Tuesday, 5 May 2015

Cloud Adoption in Asia

I was recently asked to compare and contrast the adoption and attitude towards cloud in Asia vs the UK. This is a little less straightfoward as the questioner intended, I think, since “Asia” is a very broad market, with several distinct regions and attitudes to Cloud and Software as a Service.
Still, let's look at how things stand in Japan, Korea, China, Australia, India & South-East Asia.


Japan is still generally a very conservative IT market, however Cloud & SaaS is being looked at carefully. There are definitely examples of major corporations making significant investments in cloud and especially SaaS technologies (for example Toyota is the largest customer of SalesForce.com).   While big enterprise has the resources to move into cloud, and especially private cloud, the chief adopters of public cloud are small-medium enterprise, with adoptions patterns very similar to western countries and for the same reasons of economy.

Korea can be even more conservative than Japan, and generally heavy reliance on older Windows technology (ActiveX, etc) from the early 2000's makes migration to Web 2.x & cloud/SaaS less straightforward in many cases. However, once one large enterprise moves, the rest of the market will follow quickly.

In Australia, cloud technologies are being investigated by large enterprises who have the resources to spare (eg the big banks), and up-coming businesses looking to leapfrog the old fixed-system way of working. Still, the "cloud first" mentality is still a couple of years behind the US.

China shows lots of interest in cloud, with several large service providers offering cloud and SaaS services, and there is a lot of focus on how to develop cloud services with a Chinese flavour rather than straight adoption of western (particular US) systems. In general, government still generally favours traditional application deployment, and security is a significant concern. The adoption trend shows that smaller organisations are most likely to adopt cloud, either by migrating existing process or taking the opportunity to move directly into public cloud. Private cloud are much less common. Larger organisations may see cloud as a way of leapfrogging old deployment methodologies, but the automation and TCO benefits of private cloud can be less attractive where low-cost, high-skilled labour is available. In general it’s still true that there is not a lot of radical new innovation in Chinese IT, but the scale of deployments often pushes the boundaries of technologies.

In South-East Asia, especially outside of Singapore, the scale of IT deployment is generally too small to really gain advantages of on-premise cloud. As with Australia, there are places where small businesses are looking to leapfrog the old way of working & go directly to SaaS, or at least public cloud.  Even this is really more prevalent in Singapore than in the truly emergent economies.

For home-grown enterprises in India  (as opposed to outposts of US, etc companies), cloud & SaaS is an area of  early investigation.  Even more than in China, TCO benefits of private cloud deployments are often undercut by generally low labour costs, which in turn hampers adoption of a lot of operations automation technology.

In comparison to the rest of the world, my general perception is that Asia's cloud-adoption maturity lags behind  that of the US (where adoption of the new can be much faster than elsewhere, even if it's not necessarily a good idea), and also behind the UK and similar western economies.  On the positive side, this means that Asia can take advantage of the mistakes made and lessons learned, and possibly  entirely leapfrog legacy IT concepts that western world has to continue to deal with for many years to come.

From an employment point of view, there is opportunity for people from the US, the UK & the rest of the world to bring their cloud expertise to Asia – most importantly it is business-integration-related skills and experience that are needed, rather than purely technical exposure. The biggest headache most organisations face when moving to cloud is how to integrate business needs with access to the right resources. This is a big change from the way we've been working with IT over the past 30 years (or more) and so needs different approaches & ideas. Experience from people successful in this area will be invaluable to organisations in Asia.

Wednesday, 22 April 2015

A Turning Point for OpenStack Cloud in the Enterprise

This past couple of days I've been at the CONNECT Expo in Melbourne, which included an OpenStack conference, where I participated as a speaker & panelist.

The focus of that conference was about whether OpenStack is ready for the enterprise, and included contributions from Dave Medbury from Time Warner Cable, Mike Dorman from GoDaddy, and Rik Harris from Telstra, who presented their real-world experiences with OpenStack in commercial settings.

One of the things that was very clear from the conference is that OpenStack is ready for the enterprise. This is probably not news to many who have been following the project for the past few years, but there definitely seems to have been a turning point in recent times, especially enterprise versions of the Juno release (such as SUSE OpenStack Cloud 5) that have become available in the past few months.

The turning point I've really noticed is that OpenStack is becoming much less of a developer's toy or interesting plaything, where new features are the most important aspect of development, and is transitioning (in the core areas at least) to a more practical enterprise-class framework, where component stability, infrastructure high-availability, and robust support options are necessary.  This was reflected in the composition of the audience for the event, which included many more "enterprise architect" types than have been present in the past.  It was also reflected in the questions asked during the panel sessions, which often seemed focused as much on business & organisation as on implementation and technicalities.

