2025-11-17T19:19:13.157995

A Framework for Distributed Resource Allocation in Quantum Networks

Panigrahy, Bacciottini, Hollot et al.
We introduce a distributed resource allocation framework for the Quantum Internet that relies on feedback-based, fully decentralized coordination to serve multiple co-existing applications. We develop quantum network control algorithms under the mathematical framework of Quantum Network Utility Maximization (QNUM), where utility functions quantify network performance by mapping entanglement rate and quality into a joint optimization objective. We then introduce QPrimal-Dual, a decentralized, scalable algorithm that solves QNUM by strategically placing network controllers that operate using local state information and limited classical message exchange. We prove global asymptotic stability for concave, separable utility functions, and provide sufficient conditions for local stability for broader non-concave cases. To reduce control overhead and account for quantum memory decoherence, we also propose schemes that locally approximate global quantities and prevent congestion in the network. We evaluate the performance of our approach via simulations in realistic quantum network architectures. Results show that QPrimalDual significantly outperforms baseline allocation strategies, scales with network size, and is robust to latency and decoherence. Our observations suggest that QPrimalDual could be a practical, high-performance foundation for fully distributed resource allocation in quantum networks.
academic

์–‘์ž ๋„คํŠธ์›Œํฌ์˜ ๋ถ„์‚ฐ ์ž์› ํ• ๋‹น ํ”„๋ ˆ์ž„์›Œํฌ

๊ธฐ๋ณธ ์ •๋ณด

  • ๋…ผ๋ฌธ ID: 2510.09371
  • ์ œ๋ชฉ: A Framework for Distributed Resource Allocation in Quantum Networks
  • ์ €์ž: Nitish K. Panigrahy, Leonardo Bacciottini, C. V. Hollot, Emily A. Van Milligen, Matheus Guedes de Andrade, Nageswara S. V. Rao, Gayane Vardoyan, Don Towsley
  • ๋ถ„๋ฅ˜: quant-ph (์–‘์ž๋ฌผ๋ฆฌํ•™), cs.PF (์ปดํ“จํ„ฐ ์„ฑ๋Šฅ)
  • ๋ฐœํ‘œ ์‹œ๊ฐ„: 2025๋…„ 10์›”
  • ๋…ผ๋ฌธ ๋งํฌ: https://arxiv.org/abs/2510.09371

์ดˆ๋ก

๋ณธ ๋…ผ๋ฌธ์€ ์–‘์ž ์ธํ„ฐ๋„ท์„ ์œ„ํ•œ ๋ถ„์‚ฐ ์ž์› ํ• ๋‹น ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ œ์•ˆํ•˜๋ฉฐ, ์ด๋Š” ํ”ผ๋“œ๋ฐฑ ๊ธฐ๋ฐ˜์˜ ์™„์ „ ๋ถ„์‚ฐํ™”๋œ ์กฐ์ •์— ์˜์กดํ•˜์—ฌ ์—ฌ๋Ÿฌ ๊ณต์กดํ•˜๋Š” ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์„ ์„œ๋น„์Šคํ•œ๋‹ค. ์–‘์ž ๋„คํŠธ์›Œํฌ ํšจ์šฉ ์ตœ๋Œ€ํ™”(QNUM) ์ˆ˜ํ•™ ํ”„๋ ˆ์ž„์›Œํฌ ํ•˜์—์„œ ์–‘์ž ๋„คํŠธ์›Œํฌ ์ œ์–ด ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ฐœ๋ฐœํ•˜๋ฉฐ, ์—ฌ๊ธฐ์„œ ํšจ์šฉ ํ•จ์ˆ˜๋Š” ์–ฝํž˜๋ฅ ๊ณผ ํ’ˆ์งˆ์„ ๊ฒฐํ•ฉ ์ตœ์ ํ™” ๋ชฉํ‘œ๋กœ ๋งคํ•‘ํ•˜์—ฌ ๋„คํŠธ์›Œํฌ ์„ฑ๋Šฅ์„ ์ •๋Ÿ‰ํ™”ํ•œ๋‹ค. ๊ทธ ํ›„ QPrimal-Dual์„ ๋„์ž…ํ•˜๋Š”๋ฐ, ์ด๋Š” ๋กœ์ปฌ ์ƒํƒœ ์ •๋ณด์™€ ์ œํ•œ๋œ ๊ณ ์ „ ๋ฉ”์‹œ์ง€ ๊ตํ™˜์„ ์‚ฌ์šฉํ•˜๋Š” ๋„คํŠธ์›Œํฌ ์ œ์–ด๊ธฐ๋ฅผ ์ „๋žต์ ์œผ๋กœ ๋ฐฐ์น˜ํ•˜์—ฌ QNUM์„ ํ•ด๊ฒฐํ•˜๋Š” ๋ถ„์‚ฐํ™”๋˜๊ณ  ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ์•Œ๊ณ ๋ฆฌ์ฆ˜์ด๋‹ค. ์˜ค๋ชฉ ๋ถ„๋ฆฌ ๊ฐ€๋Šฅ ํšจ์šฉ ํ•จ์ˆ˜์— ๋Œ€ํ•ด ์ „์—ญ ์ ๊ทผ ์•ˆ์ •์„ฑ์„ ์ฆ๋ช…ํ•˜๊ณ , ๋” ๊ด‘๋ฒ”์œ„ํ•œ ๋น„์˜ค๋ชฉ ๊ฒฝ์šฐ์— ๋Œ€ํ•ด ๊ตญ์†Œ ์•ˆ์ •์„ฑ์˜ ์ถฉ๋ถ„ ์กฐ๊ฑด์„ ์ œ๊ณตํ•œ๋‹ค.

