## Providing deterministic quality of service in slotted optical networks

Optics Express, Vol. 14, Issue 26, pp. 12679-12692 (2006)

http://dx.doi.org/10.1364/OE.14.012679

Acrobat PDF (394 KB)

### Abstract

This paper proposes an efficient framework for deterministic service guarantees in slot-based optical networks. The framework uses the combination of a control plane and a data plane to solve the complex problem of capacity allocation, slot-matching and traffic scheduling. The control plane implements admission control and capacity allocation to source-destination node pairs and the data plane handles traffic aggregation, buffering, and scheduling. We propose an efficient algorithm in the control plane for slot collision-resolution. In the data plane, we present a comprehensive aggregation and scheduling mechanism that realizes Service Curves assurance. We use the Time-Domain Wavelength Interleaved Network (TWIN) architecture for the proof of concept and conduct extensive simulations to assess the performance of the algorithm and scheduling mechanism.

© 2006 Optical Society of America

## 1. Introduction

1. S. Yao, S. Dixit, and B. Mukherjee, “Advances in Photonic Packet Switching: an overview,” IEEE Commun. Mag. **38**, 84–94, Feb. (2000). [CrossRef]

3. I. Widjaja, I. Saniee, R. Giles, and D. Mitra, “Light core and intelligent edge for a flexible, thin-layered and cost-effective optical transport network,” IEEE Commun. Mag. **41**, 30–36, (2003). [CrossRef]

3. I. Widjaja, I. Saniee, R. Giles, and D. Mitra, “Light core and intelligent edge for a flexible, thin-layered and cost-effective optical transport network,” IEEE Commun. Mag. **41**, 30–36, (2003). [CrossRef]

*λ*

_{j}is assigned to a destination node

*j*, and source nodes tune to this wavelength

*λ*

_{j}to transmit to node

*j*. Route between each source-destination node-pair is fixed in TWIN architecture. The routes from multiple sources to a destination form a multipoint-to-point tree (see dashed line and dotted line trees in Fig. 1). Signals are aggregated at the

*aggregate devices*(ADs) and routed in the network by passive components such as the

*wavelength-selective cross-connects*(WSXCs). Once preprovisioned, each WSXC performs self-routing of various optical signals on-the-fly based on their wavelengths. Compared to OPS [1

1. S. Yao, S. Dixit, and B. Mukherjee, “Advances in Photonic Packet Switching: an overview,” IEEE Commun. Mag. **38**, 84–94, Feb. (2000). [CrossRef]

*N*nodes with roughly

*N*

^{1/2}wavelengths. This enhancement has made TWIN more practical for implementation. In this paper, we use the TWIN architecture to demonstrate the operation of our proposed slot-based optical framework.

5. N. Golmie, T. Ndousse, and D. Su, “A differentiated optical service model for WDM Networks,” IEEE Commun. Mag. **38**, 68–73, Feb. (2000). [CrossRef]

6. B. Li and Y. Qin, “Traffic scheduling in a Photonic Packet Switching System with QoS Guarantee,” J. Lightwave Technol. **16**, 2281–2295 (1998). [CrossRef]

7. M. Yoo, C. Qiao, and S. Dixit, “Optical burst switching for service differentiation in the next generation optical Internet,” IEEE Commun. Mag. **39**, 98–104, Feb. (2001). [CrossRef]

8. Maode Ma and M. Hamdi, “Providing deterministic quality-of-service guarantees on WDM Optical Networks,” IEEE J. Sel. Areas Commun. **18**, 2072–2083 (2000). [CrossRef]

8. Maode Ma and M. Hamdi, “Providing deterministic quality-of-service guarantees on WDM Optical Networks,” IEEE J. Sel. Areas Commun. **18**, 2072–2083 (2000). [CrossRef]

1. S. Yao, S. Dixit, and B. Mukherjee, “Advances in Photonic Packet Switching: an overview,” IEEE Commun. Mag. **38**, 84–94, Feb. (2000). [CrossRef]

22. C. K. Siew and M. H. Er, “A new multiservice provisioning mechanism with service curves assurance for per-class scheduling delay guarantees” in press for IEE Proc. Communications. Available at URL: http://www.icis.ntu.edu.sg/our_institute/staff/cksiew/revised_COM_2005_0246-uncorrected%20final%20draft.pdf