So OpenStack is definitely ready for the enterprise, yet as with any complex system and organisational change, there are a few considerations to bear in mind (and many of these are relevant, regardless of the private cloud infrastructure software to be used):

  • Implementing cloud computing is an organisational as well as operational change, and should be handled carefully
  • Successful cloud implementations rely on an existing IT operations mindset that includes automation & policy-driven deployment (see my previous article).
  • Executive-level sponsorship (ideally CIO-level or more) is required to effectively marshall all the disparate IT disciplines towards making cloud deployment a success
  • Integration with the lines of business is critical – internal customers of the private cloud have to agree to use the standardised cloud resources, rather than expecting bespoke IT services at cloud prices.
  • Not every workload is suitable for cloud deployment right now, as applications really need to be well suited to that kind of environment; this is OK – it's all about using the right tool for the job, so it's best not to try to force the issue and encounter failure.
  • The OpenStack control plane must provide High Availability (99.999%+ uptime). Without a highly-available control plane, access to the cloud resources disappears, and even cloud-aware services can fail.
  • OpenStack deployment and management can be difficult without the right tools and support, so it makes sense to work with an open source infrastructure software vendor (such as SUSE) to provide the technology, integration, support and training necessary to build up your own capabilities.

There is a growing number of enterprises (including PayPal and BMW) that have adopted OpenStack as a significant part of their IT infrastructure, and this trend will continue as the framework becomes more mature, and as the vendors and other members of the OpenStack Foundation build the body of knowledge in terms of documentation and training.

On that note – SUSE is hiring. At the time of writing there are at least 13 roles open at SUSE related to OpenStack & Cloud. Check https://suse.com/jobs for details.

Monday, 3 November 2014

Is Open Source Secure?


In the light of the recent Heartbleed and ShellShock vulnerabilities (both of which were apparently innocuous code bugs),  I was approached for comment for an article about the security of open source software.  The journalist (Anthony Caruana) wanted to know about how code gets approved for inclusion in open source projects, and especially about the potential for the inclusion of "malicious" code. This is a subject I touched on in a previous posting, so it was good to be able to address it more directly.

Here are the questions & my responses:

With such a huge community of developers contributed code to open source applications, what steps are in place to prevent malicious code being injected into widely distributed applications?

I take the term "malicious code"  to mean code that has been deliberately added for some nefarious purpose – to introduce a security back-door, or capture credentials and identifies, or do some other kind of damage. 
It's important to differentiate between this kind of deliberate attack, and the kind of programming errors that can result in exploitation, such as the openSSL Heartbleed bug.
The way that open source projects are managed actively works against the inclusion of malicious code. Firstly, the fact that the source is open and available to be audited by a large audience means that there is a much greater opportunity to find deliberate attacks than there is in closed source. It was source code audits that found the bugs in openSSL and Bash.  Malicious code would have to be carefully hidden - without leaving trace that it is being hidden – to avoid detection. Obfuscated code immediately attracts attention. 
The second factor is that major open source projects operate in a community where reputation is the critical factor:  simply submitting code is not necessarily enough to have it included into the main codeline.  The author must be known to, and trusted by,  the project maintainer.  It's human nature that when a new programmer joins a project, his or her code will be vetted more carefully than already-known contributors.  Establishing and maintaining a high reputation requires significant effort, so lends against the rapid or repeated insertion of malicious code. 
In contrast with this, closed-source code is not open for wide review by a large audience, and simply requires a disgruntled employee or the instruction of some agency to add the malicious code (as has been alleged by the Snowden leaks ).


What testing is done with open source software before it is released into the community? 

Testing varies from project to project - in this sense "open source" is not some homogenous group with identical policies. One of the features of the open source community is the rapid turn-around and release of projects:  with a major project there may be several different versions of software available at once - for example  a "stable" release, a "development" release, and a "nightly build".  A stable release can be considered to have gone through the most amount of testing; a development release is a version that developers and testers are working on; and a nightly build is version of the code that has the latest changes integrated & is almost certainly still buggy.  Of course, with commercial open source organisations like SUSE,  only stable versions are released.
Since this is open source it is the community itself that does the testing. The community may include commercial organisations like SUSE, but also includes researchers and hobbyists.  Users of open source software have to decide what level of comfort they have for possible bugs when they choose which version to use. One of the biggest contributions non-programmers can make is to try development release software & report back any problems.  Again - the potentially wider audience than you can achieve in a closed-source project means that this phase can be more effective and faster than in proprietary software (the "crowd source" effect). 
Hobbyists and non-programmers may only be able to investigate to a certain level, but commercial organisations and researchers can perform more involved tests. When SUSE contributes changes – either as new product features or patches – the software goes through our QA team to test functionality, stability, scalability and for regressions including vulnerabilities, as well as integration testing with our partners. SUSE makes heavy use of testing automation, and a lot of effort goes into maintaining & refining our automated methods. SUSE also relies on individual expertise to review results and find errors. 
At the end of the day for any software - open source, closed source or embedded - it's a question of confidence.  The attractive thing about open source is that there is potential for a much higher degree of confidence than for closed source.  It is much harder to hide mistakes, much harder to secretly introduce malicious code, and much more likely that a programmer wanting to make a name for him or herself will discover (and fix) problems when the source is open and available for review.  Ultimately, if the users of open source software want to perform rigorous and exhaustive examinations of code, they can;  the option is not even there for closed source software.