36 Z1 Vol.36 No.Z1 2015 11 Journal on Communications November 2015 doi:10.11959/j.issn.1000-436x.2015275 1 1,2 1 (1. 400065 2. 401520) MDS 3 P302 A Collaboration coding to multi-node repair program under the twin-mds codes framewor in cloud storage systems XIE Xian-zhong 1, HUANG Qian 1, 2, WANG Liu-su 1 (1. Institute of Broadband Access echnologies, Chongqing University of Posts and elecommunications, Chongqing 400065, China; 2. Department of Computer Science, College Mobile elecommunications, Chongqing University of Posts and elecommunications, Chongqing 401520, China) Abstract: A multi-node exact repair code scheme, which can repair multiple system nodes or redundant nodes simultaneously, was shown and proved to against the disadvantages of the existing multi-node repair model in cloud storage. he multi-node exact repair code was combined with a twin-mds codes framewor with health cooperative nodes. In this way, repair bandwidth, the number of repair lins and the amount of data to be treated in an intermediate node were reduced, while multi-node were repaired. Finally, numerical simulation results show that this scheme has greater improvements. In particular, it reduces the load in an intermediate node. And the advantages was more obvious with the more storage nodes in cloud storage. Key words: cloud storage; multi-node exact repair code; collaboration repair; twin-mds codes model 1 / 2007 Dimais [1] / regenerating codescloud storage networ codes [2,3] Suh [4] - MDS, maximum distance separable [4] MDS E-MDS, exact MDS 2015-10-29 61271259 0872037CSC2011jjA40006 CSC2010BB2415KJ120501 KJ110530 Foundation Items: he National Natural Science Foundation of China (61271259, 0872037); he Natural Science Foundation of Chongqing (CSC2011jjA40006, CSC2010BB2415); he Scientific and echnological Research Program of Chongqing Municipal Education Commission (KJ120501, KJ110530) 2015275-1
2 36 [5~12] [5] [6,7] [8] =2 [9,10] MBRC, minimum-bandwidth regenerating codes MSRC, minimum-storage regenerating codes MBRC MSRC Shum [11] MCR, multi-node cooperative regenerating MCR [12] [13] MDS (twin-mds codes) MDS [12] [4] E-MDS MER, multi-node exact repair 2 MDS 2.1 r d d m m 1 1 4 A 1 A 2 B 1 B 2 2 1 2 MDS 2 3 4 1 3 4 m 2 A 1 A 2 m m 2 1 3 m d 1 β r α γ=(d 1)β+rα 1 6 3 2 2015275-2
Z1 3 m MDS 2.2 MDS MDS n 1 n 1 2 n 2 2 i n i A i G i i l(1 l n i ) l A i g (i,l) (1 l n i ) i l G i g (i,l) g (i,l) 3 3 MDS 2 2 MDS 1 C 1 2 C 2 2 C 1 C 2 MDS MDS F q M i=0 1 C i [n i, ] F q MDS G i G i l g (i,l) (1 l n i ) 2 2 A 1 A 2 =A t 1 t C i A i MDS 1 1 4 2 2 3 4 g (1,1) 2 1 2 3 g (2,4) 2 4 4 MDS 2.3 MDS MER MDS MDS 2015275-3
4 36 [12,13] MDS 5 2 C 1 C 2 1 2 m 1 m 2 2 r 2 1 d 1 m 1 d 1 1 1 r 1 m 2 d 2 1 α ββ=α MDS MER 3 MDS MER Suh [4] (n,, d)=(2,, 2 1) E-MDS E-MDS M=β (d +1) β=1 α = M = E-MDS 5 MDS r C 1 C 2 2 r 1 r 2 d 1 d 2 2 2 m 1 m 2 m 1 m 2 d i 1 β i r i α γ=γ 1 +γ 2 γ i =(d i 1)β i +r i α m 2 1 1 1 d=d 1 1+d 2 1=d 1 +d 2 2 d=d 1 +d 2 1 E-MDS [4] X=[x ij ], Y=[y ij ], Ψ=[ψ ij ], Φ =Ψ 1 =[φ ij ] X = YΨ, Y = XΦ (1) Ψ Φ X x 1,, x Y y 1,, y K={1,,} i K(1) x = ψ y + ψ y + + ψ y i 1i 1 2i 2 i y = ϕ x + ϕ x + + ϕ x i 1i 1 2i 2 i (2) X =(X ) 1 Ỹ =(Y ) 1 X Y X Ỹ x 1,, x ỹ 1,, ỹ Θ=[θ 1 θ ] D=[δ 1 δ ]θ i i δ i +i [4] E-MDS 1 +1 2 +1 2 1 i +i i K 2015275-4
