Performance Analysis of Leaky Buckets Integrated in ATM Networks

Abstract

Pada makalah ini dibahas analisis kinerja leaky bucket terintegrasi pada jaringan ATM dengan kombinasi control adaptif, metode marking dan metode prioritas. Juga ditinjau kinerja leaky bucket hasil dimensioning. Sebagai ukuran kinerja digunakan peluang sel hilang di jaringan dan throughput jaringan. Trafik masukan diasumsikan model 2-phase MMBP. Penelitian serupa yang telah dilakukan oleh X. Wu menggunakan model trafik masuk batch traffic dengan prores kedatangan Poisson dan panjang batch terdistribusi geometris. Hasil simulasi menunjukkan peluang sel hilang di jaringan makin berkurang dengan ukuran token pool K yang makin kecil; berkurangnya laju pembangkitan token mengakibatkan bertambahnya throughput jaringan; panjang burst rata-rata trafik masukan yang semakin besar akan memperbesar hilangnya sel di jaringan; dan kemudian, kinerja leaky bucket makin meningkat dengan digabungkannya mekanisme pengontrol kongesti lainnya seperti pengontrolan adaptif, marking, dan prioritas. Untuk leaky bucket hasil dimensioning, simulasi antara lain menunjukkan bahwa jika laju kedatangan sel conforming (sesuai dengan kontrak trafik), maka peluang sel hilang di jaringan makin kecil jika token pool makin kecil. Untuk (M+K) yang sama, leaky bucket dengan buffer menghasilkan peluang sel hilang di jaringan lebih kecil dibandingkan dengan tanpa buffer. Leaky bucket dengan buffer M besar dan pembangkitan token N besar menghasilkan peluang sel hilang di jaringan terkecil. Jika laju kedatangan sel makin jauh dari conforming, leaky bucket dengan N besar menghasilkan throughtput jaringan yang lebih besar.

Performance Analysis of Leaky Buckets Integrated in ATM Networks

This paper is an analysis of the performance of leaky buckets, which are integrated within ATM networks combined with adaptive control, marking and prioritizing. Also discussed is the performance of leaky bucket dimensioning. This performance is evaluated through network cell loss probabilities and network throughput. The arrival process is a 2-phase MMBP. Similar research has been done by Wu [10,11] using incoming traffic model batch traffic with a Poisson arrival process and a geometrically distributed batch length. The result of the simulation shows that the network cell loss probability is reduced as the token pool size K becomes smaller and smaller, the decrease of the token generation rate causes an increase in the network throughput; the network cell loss probability is higher in the case of a larger mean burst length; and the performance of the leaky bucket can be increased by incorporating other control mechanisms such as adaptive control, marking and prioritizing. For the leaky bucket dimensioning, the results of the simulation show, among other things, that if the incoming cell rate is conforming (appropriate to traffic contract) and the token pool is reduced, then the probability of network cell loss is also reduced. If (M + K) are the same, the buffered leaky bucket results in a network cell loss probability is smaller than for an unbuffered leaky bucket. The leaky bucket with big buffer M and big token generation N result in the smallest network cell loss probability. If the incoming cell rate is far from conforming, the leaky bucket with big N results in a bigger network throughput.