Network-aware VM Allocation for Predictable Performance: Christine Bassem, BU (NRG Seminar)

  • Starts: 11:00 am on Monday, April 14, 2014
  • Ends: 12:00 pm on Monday, April 14, 2014
Abstract: The talk will cover the following two papers: (1) - Towards predictable datacenter networks - SIGCOMM 2011 The shared nature of the network in today's multi-tenant datacenters implies that network performance for tenants can vary significantly. This applies to both production datacenters and cloud environments. Network performance variability hurts application performance which makes tenant costs unpredictable and causes provider revenue loss. Motivated by these factors, this paper makes the case for extending the tenant-provider interface to explicitly account for the network. We argue this can be achieved by providing tenants with a virtual network connecting their compute instances. To this effect, the key contribution of this paper is the design of virtual network abstractions that capture the trade-off between the performance guarantees offered to tenants, their costs and the provider revenue. To illustrate the feasibility of virtual networks, we develop Oktopus, a system that implements the proposed abstractions. Using realistic, large-scale simulations and an Oktopus deployment on a 25-node two-tier testbed, we demonstrate that the use of virtual networks yields significantly better and more predictable tenant performance. Further, using a simple pricing model, we find that the our abstractions can reduce tenant costs by up to 74% while maintaining provider revenue neutrality. (2) - The only constant is change: incorporating time-varying network reservations in data centers - SIGCOMM 2012 In multi-tenant datacenters, jobs of different tenants compete for the shared datacenter network and can suffer poor performance and high cost from varying, unpredictable network performance. Recently, several virtual network abstractions have been proposed to provide explicit APIs for tenant jobs to specify and reserve virtual clusters (VC) with both explicit VMs and required network bandwidth between the VMs. However, all of the existing proposals reserve a fixed bandwidth throughout the entire execution of a job. In the paper, we first profile the traffic patterns of several popular cloud applications, and find that they generate substantial traffic during only 30%-60% of the entire execution, suggesting existing simple VC models waste precious networking resources. We then propose a fine-grained virtual network abstraction, Time-Interleaved Virtual Clusters (TIVC), that models the time-varying nature of the networking requirement of cloud applications. To demonstrate the effectiveness of TIVC, we develop Proteus, a system that implements the new abstraction. Using large-scale simulations of cloud application workloads and prototype implementation running actual cloud applications, we show the new abstraction significantly increases the utilization of the entire datacenter and reduces the cost to the tenants, compared to previous fixed-bandwidth abstractions.
MCS 148

Back to Calendar