## 2. Framework configuration overview

10. H. A. Mantar, J. S. Hwang, I. T. Okumus, and S. J. Chapin, “A scalable model for interbandwidth broker resource reservation and provisioning,” IEEE J. Sel. Areas Commun. **22**, 2019–2034 (2004). [CrossRef]

## 3. Control plane

### 3.1 Overview

11. K. Ross, N. Bambos, K. Kumaran, I. Saniee, and I. Widjaja, “Scheduling bursts in Time-Domain Wavelength Interleaved Networks,” IEEE J. Sel. Areas Commun. **21**, 1441–1451 (2003). [CrossRef]

*N*, the propagation delays from nodes 1, 2, ..,

*N*-1 are

*t*

_{1},

*t*

_{2}, ..,

*t*

_{N-1}respectively. The problem is more challenging than the zero delay case (Birkhoff-von Neumann switch) or the equal delay case (star topology), which have been discussed in Refs. [11

11. K. Ross, N. Bambos, K. Kumaran, I. Saniee, and I. Widjaja, “Scheduling bursts in Time-Domain Wavelength Interleaved Networks,” IEEE J. Sel. Areas Commun. **21**, 1441–1451 (2003). [CrossRef]

### 3.2 Propagation delay compensation

*N*nodes with zero or equal propagation delay among all node pairs, we assume that traffic arrivals at all edge nodes are uniformly distributed. The utilization of the network is given by the following lemma (we define

*utilization*as the ratio of used slots to the total number of slots on the average):

*Lemma 3.2.1*: In a TWIN network of

*N*nodes and zero or equal propagation delay between each source-destination node pair, the number of slots in a period for full reachability and 100% utilization is

*N*-1.

*Proof*:

*N*-1 nodes, there must be

*N*-1 transmission in a period, with one slot transmission to each destination. In any slot, there should be

*N*-1 transmission from

*N*nodes for full reachability. Next, we show that

*N*-1 slots are sufficient for 100% utilization. With no propagation delay, a burst sent in the

*m*-th slot in the period will be received in the same slot. Hence we can give such a schedule: in the first slot, let each node select any of the

*N*-1 destinations with no two nodes selecting the same destination, e.g. node 1 selects node 2, node 2 selects node3, .., and node

*N*selects node 1. In fact, there are totally (

*N*-1)! combinations with no source-destination conflict for all source nodes in the first slot.

*N*-2 destinations without overlapping (except the destination already selected in the previous slot), e.g. node 1 selects node 3, node 2 selects node 4, .., and node N selects node 2. This procedure is repeated in the remaining

*N*-3 slots. In the last (

*N*-1)th slot, there will be just one destination left for each node. In particular, each node sends exactly one conflict-free burst to each of the other

*N*-1 destination nodes utilizing every slot in every period thus achieving 100% network utilization. This is also true for all source-destination node pairs with equal propagation delay because the delay will result in a constant shift of the arrival time at the destination without causing any collision among transmissions, a time-shifted version of the zero-propagation delay case. ▪

*N*-1 slots at the destination node. We have thus compensated for the difference in propagation delays

*virtually*using the property of the periodic

*N*-1 slots. This novel synchronization scheme is stated in the following Theorem.

*Theorem 3.2.1*: In a network of

*N*nodes, denoting the propagation delay between node

*i*and node

*j*by

*δ*

_{ij}(

*δ*

_{ij}is expressed in number of slots), if (

*δ*

_{ij}mod (

*N*-1)) (∀

*i*,

*j*) is an integer constant, the minimum number of slots in a period to achieve 100% utilization is

*N*-1.

*Proof*:

*δ*

_{ij}mod (

*N*-1)) is an integer constant” is satisfied, the conclusion is the same as zero or equal propagation delay case in lemma 3.2.1.

