Friday, 9 July 2010

IEEE CLOUD 2010 Conference, Day 5

Interesting talks from Day 5:

Keynote: Thinking outside the Box: How Cloud, Grid, and Services Can Make Us Smarter? by Ian Foster (Argonne National Lab and University of Chicago).
Ian was one of the pioneers of the grid and his talk was mostly focused on how e-science researchers are using the grid to do their experiments and archive their data. He briefly talked about the efforts underway by the Globus project to develop Sci-SaaS (to provide software-as-a-service for scientists). The link between his talk and cloud computing was that the cloud provides scientists with scalable hosting for their applications and data. One of the interesting points that Ian made was this: one of the main lessons that they learned from grid computing was that the social and human issues to do with using the grid and incentives for sharing are really important. These issues need to be explored by researchers in the cloud to prevent the same barriers (that were faced by the grid community) slowing cloud adoption.

A Case for Consumer-Centric Resource Accounting Models (CLOUD2010-3064)
Ahmed Mihoob, Carlos Molina-Jimenez, Santosh Shrivastava (Newcastle University, UK)
A detailed account of how AWS charge for their storage, down to the fine details that are not mentioned in AWS’s website, they discovered these details by doing experiments on S3.

An Architecture for Public and Open Submission Systems in the Cloud (CLOUD2010-3065)
Marcello Azambuja, Rafael Pereira, Karin Breitman (; PUC-Rio, Rio Rio de Janeiro, Brazil)
This talk was about a conceptual architecture that is used for video processing and streaming where there are high-peaks at certain times (e.g. people submitting their trial videos for Big Brother in Brazil where people submit their video using a website to the judges, but people leave it late before they submit so there is often a very high peak just before the deadline). This type of architecture has also been documented in AWS’s whitepapers.

On-Demand Dynamic Security for Risk-Based Secure Collaboration in Clouds (CLOUD2010-3066)
Michael Boniface, Mike Surridge, Martin Hall-May, Stuart Bertram, Neil Briscombe (University of Southampton IT Innovation Centre, QinetiQ Ltd)
This talk was about a platform and lifecycle that can be used to model a system, its security policies and certain events with the aim of modeling security threats to support decision making. The lifecycle involves security planning, security provisioning, active threat identification and threat assessment. I’m not too sure where the cloud relation is in this work.

IEEE CLOUD 2010 Conference, Day 4

Interesting presentations from Day 4:

Keynote: Securing data n the cloud – challenges and research directions by Prof. Elisa Bertino (Purdue University)
Presented an overview of the security challenges in cloud computing, she mentioned that some of the issues are simply about people’s perceptions of security and CC providers should act to improve people’s perceptions of security in the cloud – technical improvements can help but most issues are centered on who has control over data and servers.

Cost-Optimal Scheduling in Hybrid IaaS Clouds for Deadline Constrained Workloads (CLOUD2010-3029)
Ruben Van den Bossche, Kurt Vanmechelen, and Jan Broeckhove(Universiteit Antwerpen Antwerp, Belgium)
Kurt talked about reducing the costs of running applications on a mixture of in-house infrastructure and third party IaaS, this work was based on simulating job inputs that had deadlines and required certain resources (e.g CPU, memory), and aiming to reduce the costs of completing the jobs using the in-house infrastructure and hypothetical third party IaaS providers.

Reducing Costs of Spot Instances via Checkpointing in the Amazon Elastic Compute Cloud (CLOUD2010-3030) Sangho Yi, Derrick Kondo, and Artur Andrzejak (INRIA Grenoble Rhône-Alpes, France; Zuse Institute Berlin (ZIB), Germany)
Here’s the abstract – it was an interesting bit of work but I didn’t quite understand how significant the cost reductions were as the graphs only showed a $2 or $3 improvement - maybe this was per hour so over time it would add up and be significant: Recently introduced spot instances in the Amazon Elastic Compute Cloud (EC2) offer lower resource costs in exchange for reduced reliability: these instances can be revoked abruptly due to price and demand fluctuations. Mechanisms and tools that deal with the cost-reliability trade-offs under this schema are of great value for users seeking to lessen their costs while keeping reliability at a high level. We study how one such a mechanism, namely checkpointing, can be used to minimize the cost and volatility of resource provisioning. Based on the real price history of the spot instances we compare several adaptive checkpointing schemes in terms of monetary costs and improvement of job completion times. A trace-based simulation shows that our approach can reduce significantly both the price and the task completion time.