์—ฐ๊ตฌ ๋ฐฐ๊ฒฝ ๋ฐ ๋™๊ธฐ

๋ฌธ์ œ ์ •์˜

์–‘์ž ์ธํ„ฐ๋„ท์€ ๋‹ค์–‘ํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์„ ๋ฐฐํฌํ•˜๋Š” ์ˆ˜๋งŽ์€ ์ข…๋‹จ ๋…ธ๋“œ๋ฅผ ์›ํ™œํ•˜๊ฒŒ ์„œ๋น„์Šคํ•˜๊ธฐ ์œ„ํ•ด ์ •๊ตํ•œ ํ•˜๋“œ์›จ์–ด ๊ตฌ์„ฑ ์š”์†Œ ํŽธ์„ฑ์ด ํ•„์š”ํ•˜๋‹ค. ์ „ํ†ต์ ์ธ ์ค‘์•™ ์ง‘์ค‘์‹ ์ž์› ํ• ๋‹น ๋ฐฉ๋ฒ•์€ ๋Œ€๊ทœ๋ชจ ๋˜๋Š” ๋™์  ๋„คํŠธ์›Œํฌ์—์„œ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ๋ฌธ์ œ๊ฐ€ ์žˆ๋‹ค:

  1. ๋‹จ์ผ ์žฅ์• ์ : ์ค‘์•™ ์ง‘์ค‘์‹ ์ œ์–ด๊ธฐ๊ฐ€ ์‹œ์Šคํ…œ ๋ณ‘๋ชฉ์ด ๋จ
  2. ์™„์ „ํ•œ ๋„คํŠธ์›Œํฌ ์ง€์‹ ์š”๊ตฌ: ์ „์—ญ ํ† ํด๋กœ์ง€ ๋ฐ ์„ธ์…˜ ์ •๋ณด ํ•„์š”
  3. ์ง€์—ฐ ๋ฏผ๊ฐ์„ฑ: ์†”๋ฃจ์…˜ ๋ฐฐํฌ ์ง€์—ฐ์œผ๋กœ ์ธํ•ด ์˜ค๋ž˜๋œ ๋„คํŠธ์›Œํฌ ์ƒํƒœ ์ดˆ๋ž˜ ๊ฐ€๋Šฅ

์–‘์ž ๋„คํŠธ์›Œํฌ ๊ณ ์œ ์˜ ๊ณผ์ œ

  1. ํ•˜๋“œ์›จ์–ด ์ œํ•œ: ๋ถˆ์™„์ „ํ•œ ์–‘์ž ์ €์žฅ์†Œ ๋“ฑ ๊ทผ๊ธฐ ์–‘์ž ์žฅ์น˜์˜ ์‹ฌ๊ฐํ•œ ์ œํ•œ
  2. ํ’ˆ์งˆ ๋ฏผ๊ฐ์„ฑ: ์–‘์ž ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์€ ์ œ๊ณต๋œ ์ƒํƒœ ํ’ˆ์งˆ์— ๋งค์šฐ ๋ฏผ๊ฐํ•จ
  3. ๊ณ ์ „ ๋ฉ”์‹œ์ง€ ์š”๊ตฌ: ์–‘์ž ํ†ต์‹  ๋ถ€ํ”„๋กœํ† ์ฝœ์— ๋” ๋งŽ์€ ์กฐ์ • ํ•„์š”

์—ฐ๊ตฌ ๋™๊ธฐ

๊ณ ์ „ ์ธํ„ฐ๋„ท์˜ ๋ถ„์‚ฐ ์„ค๊ณ„ ์›๋ฆฌ๋ฅผ ์ฐธ๊ณ ํ•˜์—ฌ ์–‘์ž ๋„คํŠธ์›Œํฌ์— ์ ์šฉ ๊ฐ€๋Šฅํ•œ ์™„์ „ ๋ถ„์‚ฐํ™”๋œ ์ž์› ํ• ๋‹น ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ๊ฐœ๋ฐœํ•˜๊ณ , TCP ํ”„๋กœํ† ์ฝœ๊ณผ ์œ ์‚ฌํ•œ ํ”ผ๋“œ๋ฐฑ ๋ฉ”์ปค๋‹ˆ์ฆ˜์„ ๊ตฌํ˜„ํ•œ๋‹ค.