*i*and

*j*,

*m*-th slot of current period will arrive at the destination in slot

*q*=(

*m*+

*δ*

_{ij}) mod (

*N*-1) in a later period. Consider together the assumption of (1),

*q*=(

*m*+

*δ*

_{ij}) mod (

*N*-1)=(

*m*+

*p*) mod (

*N*-1). Hence,

*q*is only subject to

*m*. Applying the property of periodicity, an assignment with different

*m*implies different

*q*, and thus there will be no slot overlapping in any period. Hence, the requirement in lemma 3.2.1 is satisfied, fulfilling the condition for 100% utilization. ■

*N*-1 slots. To satisfy the condition (1), a small compensating delay is required to be added to each node pair (

*i*,

*j*). Denoting the

*largest*compensating delay added to each source-destination link (

*i*,

*j*) by

*β*

_{ij}(in number of slots), it is obvious that

*β*

_{ij}would be smaller than the period

*τ*. This can be derived as follows: denote

*δ*

_{ij}’ the original propagation delay of link (

*i*,

*j*) in the network and

*δ*

_{ij}the propagation delay after compensation, then

*δ*

_{ij}=

*δ*

_{ij}’+

*β*

_{ij}. According to (1), (

*δ*

_{ij}’+

*β*

_{ij}) mod (

*N*-1)=

*p*(

*p*is an integer constant and 0≤

*p*<

*N*-1). If (

*δ*

_{ij}’ mod (

*N*-1))≤

*p*,

*β*

_{ij}=

*p*-(

*δ*

_{ij}’ mod (

*N*-1))<

*N*-1; if (

*δ*

_{ij}’ mod (

*N*-1))>

*p*,

*β*

_{ij}=(

*N*-1)+

*p*-(

*δ*

_{ij}’ mod (

*N*-1))<

*N*-1. Therefore,

*β*

_{ij}<

*N*-1. Since the period

*τ*is quite small compared to the longest propagation delay, this is a reasonable scheme.

#### 3.2.1 Electrical domain delay compensation

*N*-1 specific optical delays are needed at each source node for synchronization to

*N*-1 destination nodes, which will be impractical to implement. To overcome this problem, we add compensating delays in the electrical domain and show that this is a feasible but suboptimal solution. Adding a compensating delay in the electrical domain implies queueing a burst in the buffer for the delay before transmission. Due to this delay, the ideal condition in theorem 3.2.1 will be violated. While slots at destination nodes remain conflict-free, conflicts will arise at source nodes with traffic to different destination nodes. For example, consider a traffic burst between a node pair (

*i*,

*j*) designated for transmission in slot

*t*

_{ij}according to theorem 3.2.1. The electrical delay compensation will shift the transmission slot from

*t*

_{ij}to ((

*t*

_{ij}+

*β*

_{ij}) mod (

*N*-1)). A conflict will arise if ((

*t*

_{ij}+

*β*

_{ij}) mod (

*N*-1)) maps to a slot for more than one node pair for source

*i*.

### 3.3 Conflict resolution algorithm

*Lemma 3.3.1*: In a TWIN network with

*N*nodes using a periodic

*N*-1 slots, let

*δ*

_{ij}denote the propagation delay between node

*i*and node

*j*, the average utilization of the network, when (

*δ*

_{ij}mod (

*N*-1)) (∀

*i*,

*j*) is uniformly distributed in [1,

*N*-1] and electrical domain delay compensation is used, is given by

*Str*(

*N*-1,

*k*) is the

*Stirling number of the second kind*.

*Proof*: The

*N*-1 slots in a period can be imagined as

*N*-1 boxes. If two or more transmission times (after delay compensation) fall into the same box, only one can be successful. Assume (

*δ*

_{ij}mod (

*N*-1)) is uniformly distributed in [1,

*N*-1], then the transmission slots to each destination in the period after electrical compensation are also uniformly distributed in [1,

*N*-1]. Hence, the problem becomes calculating the average throughput of each source node under uniform arrival process, equivalent to the problem in Ref. [6

6. B. Li and Y. Qin, “Traffic scheduling in a Photonic Packet Switching System with QoS Guarantee,” J. Lightwave Technol. **16**, 2281–2295 (1998). [CrossRef]

*CRA Algorithm:*

*While not the end of free slots in source node*

*Search for the next free slot at source node*

*Compute the corresponding destination slot based on the propagation delay*

*If conflict-free source-destination slots are available*

*Assign source-destination slots*

*Set Successful*

*End-If*

*End-While*

*If Successful is set*

*Accept connection request*

*Else*

*Reject connection request*

*End-If*

*N*).

