I have always been a big fan of the US space program and NASA as an organisation, especially for the research that they’ve contributed to. Over and above my curiosity for space travel, rocket ships, and things out in far flung places of the universe, I really admire NASA’s research and contributions to engineering, and to the inventions that have become spinoffs usable in every day life. Many of the things NASA originally developed for space, and through their space research have directly benefited us on planet earth, from medical devices, through to car tyres, to clothing for emergency services, to cordless vacuums for the home. You should check out the NASA Tech Benefits page for a list of some of these things. I can hear you asking “What does this have to do with the topic of this article?”

The answer is simple. Just as with NASA’s research and incubation of inventions that have come to benefit our lives, Google, Facebook, Microsoft and other large Web Scale companies have researched, developed and incubated technologies that can be made directly applicable to traditional enterprise IT. All that was needed was a visionary and talented organisation to package all of the inventions and lessons learned by the Web Scale companies into a bundle that was easy to consume by any sized organisation, and would meet the diverse requirements of enterprise IT.

We needed an organisation that would deliver the “always on”, “linearly scalable”, “elastic”, “pay as you grow” experience of web scale companies, at any scale, and could support a fully integrated system from end to end. An organisation that was singularly focused on removing the massive amount of unnecessary complexity from enterprise IT, on redefining customer experience, and  taking complex IT and making it very simple and easy to use. This would break down silos of skills within IT and allow organisations of all sizes to concentrate more on business value creation, rather than managing infrastructure. Once we had that, we needed to take the technology to the masses, take what had been incubated and provide a few more simple options, while maintaining user experience. So lets examine briefly how and why Web Scale IT is relevant to you, and take a look at the organisations that will bring it to you, or a datacenter near you.

Nutanix Software and Vision Meets Dell Hardware and Global Sales and Distribution Network

That visionary organisation I’ve mentioned, the most visionary according to the Gartner Magic Quadrant of Integrated Systems, was Nutanix (where I work), and now we’ve teamed up with Dell to massively increase their global market penetration of the Nutanix Virtual Computing Platform and to bring more choice and the benefits of Web Scale IT to organisations of all shapes and sizes. At the same time launching growth to the stars, or the outer rim of a far of galaxy. Google, Facebook, Microsoft and other massive Web Scale companies have researched and developed technologies over a decade, where they have incubated the technologies and made them much more robust. At Nutanix we have taken some of the technologies used on those massive web scale companies, the methodologies, the lessons learned, and packaged them up in a way that is simple and easy for anyone to use, to pay as you grow, no matter what size your organisation is.

The alliance and OEM agreement that has just been announced with Dell will see Nutanix Software matched up with new Dell XC Converged Appliance Hardware. The appliances will be available directly from Dell and through Dell’s channel partners, starting from the end of calendar 2014. This provides Dell a proven and robust software defined storage platform to add to their existing storage product portfolio, and an entry into the integrated or converged systems market, which was up till now missing. While for Nutanix this gives access to Dell’s massive global sales and distribution channels. Dell described this alliance as “An exciting industry transformation drive by our customers”.

I like the way Dheeraj Pandey, Nutanix co-founder and CEO summed up the announcement. “This agreement with Dell highlights our shared commitment to disrupt traditional datacenter architectures and accelerate the adoption of web-scale infrastructure in the enterprise. Dell is the right partner to leverage our software-driven technology, and introduce powerful converged infrastructure appliances with a wide range of price and performance options geared toward supporting a variety of critical enterprise workloads.” He also explains the importance of this alliance in his article SDS for the Masses: Why Dell Matters.

This really proves that the Nutanix Virtual Computing Platform is a software defined solution, and was never technically tied only to the Nutanix appliance hardware (standard white box x86 from SuperMicro). The hardware was merely a convenient and simple package for delivering the software and making it easy for customers to purchase, and easy for Nutanix to support. Now with the Dell alliance customers will have more choice, but importantly while keeping the same high quality user experience and simplicity that they have become used to. Importantly for existing Nutanix Partners Dell will be bound by the same deal registration system, putting them on equal footing, so that it doesn’t negatively impact the existing loyal and capable channel that Nutanix has already built.

