STORAGE

  • 07/28/2015
    7:00 AM
  • Rating: 
    0 votes
    +
    Vote up!
    -
    Vote down!

Hyperconvergence: Redefining The Data Center

As optimized platforms for the virtualization era, hyperconverged systems are poised to radically transform the data center.

Looking at the past 20 years, virtualization stands out as one of the true infrastructure revolutions. It brought speed, efficiency and lower cost to on-premise hardware, then fueled the success of cloud computing which, in turn, is the great enabler for mobility today. P2V or physical-to-virtual has been the de-facto slogan of IT data center success.

Now the revolution is unfolding as to where the virtual machines reside. Today’s need for data-centric businesses with distributed architectures that are both flexible and scalable has given rise to hyperconvergence. Hyperconvergence coalesces storage, compute and network hardware into appliances (typically hardware) designed for better virtual machine performance along with simple management. It is a re-thinking of the data center as virtualization-enabling/empowering building blocks with inherent scalability. Hyperconvergence is an optimized platform for the VM world. 

This is very different from today’s typical ad-hoc adds of servers and SANs that result in complexity and difficult management. Just setting up a flexible SAN for shared access by high-availability systems is usually a bear, and there is a tendency to over-spend for redundancy. Hyperconvergence promises dramatically lower data center costs because of component pre-integration; appliances tend to be almost half the cost of cobbled-together systems. Operational costs are also low with significant power savings. Easier procurement, deployment and system management frees the IT staff for other projects.

Hyperconvergence is gaining traction. IDC in April of this year put 2014 growth at 162.3% with a market value of $373 million. For 2015, this is predicted to more than double to $807 million globally. There are now more than two dozen hyperconvergence vendors.  Clearly, the data center is again hot for redefinition. 

Nutanix is one of the leaders in the hyperconvergence space. The startup offers tight integration of compute, storage, networking, and virtualization resources and claims 40% to 60% savings in hardware costs with up to 8X faster time to value on deployments. It also claims power reductions of 90%. Nutanix’s 2U appliances promise to make the infrastructure metaphorically “invisible.”

SimpliVity, recently described by Forbes as a startup “unicorn” with an expected valuation of over $1 billion, offers a hyperconverged product that combines all IT services and components below the hypervisor into an x86 building block infrastructure appliance. No need to worry about discrete devices such as servers and storage; everything is in SimpliVity’s “Omnicube” system: 8 to 24 servers with 8 to 10 terabytes of usable storage in a compact, scalable, geographically dispersible system. The software-centric design is for high-availability and performance of VMs. SimpliVity claims Infrastructure cost savings of about 40% and a 300% reduction in total cost of ownership (TCO).

Nutanix, SimpliVity and other hyperconvergence companies are viewed as a threat by VMware, which has come up with its own EVO:RAIL hardware hyperconvergence product. This is a virtual SAN or VSAN approach where the VSAN is embedded into the vSphere compute cluster. Competitors claim that this approach limits flexibility and scalability, but VMware aficionados argue that the product is ideal for those starting out with virtualization, that it has low IT human resource needs, and deployment is possible in minutes. VMware Virtual SAN Ready Nodes are offered to those with greater needs and a DIY bent.

For IT departments, choosing between vendors is a matter of understanding their virtual workload needs. SimpliVity touts cloud replication as an option; a VM image running in Amazon Web Services (AWS) can be linked to an on-premise Omnicube. Meanwhile, Nutanix offers “data locality” where when you vMotion a VM to a new node on a cluster, the data store associated with the VM is pushed to that node also (this is done in the background) for better data performance. For some, VMware’s 100-VM EVO:RAIL infrastructure box built around vSphere may be the easy choice if compatibility is a concern.

While power reduction and easy scalability and management will appeal to established enterprise VM practitioners, hyperconvergence should be equally appealing at the low end to small companies. Startups creating data centers can worry less about setting up SANs and high availability servers and opt for a do-it-all appliance. Expansion as needs evolve is easy with these rack-mounted, slim-profile hyperconverged systems.

IT departments everywhere should take heed of hyperconvergence: Simple, low-cost hardware is a compelling value proposition. At the same time, the technology is still emerging and one should be mindful of critics. Some argue that hyperconvergence is rigid and presumptuous in its assumptions of scaling by adding appliances. VMs have varying appetites as to CPU hunger or IO needs and it is rare that physical memory, CPU, storage and networking needs will run out or be needed at the same time. For now, the early adopters of hyperconvergence are singing its praises.


Comments

hyperconvergence

Hi Azmi -- Are there particular use cases where you see hyperconvergence playing a big role? Or areas where it's not such a great fit? I've read that it might not be the best fit for big data analytics.

Re: hyperconvergence

Hi Marcia -- For any virtualization environment where there are separate servers and storage in a typical 3-tier infrastructure deployment, there is something to be gained in going for the simplicity of a single, optimized, easily deployed and easily managed appliance. Some of the appealing use cases are desktop virtualization, VM-centric improved backup and DRP, easily spun-up development and test environments, infrastructure upgrades for modularity and scalability and easier remote-site management. Cost and power usage are also lower. Regarding big data analytics, if your big data application is virtualized, Nutanix claims 3X performance and scalability for applications like Splunk and Hadoop to 1TB per days ingest rates.  Deployment times are claimed to be just a few hours. Fast search and indexing for up to 500,000 events/second is claimed. Of course, many companies like Oracle, EMC and Teradata now offer Hadoop appliances. Application specifics will dictate what is best for a particular deployment.

Re: hyperconvergence

Thanks for those use cases and additional guidance Azmi. In choosing a vendor, perhaps another factor for enteprises to consider would be that many hyperconvergence vendors are startups, which could be acquired.

Re: hyperconvergence

I always thought what if after creating hyperconverged infrastructure, i want to de-integrate some of its componenets.

Re: hyperconvergence

No worries, virsing211.  You can change your mind.  Simplivity, for example, stresses that hyperconvergence is part and parcel of the software defined data center.   Automated policies around  VMs and their data gives flexibility – and moving data around along with VMs makes for an agile environment.

Re: Hyperconvergence: Redefining The Data Center

As usual, I like your approach, Azmi; not too much focus in the technical minutae of the hardware, but certainly not too little, and a healthy dose of the vendor climate and economics to guide us in what is, ultimately, a business decision. In particular, I've heard many great things about Nutanix; when you're relying on one box from one vendor to do so much, the support you get from them counts for a lot. The minute it's costing you time instead of saving you it, the upfront cost no longer matters, and I've heard from a couple people that Nutanix is really good in this regard. Likewise with ease of installation and scalability - cutting down on manhours spent around the server racks is the reason you went hyperconverged, isn't it?

Re: Hyperconvergence: Redefining The Data Center

Thank you, zerox203.  Yes, its savings on manpower, scalability, flexibility and lower power consumption that you gain with hyperconvergence – a simplification of data center structure and operations.  I think all the vendors are focused on great support.  This is a new area with huge potential; all the hand-holding helps.