November 18, 2011 | Written by: Stephen Viselli
Share this post:
After reading Alfredo Olivieri’s great post about IBM’s cloud in Rome – “Does IBM use cloud?” – I felt that I needed to write a cloud article on behalf of Australia. After all, these Italians aren’t the only people who can make a great cloud! In fact, here at IBM’s Gold Coast development laboratory, we IBMers are spoiled when it comes to clouds. Not only do we have our own private cloud exclusively for use by local lab employees, but we also have access to the Australia wide IBM cloud, part of which is hosted on the Gold Coast also.
First, let’s talk about the Gold Coast cloud. Being a development lab, we’ve always had plenty of test and development virtual machine servers in our server rooms (but never enough). However, each individual team within the lab used to own and maintain its own collection of servers, and web front ends and scripts for managing those servers. The Gold Coast team decided it was time to turn those disparate servers into a cloud – uniting them with a single management interface and a single image catalog, and sharing resources between teams!
And, the result looks like the following figure:
The lab actually hired a university student as an intern to develop the management interface. So we have a custom web interface that calls the VMWare ESXi command line tools on the back end to execute management tasks. He did such a good job on the project that he was offered a full time position before he even graduated!
While we were busy building our own private cloud, the IBM Australian Development Laboratory (ADL) organization (of which the Gold Coast is a part) decided it wanted to build a cloud too – a cloud that would service all of IBM’s development, testing and support staff in Australia. So they created the ADL cloud – a geographically dispersed cloud with hardware initially located in three sites across Australia, and with users initially from seven IBM Australia offices.
The ADL cloud is built on the IBM CloudBurst offering – a combination of the necessary hardware and software designed for rapid cloud deployment. Basically a huge server rack, a lot of IBM software, and a chosen hypervisor, all working together in perfect harmony. Obviously being a geographically dispersed cloud with hardware in multiple locations and users from all over the country, there were many challenges involved in implementing it. These included:
- Managing user authentication and authorization: The team achieved integration with IBM’s existing employee directory to achieve seamless authentication and authorization to the web client and desktop client either on site or remotely.
- Isolated test networks: One of the biggest use cases for cloud computing is flexible and isolated test environments, however providing remote access to these networks while maintaining compliance with company security policies can be very difficult.
- Syncing data between sites: Australia doesn’t have the fastest network infrastructure in the world. Syncing massive libraries of VM templates and software repositories can take a long time over the wire. So far, the best solution has been to courier hard drives between sites – a little bit of thinking outside the cloud.
- Resourcing: The team charged with building the ADL cloud is not dedicated to the cloud, it is expected to continue business as usual. These guys are building and administering a geographically dispersed cloud, and still have time to do their day jobs. This is a testament not only to how smart they are, but also to the cost benefits of cloud computing.
And here is what we cloud users see – another place to create more virtual machines!
The benefits of this entire cloud infrastructure are endless. As developers and testers using the cloud, we have more resources than we could even imagine using, easy access to a library of reusable images and software bundles, a nice user interface for managing our virtual machines, and access to our test and development machines from the office and from home.
The cloud system administrators aren’t left out either – they get a single administration interface for managing users, hardware and software. They also have well integrated monitoring capabilities so they can report on the utilization, performance, and efficiency of the cloud.
Even management loves the cloud – the optimization of hardware and virtualization capabilities result in reduced future capital expenditure, and also reduce operational costs at each lab.
So what’s the downside to all this cloud computing? Well, whenever I ask my manager for more memory for my laptop, he tells me to use the cloud instead!