ํ•ต์‹ฌ ๊ธฐ์—ฌ

  1. ๋ถ„์‚ฐํ™” QNUM ์•Œ๊ณ ๋ฆฌ์ฆ˜: ๋‘ ๊ฐ€์ง€ ์œ ํ˜•์˜ ์ƒํ˜ธ์ž‘์šฉ ์ œ์–ด๊ธฐ๋ฅผ ๋ฐฐ์น˜ํ•˜์—ฌ QNUM ์ตœ์ ํ™” ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๋Š” QPrimal-Dual ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ œ์•ˆ
  2. ์ด๋ก ์  ์•ˆ์ •์„ฑ ๋ณด์žฅ: ์˜ค๋ชฉ ํšจ์šฉ ํ•จ์ˆ˜์— ๋Œ€ํ•œ ์ „์—ญ ์ ๊ทผ ์•ˆ์ •์„ฑ ์ฆ๋ช…, ๋น„์˜ค๋ชฉ ํ•จ์ˆ˜์— ๋Œ€ํ•œ ๊ตญ์†Œ ์•ˆ์ •์„ฑ ์กฐ๊ฑด ์ œ๊ณต
  3. ์‹ค์ œ ๊ตฌํ˜„ ๋ฐฉ์•ˆ: ์ˆœ์ฐจ ์–‘์ž ๋„คํŠธ์›Œํฌ์—์„œ ์•Œ๊ณ ๋ฆฌ์ฆ˜์˜ ์‹ค์ œ ๊ตฌํ˜„ ํ”„๋กœํ† ์ฝœ ๊ฐœ์š” ์ œ์‹œ
  4. ํ™•์žฅ ๋ฐฉ์•ˆ: ์ œ์–ด ์˜ค๋ฒ„ํ—ค๋“œ ๊ฐ์†Œ ๋ฐ ์–‘์ž ์ €์žฅ์†Œ ์œ„์ƒ ์†Œ์‹ค์— ๋Œ€์‘ํ•˜๋Š” ๋ฐฉ์•ˆ ์ œ์•ˆ

๋ฐฉ๋ฒ•๋ก  ์ƒ์„ธ ์„ค๋ช…

์ž‘์—… ์ •์˜

์–‘์ž ๋„คํŠธ์›Œํฌ ๊ทธ๋ž˜ํ”„ G = (V, L)์—์„œ ์—ฌ๋Ÿฌ ์–ฝํž˜ ์„ธ์…˜์— ์ž์›์„ ํ• ๋‹นํ•˜๋ฉฐ, ๊ฐ ์„ธ์…˜ r โˆˆ R์€ ๋…ธ๋“œ ์Œ(Ar, Br)์— ํ•ด๋‹นํ•œ๋‹ค. ๋ชฉํ‘œ๋Š” ์ง‘๊ณ„ ํšจ์šฉ โˆ‘rโˆˆR Ur(Rr, Fr)์„ ์ตœ๋Œ€ํ™”ํ•˜๋Š” ๊ฒƒ์ด๋ฉฐ, ์—ฌ๊ธฐ์„œ:

  • Rr: ์„ธ์…˜ r์˜ ์ข…๋‹จ ๊ฐ„ ์–ฝํž˜๋ฅ 
  • Fr: ์„ธ์…˜ r์˜ ์ข…๋‹จ ๊ฐ„ ์ถฉ์‹ค๋„

QNUM ์ตœ์ ํ™” ๋ฌธ์ œ

QNUM: max โˆ‘rโˆˆR Ur(Rr, wโƒ—r)
์ œ์•ฝ ์กฐ๊ฑด:
โˆ‘r:lโˆˆr Rr โ‰ค dl(1-wl), โˆ€l โˆˆ L     (์šฉ๋Ÿ‰ ์ œ์•ฝ)
โˆ‘l:lโˆˆr log wl โ‰ฅ Kr, โˆ€r โˆˆ R        (์ตœ์†Œ ์ถฉ์‹ค๋„ ์ œ์•ฝ)
0 โ‰ค wl โ‰ค 1, โˆ€l โˆˆ L                (Werner ๋งค๊ฐœ๋ณ€์ˆ˜ ๋ฒ”์œ„)
Rr โ‰ฅ 0, โˆ€r โˆˆ R                    (๋น„์Œ์ˆ˜ ์œจ ์ œ์•ฝ)

QPrimal-Dual ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์•„ํ‚คํ…์ฒ˜

๋ผ๊ทธ๋ž‘์ฃผ ํ•จ์ˆ˜

A(Rโƒ—, wโƒ—, ฮปโƒ—, ฮผโƒ—) = โˆ‘rโˆˆR Ur(Rr, wโƒ—r) 
                   - โˆ‘l ฮปl[โˆ‘r:lโˆˆr Rr - dl(1-wl)]
                   - โˆ‘r ฮผr[Kr - โˆ‘l:lโˆˆr log wl]

์—…๋ฐ์ดํŠธ ๊ทœ์น™

  1. ๋งํฌ ๊ฐ€๊ฒฉ ์—…๋ฐ์ดํŠธ:
    ฮปฬ‡l(t) = [โˆ‘r:lโˆˆr Rr(t) - dl(1-wl(t))]
    ฮปl(t+1) โ† max{ฮปl(t) + kฮปl(t)ฮปฬ‡l(t), 0}
    