Open Source Cloud Computing Tools: A Case Study with a Weather Application (CLOUD2010-3056) Manuel Rodriguez-Martinez, Jaime Seguel, and Melvin Greer(University of Puerto Rico, Mayagüez; Lockheed Martin)
This was an interesting evaluation of Eucalyptus/Hadoop/Django for processing weather data, they had the same painful experience of installing Eucalyptus as we did.

Cloud Migration: A Case Study of Migrating an Enterprise IT System to IaaS (CLOUD2010-3057)
Ali Khajeh-Hosseini, David Greenwood, and Ian Sommerville (University of St Andrews, UK)
My talk was after the above presentation, it went really well and there was plenty of time for a good question and answer session as the previous presentations finished early. There was a lot of interest from industry folk who agreed that the socio-technical issues around cloud adoption are certainly important and need to be further explored.

An Evaluation of Distributed Datastores Using the AppScale Cloud Platform (CLOUD2010-3039)
Chris Bunch, Navraj Chohan, Chandra Krintz, Jovan Chohan, Jonathan Kupferman, Puneet Lakhina, Yiming Li, Yoshihide Nomura(University of California, USA; Fujitsu Labs Ltd., Japan)
Navraj gave a really good overview of the AppScale open source project, which tries to achieve a similar goal to the Eucalyptus project in giving the community a free platform that they can use and experiment with. If you haven’t already seen this project, google it, it’s worth a visit.

IEEE CLOUD 2010 Conference, Day 3

Interesting presentations from Day 3:

FlexPRICE: Flexible Provisioning of Resources in a Cloud Environment (CLOUD2010-3011) Thomas A. Henzinger, Anmol V. Singh, Vasu Singh, Thomas Wies, and Damien Zufferey(IST Austria A-3400 Klosterneuburg, Austria)
This talk was about running applications on EC2, where the user needs to decide number of CPUs, memory etc. How can they decide what they need and what is the best for their apps? FlexPRICE enables users to submit jobs to the cloud, and providers a number of suggestions for when and how to schedule the job. A job is presented as a data flow diagram with nodes as tasks and attributes such as input data, duration of job, memory required, executable file to run. As one of the questioners said at the end of this presentation, there is no clear notion of jobs in the cloud, so this work is more useful for the grid.

Workload Migration into Clouds – Challenges, Experiences, Opportunities (CLOUD2010-3020)
C. Ward, N. Aravamudan, K. Bhattacharya, K. Cheng, R. Filepp, R. Kearney, B. Peterson, L. Shwartz, and C.C. Young(B. Peterson, L. Shwartz, C. C. Young (IBM, USA)
Presented IBM’s Darwin framework, which is an integrated framework to assist application migration to the cloud.

Towards Self-Assisted Troubleshooting for the Deployment of Private Clouds (CLOUD2010-3021) Michael R. Head, Anca Sailer, Hidayatullah Shaikh, and Dennis Shea (IBM T. J. Watson, USA)
About the management of VM images in private clouds, where they can detect faults early during deployment if VM images are corrupted. This technique is based on taking snapshots of the images and using checkpoints to recover from failures.

Panel 2: Trends of Services and Cloud Computing
This panel included Andrzej Goscinski (School of IT, Deakin University Australia), Hemant K. Jain (Tata Consulting Services Professor, University of Wisconsin at Milwaukee) and Min Lou (IBM). Each panellist gave a 15-minute presentation on their thoughts about trends in cloud computing. I don’t think anything new was said here, the usual comparisons of electricity and commuting as a utility were brought up.

