ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let Un,k(t)U_{n,k}(t) denote the scaled fraction of queues with at least kk customers, let Fn,k(0)F_{n,k}(0) denote the corresponding initial tail fractions, and let uk(t)u_k(t) be the fluid-limit functions. Fix k≤Kk\le K, where KK is defined in the paper, and suppose that d(n)d(n) is the buffer-size parameter. Assume

lim⁡n→∞nd(n)=∞.\lim_{n\to\infty}\frac{n}{d(n)}=\infty.

Assume that Un,k(0)U_{n,k}(0) is deterministic for all nn and k≤Kk\le K, and that there exist v1,…,vK∈R+v_1,\dots,v_K\in\mathbb{R}_+ and v1∗,…,vK∗∈Rv_1^*,\dots,v_K^*\in\mathbb{R} such that

lim⁡n→∞nd(n)k(Un,k(0)−vk)=vk∗.\lim_{n\to\infty}\sqrt{\frac{n}{d(n)^k}}\left(U_{n,k}(0)-v_k\right)=v_k^*.

Additionally, assume

lim⁡n→∞n d(n)K+1(Fn,K+1(0)+Fn,K+2(0)+⋯ )=0.\lim_{n\to\infty}\sqrt{n\,d(n)^{K+1}}\left(F_{n,K+1}(0)+F_{n,K+2}(0)+\cdots\right)=0.

Generalized diffusion-limit conjecture. As n→∞n\to\infty,

nd(n)k(Un,k(t)−uk(t)+1d(n)uk+1(t))⇒Zk(t),\sqrt{\frac{n}{d(n)^k}}\left(U_{n,k}(t)-u_k(t)+\frac{1}{d(n)}u_{k+1}(t)\right)\Rightarrow Z_k(t),

where uK+1(t)u_{K+1}(t) is interpreted as zero. The process Z1(t)Z_1(t) is the solution of

Z1(t)=v1∗+λ B1(1)(t)−∫0t1−e−u1(s) dB1(2)(s)−∫0te−u1(s)Z1(s) ds,Z_1(t)=v_1^*+\sqrt{\lambda}\,B^{(1)}_1(t)-\int_0^t\sqrt{1-e^{-u_1(s)}}\,\mathrm{d}B^{(2)}_1(s)-\int_0^t e^{-u_1(s)}Z_1(s)\,\mathrm{d}s,

and, for k=2,…,Kk=2,\dots,K, Zk(t)Z_k(t) is the solution of

Zk(t)=vk∗+∫0tλuk−1(s) dBk(1)(s)−∫0tuk(s) dBk(2)(s)−∫0tZk(s) ds,Z_k(t)=v_k^*+\int_0^t\sqrt{\lambda u_{k-1}(s)}\,\mathrm{d}B^{(1)}_k(s)-\int_0^t\sqrt{u_k(s)}\,\mathrm{d}B^{(2)}_k(s)-\int_0^t Z_k(s)\,\mathrm{d}s,

for independent Wiener processes Bk(1)(t)B^{(1)}_k(t) and Bk(2)(t)B^{(2)}_k(t). The conjecture concerns a proposed extension of the paper's proved diffusion limit: the different scaling orders for successive queue-length levels introduce complications for the proof, and the generalization remains unproved in the source.

Record

Comments

Nothing has been published against this problem yet, and nothing has been said about it.

Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.

Sign in

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.