A google search of the term hyper convergence will bring you about 2 million results. There are thousands of tweets around hashtags.Industry witnessed a number of startups focused only on hyperverged systems.
In just about 3-5 years industry witnessed a number of start up OEMs growing exponentially to billion dollar enterprises . All the major hardware OEMs are selling at least one hyperconvrged product in the market. Clearly hyperconvrgence appears to be the future of data center hosting technologies.
And of course!!
So much has been spoken, written, blogged and tweeted about hyperconvergence already. So I promise not to repeat and reinvent the wheel. I shall instead be focusing on the technology adoption aspects. However, in this first section of a series of blogs about hyper convergence adoption we shall spend a few words to bring up to speed , those readers who has not yet had a chance or patience to look at some of the incredible writings about the technology of future Data center hosting
Hyper convergence essentially means:
- A handful of server blades and a few of Terra Bytes of hard disks capacity integrated in a small form factor rack mountable unit
- A Software defined abstraction that enables scale out of storage volumes carved out of these physical disks across several, sometimes hundreds of such small units.
- A virtualization clusters formed by server blades and the software defined storage volumes across these units
A smart management software that provides tight integration with the hypervisor and enables enhanced protection of workload and data across the entire cluster irrespective of their physical location
How does hyperconvergance make a difference?
Let us look at it with a small analogy of two freight trains travelling from a small industrial town to a port across the country
At the start of the journey, train just a carries few container cabins. As the train progresses its journey and passes via it’s stops more load – more container cabins are added, eventually has a long line of container cabins as it reaches the end point.
The Traditional Train
From starting point till the end The Traditional Train rides with a very large engine-cabin capable of pulling the entire load. So much of fuel is spent running a large engine for thousands of mile even when it is just transporting just a few tons of freight.
The Hyperconverged Train
Now let us look at a different option. A small medium powered engine is used to pull the containers at the starting point. It consumes much less energy and less pollution and so on.. As the journey progresses and more load is added to the train additional engine-cabins are added to the train just when it is needed
Hyper-convergence takes the converged systems to the next level by adopting a software centric architecture to combine compute, storage and network technologies in a single physical unit based on commodity hardware supplied and supported by one vendor. Each of these 2U rack unit system may contain one or more blade servers and auto tiered storage pools containing high performance SSD drives and high capacity disk drives. Additional capacity can be supplemented by adding new blades and disks or 2U units to the system. Adding new blade servers (nodes) add compute capacity to the hypervisor cluster and the storage capacity is added to the existing storage pool expanding the distributed file system with network level redundancy for data. Essentially the killer element that hyper-converged systems provide here is the intelligent distributed file system with network level redundancy also providing enterprise storage features such as automated Tiering and deduplication capabilities
Major OEMs of hyper-converged systems today are Nutanix, Simplivity(now part of HPE) and Scale IO. Most functionality provided by these technologies is very much alike; with a few differences in their internal architecture and philosophy. For instance, Scale IO uses its own hypervisor while Nutanix and Simplivity based systems use OEM hypervisors such as VMware ESX, Hyper V or KVM. Similarly, Scale IO and Nutanix has a fully software driven architecture while Simplivity relies on a hardware accelerator adapter for deduplication functionalities
Imperative benefits an enterprise can derive out of deploying hyper-convergence based data center are as follows:
- It allows starting with very small-lean kit and eventually scaling out to a multi petabyte system of thousands of virtual machines, as the requirement grows
- Software defined and yet simplified architecture allows a much faster time to market. Typically hyper-converged systems are deployed in a few hours while conventional discrete systems based infrastructure takes several weeks from delivery to deployment
- Simplified management of multi geo spread systems with a central intuitive console
- Better control of IS infrastructure budget with more predictable and granular increments of capacity additions
- Network based RAID or data redundancy enables Improved data availability
With all these benefits in place, any plan for introduction of hyper-converged systems cannot overlook the following risks and concerns
- Hyper-convergence is a relatively new technology in the market. Introduction of any new technology in to an enterprise estate without proper planning, service distribution and service introduction measures can lead to a calamitous end state
- Vendors and OEMs supplying hyper converged are younger organizations which are still growing to a larger global presence and support organisms. Deployment of hyper-converged systems without a proper support mechanism and commitment from the chosen supplier will be only adding on to the existing challenges of DC management
However, Hyper-convergence has have proven its value as a technology of choice for most of the technology giants such as google, AWS etc. who have adopted it for a while now. With a spectrum of choices of technology providers and gradually increasing adoption from enterprises, very clearly hyper-convergence is the data center technology of tomorrow. An earlier adoption of this technology means a smoother transition to hyper-converged DCs in future for an enterprise