CN103297356B - The method of a kind of optimization data center switching network cost - Google Patents

Abstract

The invention discloses the method for a kind of optimization data center switching network cost, it comprises the steps: that input stage interchanger accumulates the packet received, and forms traffic matrix B; ADTPT algorithm is utilized to set up tradeoff between packet delay and core switch quantity; Calculate the minimum core switch number under given time delay; The minimum cost of calculated data center switching network.The present invention has the following advantages: time delay and core switch quantity tradeoff design, reduce the quantity of core switch, flexible design by suitably increasing time delay, enhances the extensibility of exchanging transforming network of data centre design; Core switch reshuffled total cost that expense brings switching network into, perfect switching network Cost Design; Traffic matrix decomposition algorithm is expanded to the scheduling of multiple interchanger from the scheduling of single interchanger, achieve the lot size scheduling in the exchanging transforming network of data centre met under delay constraints, the perfect lot size scheduling process of multistage exchange.

Description

The method of a kind of optimization data center switching network cost

Technical field

The present invention relates to the method for a kind of optimization data center switching network cost.

Background technology

Along with the fast development of cloud service, data center receives as service supporting platform and pays close attention to widely.In order to solve the energy consumption that data center's inner exchanging network design faces large, inner connect complicated, cost is high, switching performance is difficult to the problems such as guarantees, employing light exchanges the Main way that the novel framework of data center be connected with light has become technical development.The feature of these schemes adopts OpenFlow interchanger as ToR(or Aggregation) interchanger, utilize crossbar optical exchanger to realize core switch simultaneously, and adopt high speed fibre to connect in heart switching network in the data, thus simplify the structure of switching network, enhance the extensibility of system.

But this multistage Clos framework exchanging formation, in order to realize nonblocking switching, the number needs of intergrade interchanger (i.e. core switch) will increase along with the increase of edge switcher port number in a particular manner.Such as, rearrangeable clog-free in order to realize, the number of demand fulfillment intergrade interchanger must not be less than the port number of input stage interchanger.This has made very strong restriction to the quantity of the core switch needed for data center, thus constrains the cost of exchanging transforming network of data centre.

For above-mentioned simulate spots system, usually adopt the packet switch of lot size scheduling (Batchscheduling) (Packetswitching) pattern to the switching performance of the resource utilization and system that improve switching network.In lot size scheduling process, packet is first through accumulation after a while in input-buffer, and then exchanges by interchanger is unified under certain exchange connects configuration.For the simulate spots system of single interchanger, BinWu gives the dispatching method decomposed based on traffic matrix.According to the ADAPT algorithm that it provides, for each row and and row and be no more than the traffic matrix of r × r of nA, ADAPT algorithm can resolve into N it _s(r ²-2r+2>N _s>r) individual transposed matrix { P _k(N _s>=k>=1), the one configuration of each transposed matrix respective exchanger, and the timeslot number maintained is:

But current ADAPT algorithm is only applied to the scheduling of single interchanger, in addition, the expense of reshuffling that above-mentioned employing crossbar optical exchanger is system as the major issue that core switch faces (mainly comprises core switch and is switching the time overhead needed for interconnection configuration, and between each interchanger, carry out the synchronous time consumed), along with the continuous expansion of data center's scale, the lifting of switching network complexity, reshuffle overhead issues and will seem more and more outstanding, but it is current for the optimized design studies of data center's inner exchanging network cost, what have ignored system all the time reshuffles expense, it is not too reasonable to design.

Summary of the invention

The object of the invention is to overcome the deficiencies in the prior art, the method of a kind of optimization data center switching network cost is provided, interchanger is reshuffled expense and is included within the limit of consideration of cost optimization design by it, ADAPT algorithm is expanded to multistage switching network, carry out traffic matrix decompose and exchange scheduling, and then the tradeoff design realized between packet delay and core switch quantity, make exchanging transforming network of data centre cost optimization, also breach the constraint to core switch quantity in existing design simultaneously.

The object of the invention is to be achieved through the following technical solutions: the method for a kind of optimization data center switching network cost, it comprises the steps:

S1: input stage interchanger accumulates the packet received, forms the traffic matrix B of a r × r;

S2: utilize ADTPT algorithm to set up tradeoff formula between packet delay and core switch quantity, the step setting up tradeoff formula is as follows:

S21: adopt ADTPT algorithm that traffic matrix B is resolved into N _sindividual transposed matrix P _kweighted sum;

S22: calculate packet-switching time delay A with the tradeoff formula between core switch quantity m;

S3: calculate the minimum core switch number under given time delay, computational methods are as follows:

S31: ask for about N for core switch quantity m according to tradeoff formula _ssecond order local derviation;

S32: find unique N _sm is made to obtain minimum value;

S4: the minimum cost of calculated data center switching network, computational methods are as follows:

S41: the conversion cost of definition unit time delay is τ, then total cost is intergrade interchanger cost and latency penalty sum, that is: C=m+ τ A;

S42: according to the functional relation between m and A, obtains the function of C and A;

S43: ask for the second derivative of C about A;

S44: computation delay border A and required core switch number N _s, and then obtain exchanging transforming network of data centre minimum cost C.

