. Total of 7,90,88,314 unique domain from 128 million total

domain has been taken in research and the observation found as..

Maxm-255( separated by ”.” could be 63)

But here found the maximum of 67 character with a median of 15 character long. If hosts within domain follow a similar pattern then this median goes to 30 character.

Further there is a restriction on DNS character set 6 bit required to encode each character resulting in 23 bytes and one bytes to encode the name’s length. Thus total of 24 bytes.

Examining host names as identifiers

Hence host address

would be 24 bytes which

is 50 % larger then IPv6

address.

Further if subnet id and

MAC are encoded 7

bytes more thus total of

31 bytes

These addresses can be

easily fit within modern

packet size limitations.

In this architecture, once traffic reaches the destination network, it must be forwarded based on its host name.

Packet forwarding would use three identifier…….

The host name, which a name server must be able to quickly map to a subnet ID and MAC,

A subnet ID which must be quickly routed

A MAC address for transmitting the packet at the last router.

In this approach, this step is even less work: the servers would only store one record type, requiring only about 8 bytes(2 byte for subnet ID and 6 for MAC)

Routing on subnet IDs and MACs is straightforward for routers.

For 16 bit subnet ID direct array indexing can be used which require 384 kb for 65,536 entries. And for MAC routers must simply copy to link layer.

Intra-domain routing scalability

IP addresses have been a cornerstone of the Internet for as long as we have known the TCP/IP-based Internet.

Both host names and autonomous system numbers (ASNs) were added later: host names were added to provide users with a mnemonic way of addressing machines and ASNs were added to make BGP loop free.

This topic explored a new Internet which breaks away from IP addresses and instead embraces names host as identifiers and ASNs as locators.

This design decouples routing from addressing, which IP addresses (inadvertently) entangled.

The outcome is a faster, expandable, and more scalable Internet.

We outlined the key features of this architecture and justified the choices using actual data sets from the Internet.

While there is still more work required to test the feasibility and to make the architecture practical, we hope that this topic will continue the discussion of the future of the Internet in the research community.

Conclusion

Thank you