Rack-scale Disaggregated Cloud Data Centers: The dReDBox Project Vision

Katrinis, Kostas; Syrivelis, Dimitris; Pnevmatikatos, D.; Zervas, Georgios; Theodoropoulos, Dimitris; Koutsopoulos, Iordanis; Hasharoni, Kobi; Raho, Daniel; Pinto, Christian; Espina, F.; López-Buedo, Sergio; Chen, Qianqiao; Nemirovsky, Mario; Roca, Damian; Klos, H.; Berends, T.

doi:10.3850/9783981537079_1014

Cited by 57 publications

(54 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors in [31] studied the trade-off between cost and performance of building a disaggregated memory system; they constructed a simple cost model that compares the savings expected from a disaggregated memory system to the expected costs, such as latency and bandwidth costs, and then identified the level at which a disaggregated memory system becomes cost competitive with a traditional directattached memory system. In [32], the authors proposed a software-defined architecture for the next generation data center, dRedBox, and presented a design prototype hardware architecture where system-on-chip (SoC)-based microservers, memory modules, and accelerators are placed in separated modular server trays interconnected via a high-speed, lowlatency optoelectronic system fabric and are allocated in arbitrary sets. In [33][34][35], we presented a detailed description and comprehensive review of the idea of DS, with a focus on the energy efficiency benefits of this design.…”

Section: Future Energy Efficient Data Centers With Disaggregated Serversmentioning

confidence: 99%

“…Generating the traffic demand matrix: (40) (41) (42) Constraint (40) generates the CPU-MEM traffic matrix based on the index matrix calculated previously in constraints (32) and (33), and which is the VM CPU-MEM traffic demands. Constraints (41) and (42) generate the CPU-IO and Memory-IO traffic matrices respectively, using the relevant variables.…”

Section: ) Communication Constraintsmentioning

confidence: 99%

See 1 more Smart Citation

Future Energy Efficient Data Centers With Disaggregated Servers

Ali

El-Gorashi

Lawey

et al. 2017

J. Lightwave Technol.

View full text Add to dashboard Cite

Abstract-The popularity of the Internet and the demand for 24/7 services uptime is driving system performance and reliability requirements to levels that today's data centers can no longer support. This article examines the traditional monolithic conventional server (CS) design and compares it to a new design paradigm: the disaggregated server (DS) data center design. The DS design arranges data centers resources in physical pools, such as processing, memory, and IO module pools, rather than packing each subset of such resources into a single server box. In this work, we study energy efficient resource provisioning and virtual machine (VM) allocation in DS-based data centers compared to CS-based data centers.

show abstract

Section: Future Energy Efficient Data Centers With Disaggregated Serversmentioning

confidence: 99%

Section: ) Communication Constraintsmentioning

confidence: 99%

Future Energy Efficient Data Centers With Disaggregated Servers

Ali

El-Gorashi

Lawey

et al. 2017

J. Lightwave Technol.

View full text Add to dashboard Cite

show abstract

“…With conventional DCs experiencing drawbacks in server resource utilization, power consumption and scalability, novel solutions must be explored to effectively manage this traffic trend. Disaggregation of servers into separate standalone CPU, memory, storage and network resource pools has been projected in [2,3] as a promising solution to improve resource utilization, efficiency, modularity and upgradeability in DCs. In addition, hybrid DC architectures employing both electronic and optical technologies [4][5][6] and all-optical technologies [7][8][9] have been explored and have demonstrated benefits in terms of power consumption, modularity and cost in comparison to conventional DCs.…”

Section: Introductionmentioning

confidence: 99%

In Compute/Memory Dynamic Packet/Circuit Switch Placement for Optically Disaggregated Data Centers

Peters¹,

Oikonomou²,

Zervas³

2018

J. Opt. Commun. Netw.

View full text Add to dashboard Cite

Network function services on conventional hybrid data centres (DCs) architectures such as HELIOS are hard-wired and dedicated to specific network resources. This limits flexibility and performance to handle diverse traffic patterns. Furthermore, disaggregation of server resources has shown promising potential to improve resource utilization which has been a limitation of conventional server-centric DCs. This paper presents a reconfigurable hybrid disaggregated DC (dRedBox) architecture which combines the concept of server resource disaggregation with cutting edge software, electronic and optical technologies. The dRedBox architecture provides a remarkable amount of flexibility and connectivity through hardware based multi-layer network function service programmability. This allows for multi-layer network services to be dynamically deployed at run-time to network resources and in turn handle diverse traffic patterns. Furthermore, this study proposes algorithms and strategies for selecting and deploying electronic packet switching and optical circuit switching function services to implement Virtual Machine network requests across dRedBox and conventional hybrid disaggregated architectures under different traffic patterns. Finally, the performance of the various strategies on the dRedBox and conventional hybrid disaggregated DC architectures are evaluated in terms of blocking probability, energy efficiency, network utilization and cost. Extensive results show that at 10% blocking probability, dRedBox architecture achieves 100% gain on VM placement and 92% energy savings compared to conventional hybrid disaggregated architectures.

show abstract

“…Disaggregation of traditional DC resource can be performed at different scales i.e. rack-scale [5], [17] and pod-scale [18], [19] as shown in Figure 1(a) and 1(b) respectively. Likewise, the network interconnecting disaggregated resources can be electrical, hybrid (electro-optical) or all optical.…”

mentioning

confidence: 99%

On Energy Efficiency of Networks for Composable Datacentre Infrastructures

Ajibola

El-Gorashi

Elmirghani

2018

2018 20th International Conference on Transparent Optical Networks (ICTON)

View full text Add to dashboard Cite

This paper evaluates the optimal scale of datacentre (DC) resource disaggregation for composable DC infrastructures and investigates the impact of present day silicon photonics technologies on the energy efficiency of different composable DC infrastructures. We formulated a mixed integer linear programming (MILP) model to this end. Our results show that present day silicon photonics technologies enable better network energy efficiency for rack-scale composable DCs compared to pod-scale composable DCs despite reported similarities in CPU and memory resource power consumption.

show abstract

Rack-scale Disaggregated Cloud Data Centers: The dReDBox Project Vision

Cited by 57 publications

References 6 publications

Future Energy Efficient Data Centers With Disaggregated Servers

Future Energy Efficient Data Centers With Disaggregated Servers

In Compute/Memory Dynamic Packet/Circuit Switch Placement for Optically Disaggregated Data Centers

On Energy Efficiency of Networks for Composable Datacentre Infrastructures

Contact Info

Product

Resources

About