Further, described packet-switching time delay A with the tradeoff formula result between core switch quantity m is:

$A=\frac{{\mathrm{\δN}}_S}{m-\frac{N_S}{N_S-r}}$ And $m=\frac{{\mathrm{\δN}}_S}{A}+\frac{N_S}{N_S-r};$

Wherein, δ be core switch reshuffle the overhead time.

Further, the minimum value that described core switch quantity m obtains is:

$m=\frac{{\mathrm{\δN}}_S}{A}+\frac{N_S}{N_S-r}{|}_{N_S=r+\sqrt{\frac{\mathrm{rA}}{\mathrm{\δ}}}}={(1+p)}^2;$

Wherein, $p=\sqrt{\frac{\mathrm{\δr}}{A}}.$

Further, the functional relation of described C and A is:

$C={(1+p)}^2+\mathrm{\τA}={(1+\sqrt{\frac{\mathrm{\δr}}{A}})}^2+\mathrm{\τA}.$

The invention has the beneficial effects as follows:

(1) tradeoff design of packet delay and core switch quantity, can reduce the quantity of required core switch, flexible design by suitably increasing time delay, and enhances the extensibility of exchanging transforming network of data centre design;

(2) reshuffle by core switch total cost that expense brings exchanging transforming network of data centre into, design is more reasonable, perfect switching network Cost Design model;

(3) traffic matrix decomposition algorithm is expanded to the scheduling of multiple interchanger from the scheduling of single interchanger, achieve the lot size scheduling in the exchanging transforming network of data centre met under delay constraints, the perfect lot size scheduling process of multistage exchange.

Accompanying drawing explanation

Fig. 1 is method flow diagram of the present invention;

Fig. 2 is the structure chart of three grades of switching networks.

Embodiment

Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail, but protection scope of the present invention is not limited to the following stated.

As shown in Figure 2, three grades of Clos (n, m, r) switching network framework, the input stage interchanger of r n × m is had in this network, the output interchanger of m r × r intergrade interchanger (i.e. core switch) and r m × n, is all communicated with between any two ends mouth (output port of previous stage, the input port of a rear collection) of adjacent two-stage interchanger.

As shown in Figure 1, the method for a kind of optimization data center switching network cost, it comprises the steps:

S1: input stage interchanger accumulates the packet received, the accumulation cycle is each time slot of A, forms the traffic matrix B of a r × r.The element of any row of B or arbitrarily row and be all no more than nA.Connect owing to have employed light, the inner packet transmission rate of core switch is M times (M >=n) of individual server packet transmission rate, supposes M=n here.

S2: utilize ADTPT algorithm to set up tradeoff formula between packet delay and core switch quantity, the step setting up tradeoff formula is as follows:

S21: adopt ADTPT algorithm that traffic matrix B is resolved into N _sindividual transposed matrix P _kweighted sum;

The overhead time of reshuffling supposing intergrade interchanger (namely core switch) is δ, do not collapse under pressure to make the buffer of ToR interchanger store data bag, core stage interchanger must complete the exchange of all packets in A time slot, so can obtain following restriction relation:

Wherein, m is the number of core switch, for interchanger is configured to P _ktime need maintain timeslot number, in above-mentioned restriction relation formula for the scheduling total length in A time slot, this total length is on average born by m core switch.

S22: calculate packet-switching time delay A with the tradeoff formula between core switch quantity m;

Further abbreviation is carried out to above-mentioned restriction relation formula, packet-switching time delay A can be obtained with the tradeoff formula result between core switch quantity m:

$A=\frac{{\mathrm{\δN}}_S}{m-\frac{N_S}{N_S-r}}$ With $m=\frac{{\mathrm{\δN}}_S}{A}+\frac{N_S}{N_S-r};$

Wherein, δ be core switch reshuffle the overhead time.When A value is larger, packet-switching time delay is also larger, and required core switch is also fewer.