*τ*? The period

*τ*defines the maximum delay for a service-slot or a multiple service-slot. Therefore, the period

*τ*will define the minimum delay bound. Higher delay bounds of integer multiples of

*τ*can also be provided. It appears that

*τ*should be as small as possible subject to the minimum slot size and the number of slots in each period.

### 3.4 Simulation results

^{st}column). In the table, a network load equal to 1 implies that every node has a slot of traffic to send to every other node in every period

*τ*. In this case, the conflicts reduce the utilization to 0.63 which is close to the result of lemma 3.3.1 and recent results in Ref. [13]. After applying the CRA algorithm, the resulting utilizations improve about 40 to 50% for network loads from 0.7 to 1.

^{nd}and 3

^{rd}columns). The results of experiment 2 show a similar improvement of utilization for CRA algorithm. The performance is slightly better for a network with more nodes. The reason is due to better choices for conflict resolution with more available slots.

*τ*, we conducted the third experiment: for a network of

*N*(

*N*=200) nodes, we increase the number of slots to 2(

*N*-1) in each period. We conduct this experiment to assess the effect of having double the number of slots in a period. In this experiment, each node requests for two slots and CRA can allocate non-consecutive slots in the period. The results are shown in Table 2. Small improvements can be perceived when CRA is used.

### 3.5 Periodic timeslot service

14. R. L. Cruz, “Quality of service guarantees in virtual circuit switched networks,” IEEE J. Sel. Areas Commun. **13**, 1048–1056, Aug. (1995). [CrossRef]

14. R. L. Cruz, “Quality of service guarantees in virtual circuit switched networks,” IEEE J. Sel. Areas Commun. **13**, 1048–1056, Aug. (1995). [CrossRef]

16. R. L. Cruz and C. M. Okino, “Service guarantees for window flow control,” in Proc. 34^{th} Allerton Conf. on Comm., Cont. & Comp., Oct. (1996). [PubMed]

*Definition 3.5.1*: The service offered by a stream of periodic slots to a aggregate flow

*I*between a source-destination pair is given by a

*slotted service curve*

*w*

_{I}is the number of slots assigned to the aggregate flow

*I*, and

*b*is the number of bits in a timeslot, and

*z*

_{I}τ is the period in the assignment for aggregate flow

*I*.

*w*

_{I}and

*z*

_{I}. By varying these two parameters, we could construct a

*slotted service curve*with specific timeslot duration and period. Figure 4 shows a

*slotted service curve*for

*w*

_{I}=1 and

*z*

_{I}=1. For the sake of exposition, the following notations will be used in the rest of this paper:

*S̄*

_{I}(

*t*) denotes the arrival curve of the aggregate flow

*I*,

*S*

_{I}(

*t*) denotes its service curve,

*A*

_{I}(

*t*) denotes its actual traffic envelope, and

*d*

_{I}denotes its

*queueing delay budget*. To represent an individual flow

*i*between this source-destination pair, we replace the subscript by a lowercase letter

*i*. The

*slotted service curve*offers a “staircase service” with step size

*w*

_{I}

*b*after each period

*z*

_{I}τ with no service between the periods. The proposed

*slotted service curve*has useful properties that provide deterministic QoS guarantee in relation to admission control and slot scheduling in the data plane in Section 4.

## 4. Data plane

### 4.1 Class-based traffic assembly process

*z*

_{i}

*τ*seconds. Figure 5 shows

*M*+1 buffers for

*M*real-time traffic classes and one best-effort traffic class. Also shown in the figure are flow specific traffic monitors

*i*that identify and forward conformant and non-conformant traffic to their buffers.

22. C. K. Siew and M. H. Er, “A new multiservice provisioning mechanism with service curves assurance for per-class scheduling delay guarantees” in press for IEE Proc. Communications. Available at URL: http://www.icis.ntu.edu.sg/our_institute/staff/cksiew/revised_COM_2005_0246-uncorrected%20final%20draft.pdf

### 4.2 Flow classification

22. C. K. Siew and M. H. Er, “A new multiservice provisioning mechanism with service curves assurance for per-class scheduling delay guarantees” in press for IEE Proc. Communications. Available at URL: http://www.icis.ntu.edu.sg/our_institute/staff/cksiew/revised_COM_2005_0246-uncorrected%20final%20draft.pdf

*D*={

*D*

_{M},

*D*

_{M-1}, ..,

*D*

_{1}}, where

*D*

_{M}<

*D*

_{M-1}<..<

*D*

_{1}are the delay bounds of the classes from

*M*to 1 respectively. The basic idea is that flows classified into a particular class will have the same class-based delay bound. The following definition specifies how a flow is classified into a particular class shown in Fig. 6.