Z1 5 (3) D = ωyθ X + σθψ ω σ GF(q) ΘΨ j λ j = = 1 ψ ε ε j ε θ 3 δ = j ω y x θ + σ i j i λj i= 1, j K (4) Θ = ω XD Y + σ DΦ (5) ω σ GF(q) DΦ j λ j = = 1 ϕ ε ε j ε δ 5 θ =, j ω x y δ + σ j K i j i λ (6) j i= 1 35 ω σ ω σ ωω + σσ = 1, σω + ωσ = 0 (7) E-MDS F(Θ)= ωyθ X +σθψ G(D)= ω XD Y + σ DΦ G(F(Θ))=Θ F(G(D))= D E-MDS MDS E-MDS MER, multi-node exact repair E-MDS 1 ( ) [14] m n 1 Ψ= a i b j GF(q) a b i j a i b j 0 a i a j b i b j 1 i m 1 j n C Cauchy 1 E-MDS Y Ψ ω σ ω σ 1) Y GF(q) 2) Ψ GF(q) 3) ω σ ω σ 7 4) i j K ψ ij φ ij 1 r MER E-MDS i K Θ i D +i MER r MER d=n r i y i i +i x i (+i) r 2 r r Θ D 2 r +1 +r (+i) x i θ 1,,x i θ x i θ j =ωx j λ i +σx i λ j i {1,,r } j {r+1, r+2,, }4 (+i) δ = i ω y x θ + σ ε i ε λ (8) i ε = 1 8 (x i θ 1,, x i θ ) r+1 j (+i) x i δ j =ωx j λ i +σx i λ j x λ = i i ψ x θ εi i ε ε = 1 (9) 1 σ x λ = x δ ψ j i i j ε j i ε x θ ω ω ε = 1 1 j r (+i) r 1 x λ i j j i x 1,,x λ i r=2 2 u +v u v K u +v u y θ u i y δ = ω x θ + σ y λ u j j u u j (10) 2015275-5
6 36 (+v) x θ v i x δ = ω x λ + σ x λ v j j v v j i K \{u} j K \{v} (11) ε ψ iε ϕ εj δ ij i=j ε ψ iε ϕ εj =1 i j ε ψ iε ϕ εj =0 (+v) 11 ϕ j = 1, j v ju v j x δ +(ω+σ) i= 1, y u λ v x v θ u ψ ϕ x θ i u iv vu v i ( ( ) ) ω y λ + σ σ + ω ψ ϕ x θ (12) u v uv vu v u (+v) 12 u u 10 12 ωψ iv y θ u i 10 ( ωψ ( σ ( σ ω) ψ ϕ ) ) y + + x θ ( ω x + σψ y j uj u ) θu uv u uv vu v u ( σ ( σ + ω) ψ ϕ uv vu ) x + ωψ y v uv u ω x + σψ y 1 u1 u Α = ω x + σψ y ω x ω x + σψ y u u v 1 u, v 1 u + σψ y v+ 1 u, v+ 1 u (13) (14) θ u u MER A σ (σ+ω)ψ uv ϕ vu 2 σ (σ+ω)ψ uv ϕ vu =0 14 ωψ y uv u uv u ωψ ϕ x ε ε ε = 1 Β = = ω K \{ v} X ω X K \{ v} (15) X K\{v} =[x 1 x v 1 x v+1 x ] 15 B σ (σ+ω)ψ uv ϕ vu 0 2 y u = = 1 C = C C ϕ ε ε u ε 1 2 x 14 σ σψ ϕ ωψ Φ uv vu uv \{ } x, v K v u = (16) σϕ Ψ ωι + σψ Φ vu u, K \{ v} u, K \{ v} K \{ v}, u X K \{ v} Ψ u, K\{v} =[ψ u1 ψ u,v 1 ψ u,v+1 ψ u ] Φ K\{v}, u= [ϕ 1u ϕ v 1,u ϕ v+1,u ϕ u ] I ( 1) ( 1) A 16 C C 1 C 1 C = Λ + gf 1 ( ) σ σ + ω ψ ϕ 0 uv vu = + 0 ωι ωψ uv ϕvu K \{ v}, u σ u, K \{ v Φ Ψ } (17) σ (σ+ω)ψ uv ϕ vu ω Λ Sherman-Morrison = 1+ f Λ 1 g 0C 1 ψ ϕ ε 1 iε ε j δ ij 1 ψ ϕ 1 ε = uε εu ( )( uv vu ) σ ( σ + ω) ψ ϕ σ σ + ω 1 ψ ϕ 1 1+ f Λ g = (18) 7 2 (ω 2 σ 2 )(ω 2 σ 2 )=1 1 ω 2 σ 2 σ 0 ψ uv ϕ vu 1 1+ f Λ 1 g 0 14 A MER u θ u MER E-MDS (+v) u ω x λ u +(σ (σ +ω )ϕ vu ψ uv ) y δ v +v v u 1 MER E-MDS E-MDS r 2 4 uv vu 2 2015275-6
Z1 7 MDS [12,13] [1] [15] [11] [12] =d,r =n n<2 (n,, d)(α, B) 4.1 MDS γ 1 1 (n,, d)(α, B) (4,2,2) (2,4) γ =4 γ =3 γ =3 γ =3 γ =3 (7,4,4)(21,84) γ =84 γ =51.3 γ =42 γ =42 γ =35 (14,8,8)(80,640) γ =640 γ =261.3 γ =173.3 γ =173.3 γ =160 4.2 f 2 2 (n,, d)(α, B) (4,2,2)(2,4) f=6 f=3 f=3 (7,4,4)(21,84) f=18 f=6 f=5 (14,8,8)(80,640) f=78 f=13 f=12 1 4.3 MDS 2 m 1 m 2 m 1 m 2 [12] 3 3 (n,, d)(α, B) (4,2,2)(2,4) 4 2 (7,4,4)(21,84) 21 11 (14,8,8)(80,640) 86.7 44 3 MDS MER E-MDS 1 MER MDS 5 MDS MDS E-MDS MER 2015275-7