S3: calculate the minimum core switch number under given time delay, computational methods are as follows:

S31: ask for about N for core switch quantity m according to tradeoff formula _ssecond order local derviation, due to given δ, r and A, then obtain:

$\frac{d^{2}m}{d{N_S}^2}=\frac{2r}{{(N_S-r)}^3}>0;$

Ensure that N must be had to realize switching performance _s>r.Therefore, m is N _sconcave function, and unique N can be found _sm is made to obtain minimum value.Order formula is below set up:

$\frac{\mathrm{dm}}{{\mathrm{dN}}_S}=\frac{\mathrm{\δ}}{A}-\frac{r}{{(N_S-r)}^2}=0;$

Can obtain:

$N_S=r+\sqrt{\frac{\mathrm{rA}}{\mathrm{\δ}}};$

Wherein N _sbe necessary for integer, therefore in practice should value

S32: find unique N _sm is made to obtain minimum value;

$m=\frac{{\mathrm{\δN}}_S}{A}+\frac{N_S}{N_S-r}{|}_{N_S=r+\sqrt{\frac{\mathrm{rA}}{\mathrm{\δ}}}}={(1+p)}^2;$

Wherein, $p=\sqrt{\frac{\mathrm{\δr}}{A}}.$

S4: the minimum cost of calculated data center switching network, computational methods are as follows:

S41: the conversion cost of definition unit time delay is τ, then total cost is intergrade interchanger cost and latency penalty sum, that is: C=m+ τ A;

S42: according to the functional relation between m and A, obtains the function of C and A, namely

$C={(1+p)}^2+\mathrm{\τA}={(1+\sqrt{\frac{\mathrm{\δr}}{A}})}^2+\mathrm{\τA};$

S43: ask for the second derivative of C about A:

$\frac{d^{2}C}{d{A}^2}=\frac{1}{A^3}(\frac{3\sqrt{\mathrm{\δrA}}}{2}+2\mathrm{\δr})>0;$

Therefore, C is the concave function of A, and C has unique minimum value:

$\frac{\mathrm{dC}}{\mathrm{dA}}=-\frac{\mathrm{\δr}}{A^2}-\sqrt{\frac{\mathrm{\δr}}{A^3}}+\mathrm{\τ}=-\frac{1}{\mathrm{\δr}}(p^4+p^3-\mathrm{\τ\δr})=0.$

S44: according to above-mentioned function, can find p value by the method for numerical solution, calculates time delay border A and required core switch number N _s, and then obtain exchanging transforming network of data centre minimum cost C.

Claims (4)

1. a method for optimization data center switching network cost, is characterized in that: it comprises the steps:

S1: input stage interchanger accumulates the packet received, forms the traffic matrix B of a r × r;

S2: utilize ADAPT algorithm to set up tradeoff formula between packet delay and core switch quantity, the step setting up tradeoff formula is as follows:

S21: adopt ADAPT algorithm that traffic matrix B is resolved into N _sindividual transposed matrix P _kweighted sum;

S22: calculate packet-switching time delay A with the tradeoff formula between core switch quantity m;

S3: calculate the minimum core switch number under given time delay, computational methods are as follows:

S31: ask for about N for core switch quantity m according to tradeoff formula _ssecond order local derviation;

S32: find unique N _sm is made to obtain minimum value;

S4: the minimum cost of calculated data center switching network, computational methods are as follows:

S41: the conversion cost of definition unit time delay is τ, then total cost is intergrade interchanger cost and latency penalty sum, that is:

C=m+τA；

S42: according to the functional relation between m and A, obtains the function of C and A;

S43: ask for the second derivative of C about A;

S44: computation delay border A and required core switch number N _s, and then obtain exchanging transforming network of data centre minimum cost C.

2. the method for a kind of optimization data center according to claim 1 switching network cost, is characterized in that: described packet-switching time delay A with the tradeoff formula result between core switch quantity m is:

$A=\frac{{\mathrm{\δN}}_S}{m-\frac{N_S}{N_S-r}}$ And $m=\frac{{\mathrm{\δN}}_S}{A}+\frac{N_S}{N_S-r};$

Wherein, δ be core switch reshuffle the overhead time.

3. the method for a kind of optimization data center according to claim 1 switching network cost, is characterized in that: the minimum value that described core switch quantity m obtains is:

$m=\frac{{\mathrm{\δN}}_S}{A}+\frac{N_S}{N_S-r}{|}_{N_S=r+\sqrt{\frac{rA}{\mathrm{\δ}}}}={(1+p)}^2;$

Wherein, $p=\sqrt{\frac{\mathrm{\δ}r}{A}}.$

4. the method for a kind of optimization data center according to claim 1 switching network cost, is characterized in that: the functional relation of described C and A is:

$C={(1+p)}^2+\mathrm{\τ}A={\left(1+\sqrt{\frac{\mathrm{\δ}r}{A}}\right)}^2+\mathrm{\τ}A.$