*Definition 4.2.1*: A flow i is classified into class

*k*if the queueing delay budget

*d*

_{i}is

### 4.3 Class-based scheduler

*w*slots per period

*τ*to a source-destination node pair. The switch (Fig. 5) will close for w slots transmission every period

*τ*. Just before the switch is closed, traffic is assembled into the slot according to their priority class as described in subsection 4.1.

*M*traffic classes. We define a capacity allocation vector

**={**

*α**α*

_{M},

*α*

_{M-1}, ..,

*α*

_{1}}, where

*α*

_{M}<

*α*

_{M-1}<..<

*α*

_{1}and Σ

_{k}

*α*

_{k}=1, reserving capacity for various classes of traffic. It does not preclude a network operator from allocating the full capacity to one class of traffic, a special case of class-based scheduling. Using the allocation vector, we propose the following theorem for per-flow service guarantees by means of class-based traffic treatment.

*Theorem 4.3.1*: For a

*slotted service curve*of

*wb*bits per

*τ*second allocated to an ingress-egress node pair with

*M*classes and a capacity allocation vector

**={**

*α**α*

_{M},

*α*

_{M-1}, ..,

*α*

_{1}}, a flow

*j*with arrival curve

*S̄*

_{j}(

*t*) could be admitted to class

*h*(1≤

*h*≤

*M*) by this optical slot scheduling mechanism if

*F*

_{k}(

*F*

_{h}) is the number of flows already accepted in class

*k*(

*h*).

*Proof:*

*M*classes of traffic, we apply flow classification to all QoS flows according to their delay budgets. For a flow to be admitted to class

*M*, we must ensure that the arrival traffic of class

*M*shifted by

*D*

_{M}is upper bounded by the

*slotted service curve*

*S*(

*t*). That is,

*M*classes, the following conditions are required:

*j*arrives, it could be admitted into class

*h*if and only if the service curve required is still upper bounded by the

*slotted service curve*. Hence the results in the theorem are obvious. This completes the proof. ■

### 4.4 Application examples

*j*are modeled as either the token bucket

*TB*(

*Σ*

_{j},

*ρ*

_{j}) or TSPEC (

*M*

_{j},

*p*

_{j},

*Σ*

_{j},

*ρ*

_{j}) in IETF specifications. While the token bucket provides a neat solution, the TSPEC traffic characteristic provides a more realistic model of the real traffic.

*j*could be characterized by the token bucket

*TB*(

*Σ*

_{j},

*ρ*

_{j}), where

*Σ*

_{j}and

*ρ*

_{j}denote the burstiness and long term rate of flow

*j*respectively. In this context, it is not difficult to show that either of these parameters determines whether this flow could be admitted. For ease of exposition, we neglect all out-ofprofile traffic as they are not protected by the scheduling mechanism. We apply theorem 4.3.1 to define a set of simple and concise constraints for admission control in the following proposition.

*Proposition 4.4.1*: For a

*slotted service curve*of

*wb*bits per

*τ*second allocated to an ingress-egress node pair with

*M*classes and a capacity allocation vector

**={**

*α**α*

_{M},

*α*

_{M-1}, ..,

*α*

_{1}}, a flow

*j*with requirement (

*Σ*

_{j},

*ρ*

_{j},

*d*

_{j}) could be admitted to class

*h*(1≤

*h*≤

*M*,

*D*

_{h}≤

*d*

_{j}≤

*D*

_{h}-1), by this optical slot scheduling mechanism if

*i*in class

*k*and

*i*in class

*h*and

*F*

_{h}is the set of existing flows (already admitted) in class

*h*.