  2. ์„ธ์…˜ ์œจ ์—…๋ฐ์ดํŠธ:
    Rr(t+1) โ† fr^(-1)(โˆ‘l:lโˆˆr ฮปl(t), wโƒ—r(t))
    
  3. ์ข…๋‹จ ๊ฐ„ ์ถฉ์‹ค๋„ ๊ฐ€๊ฒฉ ์—…๋ฐ์ดํŠธ:
    ฮผฬ‡r(t) = [Kr - โˆ‘l:lโˆˆr log wl(t)]
    ฮผr(t+1) โ† max{ฮผr(t) + kฮผr(t)ฮผฬ‡r(t), 0}
    
  4. ๋งํฌ ์ˆ˜์ค€ Werner ๋งค๊ฐœ๋ณ€์ˆ˜ ์—…๋ฐ์ดํŠธ:
    แบ‡l(t) = -dlฮปl(t) + โˆ‘r:lโˆˆr fl(Rr(t), wโƒ—r(t)) + โˆ‘r:lโˆˆr ฮผr(t)/wl(t)
    wl(t+1) โ† min{max{wl(t) + kwl(t)แบ‡l(t), 0}, 1}
    

์ด์ธต ์—…๋ฐ์ดํŠธ ๋ฐฉ์•ˆ

  • ๋‚ด์ธต: ๋งํฌ ๊ฐ€๊ฒฉ ๋ฐ ์„ธ์…˜ ์œจ์˜ ๋น ๋ฅธ ์—…๋ฐ์ดํŠธ
  • ์™ธ์ธต: ๋งค Touter ๋ฐ˜๋ณต๋งˆ๋‹ค ์ถฉ์‹ค๋„ ๊ฐ€๊ฒฉ ๋ฐ Werner ๋งค๊ฐœ๋ณ€์ˆ˜ ์—…๋ฐ์ดํŠธ, ์ œ์–ด๊ธฐ ๊ฐ„ ํ†ต์‹  ๊ฐ์†Œ

์ˆœ์ฐจ ์–‘์ž ๋„คํŠธ์›Œํฌ ๊ตฌํ˜„

q-datagram ํ—ค๋” ํ•„๋“œ

  • ฮ”Rr: ์„ธ์…˜ ์œจ ๋ณ€ํ™”
  • ฮ›sum_r: ๋ˆ„์  ๋งํฌ ๊ฐ€๊ฒฉ ํ•ฉ
  • Wprod_r: ๋ˆ„์  Werner ๋งค๊ฐœ๋ณ€์ˆ˜ ๊ณฑ
  • WrU'r: Wrโˆ‚Ur(Rr, wโƒ—r)/โˆ‚Wr ์ €์žฅ
  • ฮ”ฮผr: ์ถฉ์‹ค๋„ ๊ฐ€๊ฒฉ ๋ณ€ํ™”

์ œ์–ด๊ธฐ ์„ค๊ณ„

  1. ์„ธ์…˜ ์ œ์–ด๊ธฐ: ์†Œ์Šค ๋…ธ๋“œ์— ์œ„์น˜, Rr, ฮผr, Wr ์œ ์ง€
  2. ๋งํฌ ์ œ์–ด๊ธฐ: ๋งํฌ์— ์œ„์น˜, ฮปl, wl ๋ฐ ์„ธ์…˜ ํŠน์ • fl(Rr, wโƒ—r) ์œ ์ง€

์‹คํ—˜ ์„ค์ •

๋„คํŠธ์›Œํฌ ํ† ํด๋กœ์ง€

  1. ๋ค๋ฒจ ํ† ํด๋กœ์ง€: 8๊ฐœ ๋…ธ๋“œ, 7๊ฐœ ๋งํฌ, ๋ณ‘๋ชฉ ๋ฐ ํ˜ผ์žก ์„ฑ๋Šฅ ํ…Œ์ŠคํŠธ
  2. NSFNet ํ† ํด๋กœ์ง€: 14๊ฐœ ๋…ธ๋“œ, 21๊ฐœ ๋งํฌ, ํ™•์žฅ์„ฑ ํ…Œ์ŠคํŠธ

์‹œ์Šคํ…œ ๋งค๊ฐœ๋ณ€์ˆ˜

  • ๋ฐ˜๋ณต ์œจ: ฯ‡l = 100 kHz
  • ์ €์žฅ ํ๋น„ํŠธ: ๋…ธ๋“œ๋‹น ๋งํฌ๋‹น 50๊ฐœ
  • ์ฝ”ํžˆ์–ด๋Ÿฐ์Šค ์‹œ๊ฐ„: Tc = 1s (์œ„์ƒ ์†Œ์‹ค ๊ณ ๋ ค ์‹œ)
  • ์™ธ์ธต ์ฃผ๊ธฐ: Touter = 10

