Virtual Resource Management:EU FP7 EVANS Project End-to-end Virtual Resource Management across Heterogeneous Networks and Services (EVANS) “vertical'”resource management and control Horizontal"resource management by network infrastructure providers and control by service providers SP2 Virtual network #2 Virtual network #l SP 1 Mobile operator Backbone provider Backbone provider Mobile operator (P) (Optical/GMPLS)
Virtual Resource Management: EU FP7 EVANS Project EVANS Project End-to-end Virtual Resource Management across Heterogeneous Networks and Services ( ) EVANS ) “vertical” resource management and control by network infrastructure providers “Horizontal” resource management and control by service providers SP 2 Virtual network #1 SP 1 Virtual network #2 Virtual network #1 SP 1 16 Mobile operator Mobile operator Backbone provider (IP) Backbone provider (Optical/GMPLS)
NV Issues in Wired Networks Mapping and Embedding:static vs.dynamic Resource allocation Scheduling Energy Efficiency:putting some nodes or links into sleep mode 17
NV Issues in Wired Networks NV Issues in Wired Networks Mapping and Embedding: static vs dynamic Mapping and Embedding: static vs. dynamic Resource allocation & Scheduling E Effi i tti d li k i t l Energy Effi c iency: putting some no des or li n ks i n to s leep mode 17
Mapping or Embedding -Offline,Static One-off mapping done offline, ■ e.g.,Zhu and Ammar."Algorithms for Assigning Substrate Network Resources to Virtual Network Components",Infocom'06 ■ Objective:to achieve load balancing in terms of stress across either: Node: max SN(t,v) RN(t)= vEVs [∑ SN(t,v)]/IVsl vEVs Optimal link stress Link: max SL(t,e) R红(t)= eEEs [∑ SL(t,e]/八Esl eEEs Optimal Node stress ■ Bad assumption:the physical network resources is unlimited. 18
Mapping or Embedding Mapping or Embedding – Offline, Static Offline, Static One-off mapping done offline, e.g., Zh d u an d Ammar. “Al i h f Algorit hms for Assigning Substrate Network Resources to Virtual Network Components”, Infocom’06 Obj g ective: to achieve load balancin g in terms of stress across either: Node: Link: Optimal link stress Bad assumption: the ph ysical network resources is unlimited. Optimal Node stress 18
Mapping or Embedding:Online,Dynamic Yu et al."Rethinking Virtual Network Embedding:Substrate Support for Path Splitting and Migration",Sigcomm 08 Assume that the substrate network supports path splitting. Path optimization(and thus path migration)is carried out at each time slot t. New arnving New amiving VN Reg 2 Without VN Rey 2 Without Ex对isting path splitting Existing 30 e migration Reject VN Req 30 Reject NRe网I With Path splitting Accept migration Accept 15 15 15 15 10 20 2049 10 …40… 26 20 20 30 20 2 0 Substrate network at time t-I Substrate network at time t Substrate network at time t-1 Substrate network at time t 19
Mapping or Embedding: Online, Dynamic Mapping or Embedding: Online, Dynamic Yu et al. “Rethinking Virtual Network Embedding: Substrate Support for Path Splitting and Migration for Path Splitting and Migration ” , Sigcomm 08 Sigcomm 08 Assume that the substrate network supports path splitting. Path o p ( p g) timization (and thus path mi gration ) is carried out at each time slot t. 19
Resource Allocation Scheduling Static:simple but not efficient;dynamic:more efficient but less stable ■ Each substrate link periodically reassigns bandwidth shares between its virtual links. ■ Method:to max/min a utility function subject to certain constraints such as link bandwidth. Problem:not too much flavour of network virtualization-not like in mapping maximize U(x)=∑4x) subject to！ xs routing matrix or happiness of user i,a data rate function of flow rate link bandwidths 20
Resource Allocation & Scheduling Resource Allocation & Scheduling Static: simple but not efficient; dynamic: more efficient but less stable E h b t t li k i di ll i b d idth h b t it Eac h su b s tra te li n k per i odically reass igns ban d width s hares b e tween its virtual links. Method: to max/min a utilit y j function subject to certain constraints such as link bandwidth. Problem: not too much flavour of network virtualization – not like in mapping maximize U(x) = ∑i u i (x i ) subject to Rx ≤ C mapping subject to Rx ≤ C happiness of user i, a function of flow rate routing matrix or data rate link bandwidths 20