An Architecture for a Mashup Container in Virtualized Environments (CLOUD2010-3049)
Michele Stecca and Massimo Maresca(DIST University of Genova; DEI University of Padova, Italy)
A reference model for event driven composite services or mash-ups. Based on the idea that a resource can be monitored and users can be notified when something interesting happens.

Mining Twitter in the Cloud: A Case Study (CLOUD2010-3014)
Pieter Noordhuis, Michiel Heijkoop, and Alexander Lazovik (University of Groningen, The Netherlands)
An interesting talk about how to use EC2 to create social graph from twitter that has 50 million nodes and 1.8 billion edges, not too sure if I understood the motivations for doing this work but nonetheless it was an interesting case study. Their main finding was that you should create customized and basic analysis services yourself and deploy them on EC2 instances, and not use AWS’s SQS or SimpleDB as they would be expensive.

Cloud Computing: A Digital Libraries Perspective (CLOUD2010-3015)
Pradeep Teregowda, Bhuvan Urgaonkar, and C. Lee Giles (Pennsylvania State University, USA)
An interesting case study of the feasibility of migrating CiteSeerX (which is deployed on physical 22 servers) from their in-house data centre to Amazon EC2 to take advantage of the cloud’s elasticity but also to free them from hardware maintenance so they can focus on software improvements.

A Lifetime Supporting Framework for Cloud Applications (CLOUD2010-3046)
Shigeru Hosono, He Huang, Tatsunori Hara, Yoshiki Shimomura, and Tamio Arai(NEC Corporation, Japan; NEC Advanced Software Technology (Beijing) Co., Ltd., China; The University of Tokyo; Tokyo Metropolitan University, Japan)
A suite of tools being developed at NEC (for NEC’s use, not open source) for the development and maintenance of applications for the cloud. Based on a heavy-weight lifecycle management approach that can be customized for each application development project.

Maximizing Cloud Providers' Revenues via Energy Aware Allocation Policies (CLOUD2010-3017)
Michele Mazzucco, Dmytro Dyachuk, and Ralph Deters (University of Cyprus, Cyprus; University of Saskatchewan, Canada; University of Tartu, Estonia)
Presented the results of simulations that were used to show how cloud providers could increase their revenues by shutting down idle physical servers and moving their VMs to other servers (i.e. load skewing). Tried to address the problem of deciding how many servers to keep switched on in a data centre by using different heuristics. They didn’t actually consider how the performance of applications would be affected by load skewing, and assumed that one VM would run on one core. They also didn’t consider the increased heat production and the cost of cooling the over-utilized servers.

Panel 3: Killer Applications in the Cloud
An interesting panel, but many of the issues that were mentioned have already been discussed in the literature.

IEEE CLOUD 2010 Conference, Day 2

Day 2 was the start of the main research presentations; I went to a few different talks, the following presentations were the more interesting ones:

Compliant Cloud Computing (C3): Architecture and Language Support for User-Driven Compliance Management in Clouds (CLOUD2010-3031)
Ivona Brandic, Schahram Dustdar, Tobias Anstett, David Schumm, Frank Leymann,and Ralf Konrad (Vienna University, Austria; University of Stuttgart, German; T-Systems International GmbH Frankfurt, Germanyy)
Ivona talked about some of the security, privacy and trust implications of cloud computing, it was refreshing to see someone who knew all the issues but had also done some research to address the issues. Most people agree that security, privacy and trust are important issues that need to be addressed in the cloud but very little work seems to be actually focused on addressing the issues. Ivona mentioned that the grid failed to address these issues and maybe these issues were simply not important for some of the workloads that run on grids. She provided an overview of their Compliant Cloud Computing (C3) middleware that supports developers who want to make use of several IaaS providers to host data. C3 has a language that can be used by domain experts who know about the data and who should be able to access what. Once this mapping is setup, the C3 middleware takes over and deploys the data on the selected IaaS providers while respecting the privacy requirement of the data. The middleware was demonstrated using a case study that was carried out with T-Systems International in Germany.

