- Predictable Scaling Behaviour in the Data Centre with Multiple Application Servers
Proceedings of DSOM 2006 LNCS LNCS 4269 p49-60 (2006)
Load sharing in the data centre is an essential strategy for
meeting service levels in high volume and high availability
services. We investigate the accuracy with which simple, classical
queueing models can predict the scaling behaviour of server capacity
in an environment of both homogeneous and inhomogeneous hardware, using
known traffic patterns as input. We measure the performance of three
commonly used load sharing algorithms and show that the simple
queueing models underestimate performance needs significantly at high
load. Load sharing based on real-time network monitoring performs
worst on average. The work has implications for the accuracy of
Quality of Service estimates.
- Uncertainty in Global Application Services with Load Sharing Policy
Proceedings of DSOM2006 in LNCS 4269 p37-48 (2006)
With many organizations now employing multiple data centres
around the world to share global traffic load, it is important to
understand the effects of geographical distribution on service
quality. The Domain Name Service is an important component for global
load balancing. Using controllable simulations, we show that wide area
sharing can play an important role in optimization of response times
when traffic levels exceed that which can be supplied by a local
infrastructure. We compute the probability of being able to meet
Service Level Objectives as a function of DNS caching policy (Time To
Live), so that service providers can account for DNS error margins in
Service Level Agreements.
- On the Reliability of Service Level Estimators in the Data Centre
Submitted
The ability to predict achievable service levels is an
important part of business strategy in drawing up Service Level
Agreements (SLA). Service delivery is a dialogue between supply and
demand. Recently it has been argued that adaptive service levels
might be predicted using classical queueing theory combined with a
data centre model. Using generic tools, we investigate the accuracy
with which service levels can be predicted for a single server Web
service, based on expected traffic patterns, and hence we identify
approximate limits on the reliability for service levels.
- Adaptive Provisioning Using Virtual Machines and Autonomous Role-Based Management
ICAS'06 - SELF workshop - Self-adaptability and self-management of context-aware systems
A modern server must deliver a complex set of obligations towards its
users. We study the concept of roles as autonomous encapsulations of a set
of promises. Using virtualization and configuration management every
defined role for a real-life computer system can be realized as a
self-managing agent that is aware of the other related roles. A prototype implementation that provides web hotel services for customers us presented and
discussed.
- Scalable Deployment and Configuration of High-Performance Virtual Clusters
CISE/CGCS 2006: "3rd International Conference on Cluster and Grid Computing Systems"
Virtualization and high performance computing have been discussed from a
performance perspective in recent publications. We present and discuss a flexible and efficient approach to the management of virtual clusters. A virtual
machine management tool is extended to function as a fabric for cluster
deployment and management. We show how features such as saving the
state of a running cluster can be used to avoid disruption. We also compare
our approach to the traditional methods of cluster deployment and
present benchmarks which illustrate the efficiency of our approach.
