Converged Infrastructure: Talk by Bala Ganeshan, CTO QLogic
This is a live-blog of an event organized by Software Exporters Association of Pune (SEAP), in which Dr. Bala Ganeshan, Vice President and CTO of QLogic Corporation talked about Converged Infrastructure, which has also been called as Integrated Systems, or Unified Computing Systems. Specifically, he focused on the idea of having compute, storage and networking into one preconfigured package which can be delivered to the data center. This approach provides a major transition that complements the shift to cloud and virtualized computing. These trends together with new I/O centric applications and solid state storage have the potential to revolutionize computing. Many IT organizations have started using converged infrastructure for various reasons ranging from centralized management of IT resources to consolidation to lowering costs. Gartner estimated the growth by more than 50% in 2014 over 2013 to reach $6 billion. Another research firm IDC estimates that, total worldwide spending on converged infrastructure will hit $17.8 billion in 2016.
The evolution of IT infrastructure for the data center
- Initial system implementations for “open” infrastructure were inflexible, had CPUs with limited power, networking that was slow and unreliable, and storage was captive – it wasn’t easy to share data.
- SANs (Storage Area Networks) evolved to address these limitations. Here, the data and storage was divorced from the servers, and powerful and flexible standalone storage systems were set up that could be accessed by any servers from anywhere else on the network. But SANs are complex and expensive, required specialised skills, relative high effort to provision new infrastructure, and ended up requiring lots of efforts to manage operationally in large environments. Over time, interoperability between different pieces of the SAN became more and more difficult.
- Converged Infrastructure was a response to the problems with SAN. The focus of the new systems was ease of use, faster time to deployment, and ease of management.
Factors that drove the disruption of SANs and IT infrastructure:
- Web Scale: Suddenly, the requirements on the IT infrastructure grew by orders of magnitude, requiring radically different solutions and architectures
- Open Source: Putting pressure on proprietary solutions
- Software Defined Networking in particular, and in general Software Defined Everything
- Storage Scale-out: more storage generated in the last 2 years than the history of computing before that.
- Server vendors started incorporating storage in their offerings
- Storage vendors started building intelligence and apps in their hardware
- No end user specialized skills required. No zoning, no LUN masking.
- Automated provisioning: use of single-click, drag-n-drop interfaces for creating new storage and associating it with servers and applications
- Tightly integrated management
- Focused on application deployment: the business doesn’t really care about disks and LUNs and servers. They care about their sales orders, and websites, and SAP instances. So all management and deployment should be in terms of the applications, and the details of the underlying infrastructure should be hidden, and automatically handled by the system
- Combines compute, network, and storage resources
- Improves time to deploy new applications
Not all of these goals are easy to achieve, but any progress on any of these parameters is a huge improvement.
There are three approaches to converged infrastructure.
- SAN based infrastructure
- Where servers and storage an infrastructure unit contains a bunch of servers and the storage is hidden under the servers, and the servers expose virtualized storage
- This gives a flexible architecture. Servers and storage can be expanded independently
- Reduction in OpEx
- Management is convoluted, and is at the level of the virtualized storage
- Here hardware appliances contain both storage as well as compute nodes
- The convergence across appliances is provided via a software management layer
- Provides an Application Centric view to the users
- Fixed granularity of scale – need to add appliances; can’t scale the storage without scaling the compute, and vice versa
- Open Source Technologies:
- The kinds of technology being built/used by Facebook/Google
- This is currently a small fraction of the worldwide infrastructure
A lot of small startups are playing in the “Appliance” space, because the ease of management is an important consideration in the SMB market. Larger, more established companies are doing SAN based converged infrastructure.
Here is a random collection of interesting points made during the discussion, captured in no particular order:
- Although the impact of the cloud has been growing, enterprises have still been slow to adopt the cloud. Security is one of the problems, but also the need to move lots of data across the network is a bottleneck. So, private clouds are getting good traction; and Microsoft is one of the companies that is responding quite well to this demand.
- For now, applications like SAP will continue to be in-house and will not easily move to the cloud. This is notwithstanding the fact that SAP does run on the cloud – it’s just not an ideal setup. Initially, the applications that are moving to the cloud are things like websites, mobile apps, and other low-hanging fruit.
There was a panel discussion on converged infrastructure. The panelists were:
- Aalop Shah: Technical Director at Druva; handling the laptop backup product.
- Maneesh Bhandari: Angel Investor.
- Nitin Deshpande: President, Allscripts India.
- Viswanthan K: Vice President, Corporate IT, APAC, at Eaton Corp
- Yogesh Zope: Group CIO for Bharat Forge Limited, and CEO of Kalyani Technologies.
Here are some of the comments made by the panelists:
- Yogesh: For years, different companies have called this concept by different names. And yet, it is still not something that is really useful for a company like Kalyani. Converged Infrastructure and all this flexibility might be useful for a company that needs to handle big data and lots of unpredictability as far as the storage and compute requirements are concerned, but that is not really an issue for many companies. So we’re happy with the old style infrastructure. Another important point is that the whole issue of “manageability” is more of a consideration in the west. In India, talent is much cheaper, and we don’t have the same problems.
- Bala: The needs and the trade-offs in the Indian market might be different. While some of the problems being solved by the current players in the converged infrastructure might not be pain-point in India, there are other problems that are faced here. So while the current players are not really addressing those requirements, at some point they’ll start addressing India and then they might make inroads in the market.
- Vish: One of the aspects of this that has not yet been addressed is the power requirements. The data-center requires a lot of power, and the requirements of a converged infrastructure data-center are unique. We (i.e. Eaton) are now playing in this space, and providing “convergence” not just at the level of the software/hardware infrastructure, but at the level of the entire physical data-center, including the power.