*Proof:*

*j*into class

*M*, we must ensure that the traffic of the flow

*j*is upper bounded by the envelope of the arrival curve deduced from its service curve. That is,

*i*in class

_{M}and

*F*

_{M}is the set of existing flows in class

*M*. Inequalities (14) and (15) ensure that the class

*M*traffic arrival curve is upper bounded by the

*slotted service curve*shifted left by

*τ*and thus bound the delay and limit the capacity allocated to class

*M*traffic. When

*α*

_{M}=1, all the capacity is allocated to class

*M*and we have the scenario of single-class case.

*j*could be admitted in class

*M*-1 (delay bound

*D*

_{M-1}) if

*h*, the results follow. This completes the proof. ■

*j*could be admitted depends on whether its burstiness or its long-term rate causes the resulting traffic curve to exceed the guaranteed arrival curve, which is deduced from the

*slotted service curve*. By means of class-based traffic assembly and scheduling, the queueing delay bound achieved for higher priority class traffic is smaller than that of lower priority class traffic independent of capacity allocation.

*j*whose characteristics are given by the quadruple (

*M*

_{j},

*p*

_{j},

*Σ*

_{j},

*ρ*

_{j}) and a queueing delay budget

*d*

_{j}. Instead of providing full derivation based on theorem 4.3.1, we sketch a simple graphical solution in Fig. 7 due to lack of space. In this figure, we plot the arrival curves of the flow based on

*TB*(

*Σ*

_{j},

*ρ*

_{j}) and

*TSPEC*(

*M*

_{j},

*p*

_{j},

*Σ*

_{j},

*ρ*

_{j}) as

*S̄*

_{1}(

*t*) and

*S̄*

_{2}(

*t*). We sketch the rates

*r*

_{1}and

*r*

_{2}which are needed to provide the deterministic guarantees according to

*S̄*

_{1}(

*t*) and

*S̄*

_{2}(

*t*). The graphical solution shows that for a given queueing delay budget

*d*

_{j},

*r*

_{1}≥

*r*

_{2}. The tight arrival curve based on IETF specification allows a lower minimum rate for the same queueing delay budget. It allows a more accurate computation of the service curve based on a more realistic arrival curve. This result shows that by means of the realistic IETF specification, it could admit more flows for a given capacity, leading to a more efficient utilization of resources.

### 4.5 Simulation result

*ρ*,

*Σ*), in which

*ρ*=0.5Mb/s and

*Σ*=40kbit. The delay bound of the three classes are 20ms, 40ms, 80ms respectively and the capacity allocation vector {

*α*

_{3},

*α*

_{2},

*α*

_{1}}={0.3, 0.3, 0.4}. We compute the number of flows admitted into each class without violating the service contract.

*n*

_{3}that can be admitted in class 3 (highest priority) must satisfy:

*n*

_{3}×0.5×10

^{6}≤0.3×4×10

^{6}⇒

*n*

_{3}=2 Similarly, the number of flows

*n*

_{2}that can be admitted in class 2 must satisfy:

*n*

_{2}×0.5×10

^{6}≤0.3×4×10

^{6}⇒

*n*

_{2}=2

*n*

_{1}that can be admitted in class 1 must satisfy:

*n*

_{1}×0.5×10

^{6}≤0.4×4×10

^{6}⇒

*n*

_{1}=3

## 5. Conclusion

*N*) for a

*N*edge nodes optical network. We have proposed a comprehensive aggregation, buffering, and scheduling mechanism to provide deterministic QoS guarantees using class-based traffic treatment. Through our analysis, we prove that the framework realizes Service Curves assurance that provides class-based queueing-delay bounds. Our simulation results on our framework, for the TWIN Network, have validated our analysis and shown that the class-based delay bounds are maintained. The simplicity and scalability of this framework make it suitable for implementation in slotted optical networks.

