Document 17673011

advertisement

1

U

SING

F

IGHT OR

F

LIGHT

M

ECHANISM

T

O

R

EDUCE THE

BGP C

ONVERGENCE

T

IME

E.A. Alabdulkreem

,

H.S. Al-Raweshidy

,

M.F. Abbod

Electronic & Computer Engineering/

School of Engineering & Design

Third Year PhD student

Keywords—BGP, Convergence time, Fight and Flight.

I.

INTRODUCTION

The internet consists of thousands of computer networks from all around the world [1]. Each network is typically known as a domain or an Autonomous System (AS), which is controlled by a common administrative entity. The Border

Gateway Protocol (BGP) is the de facto routing protocol between various AS types in the internet [2]. This protocol allows us to receive email, surf the web, and connect to any available internet data within milliseconds. That is why it is considered the glue that binds networks together in the internet.

For any routing protocol like the BGP, an important metric is convergence time, which is the time required to reroute the packet after a failure. The first significant researches were carried out about the BGP convergence time using the internet measurements that showed the BGP convergence time was slow [3]. This slow convergence is because of the fact that, for a single link failure the global internet will be forced to exchange a large number of advertisements while exploring a new path.

When a BGP router comes up to the internet, it establishes connection with its neighbors, which are the other BGP

routers that

directly communicate with it, then it will download the entire routing table of each neighbor router.

When a BGP router detects a routing change, an update messages will be sent, the message could be announcing a new path or a withdrawal for a path that does not exist anymore.

The convergence time is the time required to reroute packets after a routing change. It has been found that the current convergence delay could stretch into more than hundreds of seconds and can lead to high packet drop rates [4]. One of the reasons for this delay is the path exploration process. This is because the path exploration process requires checking different parameters. The parameters are lowest local preference, shortest AS-path, lowest MED, eBGP or iBGP, metric to next hop, older router, and lower ID BGP router.

This paper aim to speed up this process by using the concept of the Fight or Flight responds mechanism, which will explain next.

TABLE I. HE ORIGINS PRIORITY IN THE SHORT–TERM AND

LONG-TERM T ABLE E XAMPLE

Short-term survival

Saliva flow decrees.

Food movement slows down.

Muscles will become more tens.

Deep and quicker breathing will accrue.

Heart will beat faster and harder.

Digestive enzymes decrease.

Blood pressure increase.

Long-term survival

Saliva flow increase.

Food movements speed up.

Muscles will relax.

Normal breathing will accrue.

Heart will beat normally.

Digestive enzymes increase.

Blood pressure goes back to normal.

II.

USING THE FIGHT OR -FLIGHT MECHANISM

The fight or flight response occurs in the presence of something that is terrifying, either mentally or physically. The human body is either in the long-term survival or short-term survival. In each case different origins will have the priority as seen in table.1.

This idea will used to modify the path exploration to speed up this process. When there is a withdraw message which mean that there is a failure. In the internet the failure is the threat that will change the situation to short-term survival, and that lead to ignore some parameters and concentrate on just the older BGP router and the BGP router ID. When there is an announcement message, which means that there is a new path in the internet. This is not an emergency situation, so it will be a long-term survival, and in this case all the parameters are considered. Figure 1 shows the flow chart for the original BGP path exploration and the modification suggested by this paper.

III.

R ESULTS AND D ISCUSSION

The new algorithm was implemented using OPNET 16 which is a powerful simulation tool. Figure 2 shows the result with the x-axis displaying the time, as the scenario took 20 min, starting at 19:30 and stopped at 19:50. The y-axis displayed the number of seconds took the BGP protocol to converge.

After 10 minutes At 19:40 when one of the links failed the

BGP protocol needed to converge. The original BGP as we can see in the figure, the blue curve, has reached 32 seconds to

converge. The original BGP converge in 73 seconds, while the modified BGP the convergence curve in red reached 45 seconds. This shows that the convergence time is less by 35%.

Incoming update

Message = withdraw

Yes

BGP table

Pick highest

No

No

No o same weight?

Yes

Trash

Pick highes t

Do they have the same local preference?

Pick route

Pick shortes t

Pick lowest

Pick lowest

Pick eBGP

Yes

No

No

No

Yes

Yes

Is one of the paths originated by this router

No

Are the ASpaths having same length?

Yes

Is the origin code the same?

Yes

Is the MED value the same?

Yes

Is one of the bathes from eBGP?

No

Pick shortest

Pick oldest

No

Yes

Does paths to the next hope equal

Is one of the routers older?

Yes

No

Select the lower

BGP router ID

Figure.1. Original and modified (in red )BGP path exploration flow chart.

IV.

CONCLUSION

This paper aimed to modify the path exploration process to maximally reduce the convergence time. The Fight or Flight response mechanism were used to speed up the process. This is done by ignoring most of the parameters in case the update

2 message is a withdraw as this is mean that there is a failure, and this is considered to be an emergency.

Original BGP

Modified BGP

Time figure.2. Comparison between the convergence duration with the original

BGP and the optimized BGP .

On the other hand, take all the parameters in consecration when the message is an announcement, as this mean there is a new path.

The main conclusion is that with this modification, the convergence time was reduced by 35% as seen in figure 2.

This result proves how sensitive the BGP is to minor modification of some of its parameters.

V.

F UTURE P LAN / D IRECTIONS

Future work will be related to the route flap damping which refers to the roots that change too often. There should be a technique that copes with routers that regularly advertise and shortly after withdraw their routes. There is a need to reduce the number of BGP advertisements exchanged, without increasing BGP convergence time. Also route flap damping mechanism should significantly exacerbate the convergence times of relatively stable routes because a route to a prefix that is withdrawn exactly once and re-announced can be suppressed for up to an hour.

R EFERENCES

1.

G. Huaming, S. Wei, Z. Hongke, and K. Sy-yen, “On the convergence condition and convergence time of BGP”.

Elsevier , vol. 34, no. 2, pp. 192-199, 2011.

2.

Y. Rekhter, and T. Li, “A Border Gateway Protocol 4

(BGP-4) RFC 1771”, 1995.

3.

C. Labovitz, A. Ahuja,, A. Bose, and F. Jahanian,

“Delayed Internet Routing Convergence”, IEEE/ACM

Transactions on Networking, vol. 9, no. 3, pp. 293-306,

2001.

4.

T. Griffin, and B. Presmore, “An Experimental Analysis of

BGP Convergence”, IEEE ICNP , pp. 53-61, 2001.

Download