ํšจ์šฉ ํ•จ์ˆ˜

  1. ๋น„๋ฐ€ ํ‚ค ์œจ(SKR): BB84 QKD ํ”„๋กœํ† ์ฝœ ๊ธฐ๋ฐ˜
  2. ์–ฝํž˜ ์Œ์„ฑ(NEG): ์–ฝํž˜ ์Œ์„ฑ ์ธก๋„ ๊ธฐ๋ฐ˜

๋น„๊ต ๋ฐฉ๋ฒ•

QTCP ํ”„๋กœํ† ์ฝœ: ๊ณ ์ • Werner ๋งค๊ฐœ๋ณ€์ˆ˜ wl โ‰ˆ 0.967์˜ ๊ธฐ์ค€ ๋ฐฉ๋ฒ•

์‹คํ—˜ ๊ฒฐ๊ณผ

์ฃผ์š” ๊ฒฐ๊ณผ

์•ˆ์ •์  ์ˆ˜๋ ด ์„ฑ๋Šฅ

  • ์™ธ์ธต ์—…๋ฐ์ดํŠธ ์ฃผ๊ธฐ Touter โˆˆ 1, 50์€ ์ˆ˜๋ ด ๋ณด์žฅ
  • Touter โ‰ฅ 250์€ ๋ถˆ์•ˆ์ •์„ฑ ์ดˆ๋ž˜ ๊ฐ€๋Šฅ

์ •์ƒ ์ƒํƒœ ์„ฑ๋Šฅ ๋น„๊ต

  1. ์œ„์ƒ ์†Œ์‹ค ์—†๋Š” ๊ฒฝ์šฐ:
    • QPrimal-Dual ๋ฐ QPrimal-Dual-approx๋Š” ์ด๋ก ์  ์ƒํ•œ๊ณผ 5% ๋ฏธ๋งŒ์˜ ์ฐจ์ด
    • QTCP ๊ธฐ์ค€ ๋ฐฉ๋ฒ•๋ณด๋‹ค ํ˜„์ €ํžˆ ์šฐ์ˆ˜
  2. ์œ„์ƒ ์†Œ์‹ค์ด ์žˆ๋Š” ๊ฒฝ์šฐ:
    • QPrimal-Dual-DA ๋ฐ QPrimal-Dual-PI๋Š” ์„ฑ๋Šฅ์„ ํšจ๊ณผ์ ์œผ๋กœ ๋ณต๊ตฌ
    • QPrimal-Dual-DA-approx๋Š” ํ†ต์‹  ์˜ค๋ฒ„ํ—ค๋“œ ๊ฐ์†Œ ์‹œ ์œ ์‚ฌ ์„ฑ๋Šฅ ์œ ์ง€

๋™์  ์ ์‘์„ฑ

  1. ์žฅ์•  ๋ณต๊ตฌ: ๋งํฌ ์žฅ์•  ํ›„ ์ƒˆ๋กœ์šด ์ตœ์ ๊ฐ’์œผ๋กœ ๋น ๋ฅด๊ฒŒ ์ ์‘
  2. ๋™์  ์›Œํฌ๋กœ๋“œ: ์„ธ์…˜ ์ „ํ™˜ ํšจ์šฉ ํ•จ์ˆ˜ ์‹œ Werner ๋งค๊ฐœ๋ณ€์ˆ˜ ๋น ๋ฅด๊ฒŒ ์กฐ์ •

ํ™•์žฅ์„ฑ

NSFNet ํ† ํด๋กœ์ง€์—์„œ ์„ธ์…˜ ์ˆ˜ ์ฆ๊ฐ€์— ๋”ฐ๋ผ QPrimal-Dual ๋ณ€ํ˜•์€ ํ•ญ์ƒ QTCP๋ฅผ ๋Šฅ๊ฐ€

์ด๋ก ์  ๋ถ„์„

์•ˆ์ •์„ฑ ์ •๋ฆฌ

์ •๋ฆฌ 3.1 (์˜ค๋ชฉ ํšจ์šฉ ํ•จ์ˆ˜)

Ur(Rr, wโƒ—r)์ด (Rr, wโƒ—r)์—์„œ ์˜ค๋ชฉ์ด๊ณ  ๋ถ„๋ฆฌ ๊ฐ€๋Šฅํ•˜๋ฉฐ ๋‹ค๋ฅธ ๊ฐ€์ • ์กฐ๊ฑด์„ ๋งŒ์กฑํ•œ๋‹ค๋ฉด, ํ‰ํ˜•์  (Rโƒ—*, wโƒ—*, ฮปโƒ—*, ฮผโƒ—*)์€ ์ „์—ญ ์ ๊ทผ ์•ˆ์ •์ด๋‹ค.