## References and links

1. | S. Yao, S. Dixit, and B. Mukherjee, “Advances in Photonic Packet Switching: an overview,” IEEE Commun. Mag. |

2. | C. Qiao and M. Yoo, “Optical Burst Switching (OBS) - A new Paradigm for an Optical Internet,” J. High Speed Nets. |

3. | I. Widjaja, I. Saniee, R. Giles, and D. Mitra, “Light core and intelligent edge for a flexible, thin-layered and cost-effective optical transport network,” IEEE Commun. Mag. |

4. | C. Nuzman and I. Widjaja, “Time-domain wavelength interleaved networking with wavelength reuse,” in Proc. IEEE INFOCOM’06, Mar. (2006). |

5. | N. Golmie, T. Ndousse, and D. Su, “A differentiated optical service model for WDM Networks,” IEEE Commun. Mag. |

6. | B. Li and Y. Qin, “Traffic scheduling in a Photonic Packet Switching System with QoS Guarantee,” J. Lightwave Technol. |

7. | M. Yoo, C. Qiao, and S. Dixit, “Optical burst switching for service differentiation in the next generation optical Internet,” IEEE Commun. Mag. |

8. | Maode Ma and M. Hamdi, “Providing deterministic quality-of-service guarantees on WDM Optical Networks,” IEEE J. Sel. Areas Commun. |

9. | B. Teitelbaum et al., QBone Architecture (v1.0), Internet 2 QoS Working Group Draft, Aug. 1999 |

10. | H. A. Mantar, J. S. Hwang, I. T. Okumus, and S. J. Chapin, “A scalable model for interbandwidth broker resource reservation and provisioning,” IEEE J. Sel. Areas Commun. |

11. | K. Ross, N. Bambos, K. Kumaran, I. Saniee, and I. Widjaja, “Scheduling bursts in Time-Domain Wavelength Interleaved Networks,” IEEE J. Sel. Areas Commun. |

12. | M. Chen and T. -S. Yum, “A conflict-free protocol for Optical WDMA Networks,” in Proc. IEEE GLOBECOM, Dec. (1991). |

13. | I. Saniee and I. Widjaja, “Simplified layering and flexible bandwidth with TWIN,” in Proc. Workshop on Future Directions in Network Architecture, SIGComm, (2004). |

14. | R. L. Cruz, “Quality of service guarantees in virtual circuit switched networks,” IEEE J. Sel. Areas Commun. |

15. | J. -Y. Le Boudec and P. Thiran, |

16. | R. L. Cruz and C. M. Okino, “Service guarantees for window flow control,” in Proc. 34 |

17. | H. Sariowan, “A Service-curve approach to performance guarantees in integrated-service networks,” Ph.D. thesis, Dept of Electrical & Computer Engineering, UCSD, June (1996). |

18. | J. Schmitt, P. Hurley, M. Hollick, and R. Steinmetz, “Per-flow guarantees under class-based priority queueing,” in Proc. IEEE GLOBECOM, Nov. (2003). |

19. | J. Zheng, V. O. K. Li, and X. Yuan, “An adaptive flow classification Algorithm for IP switching,” in Proc. IEEE GLOBECOM, Nov. (1999). |

20. | W. Wang and C. C. Shen, “An adaptive flow classification scheme for data-driven label switching networks,” in Proc. IEEE ICC, June (2001). |

21. | K. Yasukawa, K. Baba, and K. Yamaoka, “Dynamic class assignment for stream flows considering characteristics of non-stream flow classes,” ICICE Trans. Commun. |

22. | C. K. Siew and M. H. Er, “A new multiservice provisioning mechanism with service curves assurance for per-class scheduling delay guarantees” in press for IEE Proc. Communications. Available at URL: http://www.icis.ntu.edu.sg/our_institute/staff/cksiew/revised_COM_2005_0246-uncorrected%20final%20draft.pdf |

**OCIS Codes**

(060.4250) Fiber optics and optical communications : Networks

(060.4510) Fiber optics and optical communications : Optical communications

**ToC Category:**

Fiber Optics and Optical Communications

**History**

Original Manuscript: August 29, 2006

Revised Manuscript: November 27, 2006

Manuscript Accepted: December 12, 2006

Published: December 22, 2006

**Citation**

Chee Kheong Siew, Daojun Xue, Yang Qin, and Jens Schmitt, "Providing deterministic quality of service in slotted optical networks," Opt. Express **14**, 12679-12692 (2006)

