Wednesday 29 September 2010

Book Review: The Art of Scalability


The Scalable Computing Programme examines several dimensions and delivery models of scalable digital systems. Cloud Computing, Multi-core processors, Large Scale Complex IT Systems are the channels we have chosen to examine process, people and technology challenges associated with truly scalable, reliable digital systems. But addressing scalability – across technologies, architectures, delivery models and organisation is quite a challenge.
I found a book on scalability that covers technical, human, managerial, procedural, practical and theoretical dimensions of Scalability. I am presenting a quick overview of the same here.
The book is a first of its kind – as it is well rounded in its approach. It is written by two of the brightest minds who have actually have worked with challenging enterprise architecture models. They have built reliable, round the clock scalable applications – eBay, Paypal, Quigo and many more.
The book has four major sections
  1. Staffing a Scalable Organisation: In this section, the book examines the impact of people and leadership on scalability and suggests a set of roles for a scalable technology organisation. Starting with basic definition, authors have analysed how various elements of organisation type, management, and leadership impact scalability. The section also contains practical advice on building a strong business case for scale.
  2. Building Processes For Scale: Authors emphasise the fact that processes are quite critical to scale. They do so by critically examining role of well designed processes to manage incidents, problems, crisis and escalations. They have also examined how change in production environments affect the scale and needs to be managed.
Having laid a strong process foundation for scale, the authors present twelve architectural principles (N+1 Design, Stateless Systems, Scale Out Not Up etc.) and six scalability principles (Design to be monitored, design at least two axes of scale etc.) The true experience and maturity of authors is visible in these chapters.
There is also an in-depth discussion on typical trade-offs – such as Build or Buy, Fast or Right and impact of each choice on scalability.
The authors provide four handles – organisational tools – to manage scalability: JAD (Joint Architecture Design), ARB (Architecture Review Board), PST (Performance and Stress Testing for Scalability) and Barrier Conditions & Rollback. There is very practical advice on using these tools and integrating them into the organisation’s process framework to achieve scale and performance.
  1. Architecting Scalable Solutions: This section begins by refreshing design principles. Here, the authors have introduced concepts such as Technology Agnostic Architecture (TAD) and Technology Agnostic Design (TAD). But aren’t architecture and design technology agnostic, by definition?
The authors provide AKF Scale Cube – a three axes method to model scalability.
·    X-Axis: Represents cloning of services or data such that work can be distributed across instances. Implementation of X-axis is said to be relatively easy and inexpensive.
·    Y-Axis: Represents separation of work by responsibility, action or data. Going along Y-axis helps scaling transactions.
·    Z-Axis: Represents separation of work by customer or requestor. It is expensive to go along Z-axis.
The authors apply scale cube to explain how applications and data can be split to achieve scale. Synchronisation issues and impact of various types of caching methods on scalability are also discussed.
  1. Solving Other Issues and Challenges: This section discusses problems associated with proliferation of data and how they can be handled. Also included are the detailed comparative discussion on clouds and the grids. There is also a valuable chapter on designing data centres based on Three Magic Rules of Three. Interestingly, three is a magic number for data centres!
This section and everything else explained in this book culminates in three case studies: eBay, Quigo and ShareThis
The appendix consists of formulae to calculate availability, capacity, load and performance.
It is quite an interesting book, worthy of being a text on scalable computing.
Book Details:
Full Title              The Art of Scalability: Scalable Web Architecture, Processes, and Organizations for the Modern Enterprise
Authors                Martin L. Abbott; Michael T. Fisher
Edition                  1st Edition
Publisher             Addison-Wesley Professional
Pub. Date             December 16, 2009
ISBN Numbers      Print (1) 0-13-703042-8 (2) 978-0-13-703042-2 Web (1) 0-13-703143-2 (2) 978-0-13-703143-6
Pages                   592 in Print Edition
Rating                   4.5 out of 5 rating [12 Ratings]
Price                    Around 24£ to 20£

Sustainable IT: EC Code of Conduct for Data Centres

This article examines the European Commission’s Code of Conduct on Data Centres Energy Efficiency.

What is it? It is a voluntary initiative that aims to inform and stimulate data centre operators and owners to reduce energy consumption in a cost-effective manner without hampering the mission critical function of data centres.


Why do we need it? Many data centres are poorly designed with large tolerances allowing for capacity changes, and possible future expansion. Thus they end up using significant power, the majority of which is consumed by redundant power supplies and cooling systems. Hence selectively switching off IT systems does not result in significant energy savings.

Power used by data centres contributes substantially to the overall electricity consumed in the European Union commercial sector. Also, power costs are rising as a percentage of overall IT costs. Hence there is an urgent need to encourage data centres to take remedial action.

But don't we already have such measures within the industry? Yes, we do. But there is a risk of confusion, mixed messages and uncoordinated activities. A central, EC-wide Code of Conduct helps the cause.