์ •๋ฆฌ 3.2 (๋น„์˜ค๋ชฉ ํšจ์šฉ ํ•จ์ˆ˜)

Ur(Rr, wโƒ—r)์ด ๋ถ„๋ฆฌ ๊ฐ€๋Šฅํ•˜์ง€๋งŒ ๋ฐ˜๋“œ์‹œ ์˜ค๋ชฉํ•˜์ง€ ์•Š๊ณ  U''wโ„“(wโ„“) < โˆ‘r:โ„“โˆˆr ฮผr/w*2โ„“๋ฅผ ๋งŒ์กฑํ•œ๋‹ค๋ฉด, ํ‰ํ˜•์ ์€ ๊ตญ์†Œ ์ ๊ทผ ์•ˆ์ •์ด๋‹ค.

์ฆ๋ช… ๊ฐœ์š”

๋ฆฌ์•„ํ‘ธ๋…ธํ”„ ํ•จ์ˆ˜ ๋ฐ ๋ผ์‚ด ๋ถˆ๋ณ€์„ฑ ์›๋ฆฌ๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ์•ˆ์ •์„ฑ์„ ์ฆ๋ช…ํ•˜๋ฉฐ, ํ•ต์‹ฌ์€ ์ ์ ˆํ•œ ๋ฆฌ์•„ํ‘ธ๋…ธํ”„ ํ›„๋ณด ํ•จ์ˆ˜๋ฅผ ๊ตฌ์„ฑํ•˜๊ณ  ๊ทธ ๋„ํ•จ์ˆ˜๊ฐ€ ๋น„์–‘์ˆ˜์ž„์„ ์ฆ๋ช…ํ•˜๋Š” ๊ฒƒ์ด๋‹ค.

ํ™•์žฅ ๋ฐฉ์•ˆ

QPrimal-Dual-approx

์ง€์ˆ˜ ํ‰๊ท ์„ ํ†ตํ•ด ๋งํฌ ์„ธ์…˜ ์œจ ํ•ฉ์„ ์ถ”์ •ํ•˜์—ฌ ฮ”Rr ํ•„๋“œ๋ฅผ ์ œ๊ฑฐํ•˜๊ณ  ํ†ต์‹  ์˜ค๋ฒ„ํ—ค๋“œ ๊ฐ์†Œ:

Tint โ† ฮฑTint + (1-ฮฑ)(t'' - t')
Rsum_l โ† 1/Tint

QPrimal-Dual-DA (์œ„์ƒ ์†Œ์‹ค ์ธ์‹)

๋Œ€๊ธฐ์—ด ์ง€์—ฐ์„ ๊ณ ๋ คํ•˜๋„๋ก ๋งํฌ ์šฉ๋Ÿ‰ ์ œ์•ฝ ์ˆ˜์ •:

โˆ‘r:lโˆˆr Rr โ‰ค dl(1-wl) - G/Tc

์—ฌ๊ธฐ์„œ G > 1์€ ์กฐ์ • ๊ฐ€๋Šฅํ•œ ๋งค๊ฐœ๋ณ€์ˆ˜๋กœ, ๋Œ€๊ธฐ ์‹œ๊ฐ„ Tl_W โ‰ค Tc/G๋ฅผ ๋ณด์žฅํ•œ๋‹ค.

QPrimal-Dual-PI

2๋‹จ๊ณ„ ๋ฐฉ๋ฒ•: ๋จผ์ € QPrimal-Dual์„ ์‚ฌ์šฉํ•˜์—ฌ ์ˆ˜๋ ดํ•œ ํ›„, wl์„ ๊ณ ์ •ํ•˜๊ณ  QTCP ๋ฐ PI ์ œ์–ด๊ธฐ๋กœ ์ „ํ™˜.

๊ด€๋ จ ์—ฐ๊ตฌ

์–‘์ž ๋„คํŠธ์›Œํฌ ์ž์› ํ• ๋‹น

  • ๋Œ€๋ถ€๋ถ„์˜ ๊ธฐ์กด ์ „๋žต์€ ์ค‘์•™ ์ง‘์ค‘์‹ ๋ชจ๋ธ ์ฑ„ํƒ
  • ๋ถ„์‚ฐํ™” ๋ฐฉ๋ฒ•์€ ์ œํ•œ์ ์ด๋ฉฐ ์ฃผ๋กœ TCP ์ ์‘ ๋ฐฉ์•ˆ
  • ๊ธฐ์กด ๋ฐฉ๋ฒ•์€ ์ถฉ์‹ค๋„๋ฅผ ๊ณ ๋ คํ•˜์ง€ ์•Š๊ฑฐ๋‚˜ ์ด๋ก ์  ๋ณด์žฅ ๋ถ€์กฑ

๊ณ ์ „ NUM ์—ฐ๊ตฌ

  • ๊ณ ์ „ ๋„คํŠธ์›Œํฌ๋Š” ๋งŽ์€ ๋ถ„์‚ฐํ™” NUM ์†”๋ฃจ์…˜ ๋ณด์œ 
  • ๊ทธ๋Ÿฌ๋‚˜ ์–‘์ž ํšจ๊ณผ(์ถฉ์‹ค๋„ ์†์‹ค, ์ €์žฅ์†Œ ์œ„์ƒ ์†Œ์‹ค ๋“ฑ)๋กœ ์ธํ•ด ์–‘์ž ๋„คํŠธ์›Œํฌ์— ์ง์ ‘ ์ ์šฉ ๋ถˆ๊ฐ€