http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-14-26-12679

Sort: Year | Journal | Reset

### References

- S. Yao, S. Dixit, and B. Mukherjee, "Advances in Photonic Packet Switching: an overview," IEEE Commun. Mag. 38, 84-94, Feb. (2000). [CrossRef]
- C. Qiao and M. Yoo, "Optical Burst Switching (OBS) - A new Paradigm for an Optical Internet," J. High Speed Nets. 8, 69-84 Jan. (1999).
- I. Widjaja, I. Saniee, R. Giles, and D. Mitra, "Light core and intelligent edge for a flexible, thin-layered and cost-effective optical transport network," IEEE Commun. Mag. 41, 30-36, (2003). [CrossRef]
- C. Nuzman and I. Widjaja, "Time-domain wavelength interleaved networking with wavelength reuse," in Proc. IEEE INFOCOM’06, Mar. (2006).
- N. Golmie, T. Ndousse and D. Su, "A differentiated optical service model for WDM Networks," IEEE Commun. Mag. 38, 68-73, Feb. (2000). [CrossRef]
- B. Li and Y. Qin, "Traffic scheduling in a Photonic Packet Switching System with QoS Guarantee," J. Lightwave Technol. 16, 2281-2295 (1998). [CrossRef]
- M. Yoo, C. Qiao and S. Dixit, "Optical burst switching for service differentiation in the next generation optical Internet," IEEE Commun. Mag. 39, 98-104, Feb. (2001). [CrossRef]
- Maode Ma and M. Hamdi, "Providing deterministic quality-of-service guarantees on WDM Optical Networks," IEEE J. Sel. Areas Commun. 18, 2072-2083 (2000). [CrossRef]
- B. Teitelbaum et al., QBone Architecture (v1.0), Internet 2 QoS Working Group Draft, Aug. 1999
- H. A. Mantar, J. S. Hwang, I. T. Okumus and S. J. Chapin, "A scalable model for interbandwidth broker resource reservation and provisioning," IEEE J. Sel. Areas Commun. 22, 2019-2034 (2004). [CrossRef]
- K. Ross, N. Bambos, K. Kumaran, I. Saniee, and I. Widjaja, "Scheduling bursts in Time-Domain Wavelength Interleaved Networks," IEEE J. Sel. Areas Commun. 21, 1441-1451 (2003). [CrossRef]
- M. Chen and T. -S. Yum, "A conflict-free protocol for Optical WDMA Networks," in Proc. IEEE GLOBECOM, Dec. (1991).
- I. Saniee and I. Widjaja, "Simplified layering and flexible bandwidth with TWIN," in Proc. Workshop on Future Directions in Network Architecture, SIGComm, (2004).
- R. L. Cruz, "Quality of service guarantees in virtual circuit switched networks," IEEE J. Sel. Areas Commun. 13, 1048-1056, Aug. (1995). [CrossRef]
- J. -Y. Le Boudec and P. Thiran, Network Calculus - A Theory of deterministic queueing systems for the Internet, (Lecture Notes in Computer Science, Springer 2001).
- R. L. Cruz and C. M. Okino, "Service guarantees for window flow control," in Proc. 34th Allerton Conf. on Comm., Cont. & Comp., Oct. (1996). [PubMed]
- H. Sariowan, "A Service-curve approach to performance guarantees in integrated-service networks," Ph.D. thesis, Dept of Electrical & Computer Engineering, UCSD, June (1996).
- J. Schmitt, P. Hurley, M. Hollick, and R. Steinmetz, "Per-flow guarantees under class-based priority queueing," in Proc. IEEE GLOBECOM, Nov. (2003).
- J. Zheng, V. O. K. Li and X. Yuan, "An adaptive flow classification Algorithm for IP switching," in Proc. IEEE GLOBECOM, Nov. (1999).
- W. Wang, C. C. Shen, "An adaptive flow classification scheme for data-driven label switching networks," in Proc. IEEE ICC, June (2001).
- K. Yasukawa, K. Baba and K. Yamaoka, "Dynamic class assignment for stream flows considering characteristics of non-stream flow classes," ICICE Trans. Commun. 11, 3242-3254, Dec. (2004).
- C. K. Siew and M. H. Er, "A new multiservice provisioning mechanism with service curves assurance for per-class scheduling delay guarantees" in press for IEE Proc. Communications. Available at URL: http://www.icis.ntu.edu.sg/our_institute/staff/cksiew/revised_COM_2005_0246-uncorrected%20final%20draft.pdf

## Cited By |
Alert me when this paper is cited |

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.

« Previous Article | Next Article »

OSA is a member of CrossRef.