A Framework for Optimized Distribution of Tenants in Cloud Applications (CLOUD2010-3032)
Christoph Fehling, Frank Leymann, and Ralph Mietzner (University of Stuttgart Universit¨atsstr Germany)
Christoph talked about the distribution of application components amongst resources (where a resource can be infrastructure or platforms), and provided details of a framework that enables cloud providers to model their resources, users and their applications to find an optimum distribution.

Enterprise Cloud Service Architecture (CLOUD2010-3004)
Longji Tang, Jing Dong, Yajing Zhao, and Liang-Jie Zhang (University of Texas at Dallas IBM T.J. Watson Research Center,USA)
Longji talked about enterprise cloud services architecture, I didn’t quite understand why an enterprise would want to do this.

Towards Living Landscape Models: Automated Integration of Infrastructure Cloud in Enterprise Architecture Management (CLOUD2010-3005)
Matthias Farwick, Berthold Agreiter, Ruth Breu, Matthias Häring, Karsten Voges, and Inge Hanschke (University of Innsbruck, Austria, iteratec GmbHMunich, Germany)
Matthias talked about the problems of keeping enterprise architecture models of an enterprise’s IT infrastructure up to date and in-sync with their actual setup. Their aim is to keep the EA models in-sync and support enterprises to transition from an as-is model to a to-be model of their IT infrastructure. Currently EA models are created manually using information from project documentation and interviews, and it becomes challenging to keep these models up to date when IaaS is used as things can be easily changed. Matthias et al have started work on a tool that automatically queries virtual machines and their applications in order to keep the EA models up to date.

A Review of Cloud Business Models and Sustainability (CLOUD2010-3006)
Victor Chang, Gary Wills, and David De Roure(University of Southampton, UK)
Victor talked about the different business models that cloud providers use to make money.

Panel 1: Cloud Computing Standards
This panel included Wu Chou (Avaya Labs), David Bernstein (Huawei Ltd), James Wendorf (IEEE), Billy Cox (Intel), Don lee (NIST), and Alexander (IEEE). Each panelist gave a short talk about their thoughts on standards in cloud computing – they all agreed that standards were important and required. There are currently 12 organizations who are working on standards in cloud computing, they include CSA, NIST, OGF, Open Cloud Consortium, SNIA, Cloud Working Group, and the ITU focus group on CC. It was interesting to hear one of the panelists highlight the non-technical challenges of cloud adoption as being physiological as humans don’t always make rational decisions. He suggested that standards and legislation will help as they provide the stamp of approval for IT decision makers.

Keynote: Cloud Computing in an Outcome Centric World by Chung-Sheng Li (IBM T.J. Watson)
An interesting view of how more and more businesses are using services and only paying for them when they feel that they have achieved the outcome that they wanted. I guess measuring the value of different outcomes is the main challenge here as each business has some unique requirements from its IT systems.

Monday, 5 July 2010

IEEE CLOUD 2010 Conference, Day 1

So I'm presenting a paper at the IEEE 3rd International Conference on Cloud Computing (CLOUD2010) in Miami. The conference is co-located with 3 other conferences so it's a fairly big event and the organisers pride themselves on managing to squeeze 400+ talks during the conference (5-10 July). Today was the 1st day of the conference and it kicked off with an opening talk by the conference chairs. There were a few workshops but the main research presentations start from tomorrow, so stay tuned. One of the workshop talks I went to was about interoperability between clouds and the presenter (David Bernstein from Huawei Technologies, USA) pointed out a few links which I had not come across before:
  1. Global Inter-Cloud Technology Forum
  2. Open Cloud Consortium
  3. Inter Cloud (can't find a proper link to this)
  4. Standards and Interoperability for eInfrastructure Implementation Initiative
I have picked around 30 presentations that I want to attend during the next 4 days, but I'm not sure if I can see all of them since the organisers are running 7 tracks in parallel. There are also 4 keynotes and 3 discussion panels, as well as a poster session that I'm planning to go to. It will be a crazy few days.