Who should take note of this Code of Conduct? Data centre owners and operators, data centre equipment and component manufacturers, service providers, and other large procurers of such equipment.

The Code of Conduct classifies organisations as Participants (Data Centre Owners and Operators) and Endorsers (supply chain and service providers including vendors, consultancies, utilities, Government, Industry Associations/Standards bodies, educational institutions).

How does it aim to achieve its objective? By proposing general principles and practical actions to be followed by all parties involved in data centres, operating in the EU.

The Code of Conduct considers the entire data centre as a complete system. It provides guidelines and best practices for existing and new data centres covering IT Load and Facilities Load at Equipment Level and System Level.

What are the components of the Code of Conduct? It has a Secretariat and three Working Groups to establish and monitor commitments, and oversee the Code. 

Working Group
Chairman
Objectives
Best practices
Liam Newcombe
BCS Data Centre Specialist Group
Explore and exploit energy saving opportunities in the data centre
Energy efficiency metrics and measurements
Jan Viegand
Danish Energy Agency and the Danish Electricity Saving Trust
To develop a method to measure the electricity consumption and energy efficiency of data
centres and server rooms
Data collection and analysis
Anson Wu
UK Department for Environment (DEFRA)
To measure the energy consumption, calculate the energy efficiency of data centres and establish performance benchmarking

The Secretariat is composed of representatives of the European Commission DG JRC and the chair persons of the three working groups

What could we do next?
  • Initially: To qualify as a participant, submit an initial report describing the simple physical and operational characteristics of your data centre along with the most recent one month facility and IT energy consumption details.
  • Regularly provide energy usage details to the designated working group of the Code of Conduct.
  • Within 3 years plan to meet Expected Minimum Levels of energy saving. These levels are identified out of best practices documented by the Code of Conduct.
Who is already part of the Code of Conduct? Some of the participants include: Fujitsu, HP, Intel, the Met Office, Microsoft, Business&Decision, evoswitch, Lamda Hellix Datacentres, Memset, Petroleum Geo-Services, reed.co.uk, TCN, Telecity Group, Telekom Austria, VCD and Vodafone

DSKTN View: The EC Code of Conduct provides a framework which helps data centres to target energy reductions in a more structured manner, leading to financial, environmental and infrastructure benefits. Also, flexibility and continuous improvement methods built into the code make it easier to adapt to a variety of national efficiency programmes, climates and energy infrastructures.

We suggest participants consider the following to help comply with the code
  • Modern process architectures such as multi-cores
  • Virtualisation to increase hardware utilisation
  • Carbon costs associated with software enhancements, replacements, retirement
  • Value of IT to minimise/avoiding travel and deliver services
  • Embrace cloud based services where possible – even the futuristic Data Center as a a Service
We will endeavour to bring experts associated with the Code of Conduct to you as soon as possible. We encourage you to join our Sustainable IT Special Interest Group on our website.

Tuesday 28 September 2010

Exciting webinars on fresh topics

Currently we are running a new webinar series on fresh topics related to Scalable Computing, Cloud Computing and Sustainable IT. Each webinar features an industry expert and his/her area of work.

Our idea is to deliver around 13 webinars over 15 weeks. As of now, we have delivered 2 webinars and 6 others are already scheduled.


  • Sharing Research Assets Over The Cloud: 13th Oct 2010 1400-1430 GMT.
Listen to Prof. Jim Austin - Lead, Advanced Computer Architectures Group, University of York about various challenges associated with sharing research assets over the cloud.

  • A Cloud Platform for R&D: 3rd Nov 2010 1400-1430 GMT.
Prof. Paul Watson - Director of North East Regional e-Science Centre, Newcastle University will introduce a cloud based platform for managing research and development activities.
 
  • HPC Scalability: Multicore to Exascale: 10th Nov 2010 1400-1430 GMT.
John Barr - Research Director & Financial Markets and Head of EU Research of The 451 Group will examine nature of HPC and challenges in managing its scalability highlighting Multi-scale processors, their impact on HPC provisioning and other associated challenges.
 
  • Processing and communications challenges in building the world’s largest radio telescope: 17th Nov 2010 1400-1430 GMT.
Know more about Square Kilometre Array (SKA) programme, a €1.5 billion international project to develop the world’s largest telescope and the UK's contribution to it. Specifically learn about Data Processing and Communication challenges associated with this telescope. And learn about potential opportunities for the UK businesses.
Speaker: Andrew Faulkner, Project Engineer - European SKA Design Studies with University of Cambridge
 
  • Elastic Cloud Services using Amazon EC2: 24th Nov 2010 1400-1430 GMT.
Matt Wood, EMEA Evangelist with Amazon UK will talk about cloud services, advantages that elastic infrastructure, and on-demand provisioning can have for capacity planning, agility and bringing new ideas to market quickly.

You are welcome to participate in these webinars.