๊ฒฐ๋ก  ๋ฐ ๋…ผ์˜

์ฃผ์š” ๊ฒฐ๋ก 

  1. ๊ณ ์ „ NUM ์ด๋ก ์„ ์–‘์ž ๋„คํŠธ์›Œํฌ๋กœ ์„ฑ๊ณต์ ์œผ๋กœ ํ™•์žฅ
  2. ์ด๋ก ์  ์•ˆ์ •์„ฑ ๋ณด์žฅ์ด ์žˆ๋Š” ๋ถ„์‚ฐํ™” ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ œ๊ณต
  3. ์‹ค์ œ ๊ตฌํ˜„ ๋ฐฉ์•ˆ์ด ํ˜„์‹ค์  ์กฐ๊ฑด์—์„œ ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ ๋ฐœํœ˜
  4. ํ™•์žฅ ๋ฐฉ์•ˆ์ด ์–‘์ž ๊ณ ์œ  ๊ณผ์ œ๋ฅผ ํšจ๊ณผ์ ์œผ๋กœ ์ฒ˜๋ฆฌ

ํ•œ๊ณ„

  1. ์•ˆ์ •์„ฑ ๋ถ„์„์€ ์—ฐ์† ์‹œ๊ฐ„ ๋™์—ญํ•™ ๊ธฐ๋ฐ˜, ์‹ค์ œ ์‹œ์Šคํ…œ์€ ์ด์‚ฐ์ 
  2. ๊ณ ์ „ ์ •๋ณด ๊ตํ™˜ ์†์‹ค ์—†์Œ ๊ฐ€์ •
  3. ์ฃผ๋กœ ์ˆœ์ฐจ ์–‘์ž ๋„คํŠธ์›Œํฌ ์•„ํ‚คํ…์ฒ˜ ๋Œ€์ƒ
  4. ๋ถ„๋ฆฌ ๊ฐ€๋Šฅ ํšจ์šฉ ํ•จ์ˆ˜ ๊ฐ€์ • ํ•„์š”

ํ–ฅํ›„ ๋ฐฉํ–ฅ

  1. ํ”ผ๋“œ๋ฐฑ ์ง€์—ฐ์„ ๊ณ ๋ คํ•œ ์•ˆ์ •์„ฑ ๋ถ„์„
  2. ๋‹ค๋ฅธ ์–ฝํž˜ ๊ตํ™˜ ์•„ํ‚คํ…์ฒ˜๋กœ ํ™•์žฅ
  3. ๊ณ ์ „ ํ†ต์‹  ์†์‹ค ์ฒ˜๋ฆฌ
  4. ๋ถ„๋ฆฌ ๊ฐ€๋Šฅ์„ฑ ๊ฐ€์ • ์™„ํ™”

์‹ฌ์ธต ํ‰๊ฐ€

์žฅ์ 

  1. ๊ฒฌ๊ณ ํ•œ ์ด๋ก ์  ๊ธฐ์—ฌ: ์—„๊ฒฉํ•œ ์•ˆ์ •์„ฑ ์ฆ๋ช… ์ œ๊ณต, ์–‘์ž ๋„คํŠธ์›Œํฌ ๋ถ„์‚ฐํ™” ์ œ์–ด ์ด๋ก ์˜ ๊ณต๋ฐฑ ํ•ด์†Œ
  2. ๋†’์€ ์‹ค์šฉ์„ฑ: ์ˆœ์ฐจ ์–‘์ž ๋„คํŠธ์›Œํฌ์—์„œ ์™„์ „ํ•œ ๊ตฌํ˜„ ๋ฐฉ์•ˆ ์ œ๊ณต
  3. ์šฐ์ˆ˜ํ•œ ์ ์‘์„ฑ: ์œ„์ƒ ์†Œ์‹ค ๋ฐ ํ†ต์‹  ์˜ค๋ฒ„ํ—ค๋“œ ๋“ฑ ์‹ค์ œ ๊ณผ์ œ๋ฅผ ์ฒ˜๋ฆฌํ•˜๋Š” ๋‹ค์–‘ํ•œ ํ™•์žฅ ๋ฐฉ์•ˆ
  4. ์ถฉ๋ถ„ํ•œ ์‹คํ—˜: ๋‹ค์–‘ํ•œ ์‹œ๋‚˜๋ฆฌ์˜ค์—์„œ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์„ฑ๋Šฅ ๊ฒ€์ฆ