8 36 MER MDS MDS MER MER [1] DIMAKIS A G, GODFREY P B, WAINWRIGH M J, et al. Networ coding for distributed storage systems[a]. IEEE International Conference on Computer Communications (INFOCOM)[C]. 2007. 2000-2008. [2] DIMAKIS A G, RAMCHANDRAN K, WU Y, et al. A survey on networ codes for distributed storage[j]. Proceedings of the IEEE, 2011, 99(3): 476-489. [3] RASHMI K V, SHAH N B, KUMAR P V. Optimal exact regenerating codes for distributed storage at the MSR and MBR points via a product-matrix construction[j]. IEEE ransactions on Information heory, 2011, 57(8): 5227-5239. [4] SUH C, RAMCHANDRAN K. Exact-repair MDS code construction using interference alignment[j]. IEEE ransactions on Information heory, 2011, 57(3): 1425-1442. [5] HU Y, XU Y, WANG X, et al. Cooperative recovery of distributed storage systems from multiple losses with networ coding[j]. IEEE Journal on Selected Areas in Communications, 2010, 28(2): 268-275. [6] WANG X, XU Y, HU Y, et al. MFR: multi-loss flexible recovery in distributed storage systems[a]. IEEE International Conference on Communications (ICC)[C]. 2010. 1-5. [7],,,. [J]. 2014, 42 (10): 1873-1881. XIE X Z, HUANG Q, WANG L S, et al. A multi-node exact repair method in cloud storage based on interference alignment[j], Acta Electronica Sinica, 2014, 42 (10): 1873-1881. [8] LE S N. Exact scalar minimum storage coordinated regenerating codes[a]. IEEE International Symposium on Information heory Proceedings (ISI)[C]. 2012. 1197-1201. [9] WANG A, ZHANG Z. Exact cooperative regenerating codes with minimum-repair-bandwidth for distributed storage[a]. IEEE International Conference on Computer Communications (INFOCOM)[C]. 2013. 400-404. [10] LI J, LI B. Cooperative repair with minimum-storage regenerating codes for distributed storage[a]. IEEE International Conference on Computer Communications (INFOCOM)[C]. 2014. 316-324. [11] SHUM K W, HU Y. Cooperative regenerating codes[j]. IEEE ransactions on Information heory, 2013, 59(11): 7229-7258. [12],,,. [J].. 2014, 37(1):52-56. XIE X Z, WANG L S, HUANG Q, et al. Efficient multi-node regenerating program with healthy nodes collaboration in distributed storage systems[j]. Journal of Beijing University of Posts and elecommunications, 2014, 37(1):52-56. [13] RASHMI K V, SHAH N B, KUMAR P V. Enabling node repair in any erasure code for distributed storage[a]. IEEE International Symposium on Information heory(isi)[c]. Saint Petersburg, Russia, 2011. 1235-1239. [14] LI X, ZHENG Q, QIAN H, et al. oward optimizing cauchy matrix for cauchy reed-solomon code[j]. IEEE Communications Letters, 2009, 13(8): 603-605. [15] WU Y, DIMAKIS A G. Reducing repair traffic for erasure coding-based storage via interference alignment[a]. IEEE International Symposium on Information heory Proceedings (ISI)[C]. Seoul, 2009. 2276-2280. 1966-1988- 1990-2015275-8