๋ถ€์กฑํ•œ ์ 

  1. ๊ฐ€์ •์˜ ์ œํ•œ: ๋ถ„๋ฆฌ ๊ฐ€๋Šฅ ํšจ์šฉ ํ•จ์ˆ˜ ๋ฐ ๊ณ ์ „ ์†์‹ค ์—†์Œ ๊ฐ€์ •์ด ์ ์šฉ์„ฑ ์ œํ•œ ๊ฐ€๋Šฅ
  2. ์•„ํ‚คํ…์ฒ˜ ํ•œ๊ณ„: ์ฃผ๋กœ ์ˆœ์ฐจ ์–‘์ž ๋„คํŠธ์›Œํฌ ๋Œ€์ƒ, ๋‹ค๋ฅธ ์•„ํ‚คํ…์ฒ˜ ์ ์šฉ์„ฑ ๋ฏธ๊ฒ€์ฆ
  3. ๋งค๊ฐœ๋ณ€์ˆ˜ ๋ฏผ๊ฐ์„ฑ: ์Šคํ… ํฌ๊ธฐ ๋งค๊ฐœ๋ณ€์ˆ˜ ์„ ํƒ์ด ์ˆ˜๋ ด ์„ฑ๋Šฅ์— ์ค‘์š”ํ•œ ์˜ํ–ฅ, ์ฒด๊ณ„์  ์ง€์นจ ๋ถ€์กฑ
  4. ๋ณต์žก์„ฑ ๋ถ„์„ ๋ถ€์žฌ: ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๋ณต์žก์„ฑ ๋ฐ ํ†ต์‹  ๋ณต์žก์„ฑ์˜ ์ƒ์„ธ ๋ถ„์„ ๋ถ€์กฑ

์˜ํ–ฅ๋ ฅ

  1. ์ด๋ก ์  ๊ฐ€์น˜: ์–‘์ž ๋„คํŠธ์›Œํฌ ์ œ์–ด ์ด๋ก ์˜ ๊ธฐ์ดˆ ๋งˆ๋ จ, ๊ณ ์ „ TCP๊ฐ€ ์ธํ„ฐ๋„ท์— ๋ฏธ์นœ ์˜๋ฏธ์™€ ์œ ์‚ฌ
  2. ์‹ค์šฉ์  ๊ฐ€์น˜: ํ–ฅํ›„ ์–‘์ž ์ธํ„ฐ๋„ท์„ ์œ„ํ•œ ์‹คํ–‰ ๊ฐ€๋Šฅํ•œ ๋ถ„์‚ฐํ™” ์ œ์–ด ๋ฐฉ์•ˆ ์ œ๊ณต
  3. ์˜๊ฐ ์ œ๊ณต: ์ž‘์—… ๋ฐฉ๋ฒ•๋ก ์„ ๋‹ค๋ฅธ ์–‘์ž ๋„คํŠธ์›Œํฌ ๋ฌธ์ œ๋กœ ํ™•์žฅ ๊ฐ€๋Šฅ

์ ์šฉ ์‹œ๋‚˜๋ฆฌ์˜ค

  1. ๋Œ€๊ทœ๋ชจ ์–‘์ž ๋„คํŠธ์›Œํฌ์˜ ๋ถ„์‚ฐํ™” ์ž์› ๊ด€๋ฆฌ
  2. ๋‹ค์ค‘ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ์–‘์ž ๋„คํŠธ์›Œํฌ์˜ ๊ณต์ •์„ฑ ๋ณด์žฅ
  3. ๋™์  ์–‘์ž ๋„คํŠธ์›Œํฌ ํ™˜๊ฒฝ์—์„œ์˜ ์ž์ ์‘ ์ œ์–ด
  4. ์–‘์ž ์ธํ„ฐ๋„ท ๊ธฐ๋ฐ˜ ์‹œ์„ค์˜ ํ”„๋กœํ† ์ฝœ ์„ค๊ณ„

์ฐธ๊ณ  ๋ฌธํ—Œ

์ฃผ์š” ์ฐธ๊ณ  ๋ฌธํ—Œ:

  1. Kelly ๋“ฑ์˜ ๊ณ ์ „ NUM ์ด๋ก  6,7
  2. Vardoyan ๋“ฑ์˜ QNUM ํ”„๋ ˆ์ž„์›Œํฌ 5
  3. ์–‘์ž ๋„คํŠธ์›Œํฌ TCP ์ ์‘ ์—ฐ๊ตฌ 32,49
  4. ์–‘์ž ์–ฝํž˜ ๋ถ„๋ฐฐ ๋ฐ ๊ตํ™˜ ๊ด€๋ จ ์—ฐ๊ตฌ 3,15,16

์ด ์—ฐ๊ตฌ๋Š” ์–‘์ž ์ธํ„ฐ๋„ท์˜ ๋ถ„์‚ฐํ™” ์ œ์–ด๋ฅผ ์œ„ํ•œ ์ค‘์š”ํ•œ ์ด๋ก ์  ๊ธฐ์ดˆ์™€ ์‹ค์šฉ์  ๋ฐฉ์•ˆ์„ ์ œ๊ณตํ•˜๋ฉฐ, ์–‘์ž ๋„คํŠธ์›Œํฌ ํ”„๋กœํ† ์ฝœ ์Šคํƒ์˜ ๊ธฐ์ดˆ ๊ตฌ์„ฑ ์š”์†Œ๊ฐ€ ๋  ๊ฒƒ์œผ๋กœ ์˜ˆ์ƒ๋œ๋‹ค.