Legacy Architectures Ultimate Decline

Collaboration between the research and engineering teams at both Nutanix and Dell will I’m sure bring to market some very interesting offerings to customers in the future, and expand the existing target use cases into different areas. I’m excited by the possibilities I’m already seeing specifically for Business Critical Apps and Monster VM’s from my existing work, and also from this new alliance. All of which will shake the status quo (which is already quaking in its boots and spreading FUD to try and discredit this shift in technology).

Legacy compute and storage architectures will have to fight to remain relevant. Too often they are unnecessarily and unjustifiably complex for the majority of common enterprise IT workloads. There is no reason it has to be so hard or complex for the majority of workloads, or the majority of use cases (90%). While they might be ok at large scale, they can’t cost effectively scale down, or be operated with the same benefits by smaller organisations. I know this first hand because I have helped architect and implement many successful large scale enterprise IT solutions across the globe, that at the time were very complex, for good reason, and with justification, but times have changed. I’ve deployed hundreds of legacy Unix systems in the past, when they were the best option to meet customer requirements, and also migrated many applications off Unix systems and Mainframes more recently to virtualized platforms. I know of dozens of projects I could have architected differently, if the platform had existed at the time, to better meet customer requirements. Legacy compute and storage architectures have already started a slow decline, which I anticipate will be similar to what we’ve seen for Unix systems and Mainframes.

The legacy will still be around for years to come, used primarily for the workloads that don’t fit well yet with Web Scale IT architectures. It’s part of my job at Nutanix to help develop the Web Scale IT architectures for the future to fit the Monster VM’s and Business Critical Apps, and attack some of the areas that don’t quite fit yet. While at the same time helping customers and prospects understand why the 90%+ of workloads fit just fine right now, and how best to benefit from the next evolution in their software defined datacenters and hybrid cloud solutions.

Just because the legacy will be with us for a long time shouldn’t stop you from adopting new technology that will significantly simplify your environments when there is a fit that meets your requirements. By migrating to a new architecture you should be able to get the best of both worlds. Getting the most out of your legacy investments, while at the same time preparing for and taking advantage of the next wave of innovation (which has already been with us in some form for 10 years, but only accessible to hyper scale web scale companies).

What is this Web Scale IT thing you speak of?

Web Scale IT. It’s Here. It’s Now. So is this just marketecture (something that looks good on a marketing slide but can’t be implemented the same in the real world) and another industry buzz word created to sell a solution looking for a problem? It might have been if there were no substance to it, and if there wasn’t a fundamental problem with the way IT is done in enterprises and smaller businesses. It is broken, it is way more complicated than it needs to be, and it’s time that it changed. Why spend so much time feeding and watering IT infrastructure when you could be much more focused on business value creation and your customers. Wasn’t the promise of virtualization and cloud that you would spend less time and money on operations and more time on innovation, customers and value creation? Well the problem is that although we virtualized things (and had less overall to manage), we didn’t change how IT was done, so we ended up with the same problem, and some new ones at the same time. We had less to manage sure, but the proportion focused on maintenance and operations stayed the same (with declining budgets to boot), because the way we architected the systems didn’t change. Traditional converged systems (often just a converged PO) go some way forward, other hyper converged systems go a bit further, but you need to take the principles of web scale before you can really change the game.

Just like consumerization of IT, mobility, and containers are having a profound impact on the way that applications architected, developed, accessed and business value is created, Web Scale IT will have a profound impact on the way enterprise and cloud provider datacenters are built and run, and how smaller business purchase and consume IT. Web Scale IT is a way of describing a group of capabilities that allow normal businesses access to some of the same benefits that hyper scale web companies have enjoyed for over a decade. It’s more than just a single technology, and it’s more than just a single implementation of any technology. Web Scale IT is a system and methodology for designing, deploying and managing infrastructure at any scale. It describes a solution that takes technologies that were incubated and made robust by the hyper scale, web scale companies (and the consumers using them), and packages it up in a way that is suitable for diverse application requirements, and can fit any scale.

In a nutshell Web Scale IT infrastructure can be defined by the following properties:

1. Out of the box scalability, ability to start small (suitable even for SME businesses), pay as you grow without disruption, linear performance
2. Hyper-convergence on x86 hardware, integrated compute and storage, linearly scalable networking based on standard ethernet / Internet protocols (Ethernet Always Wins) – PowerPC and SPARC don’t get a look in, highly proprietary hardware architectures are not a fit in this new architecture paradigm
3. All intelligence in software – 100% software defined, don’t get software defined confused with hardware appliance packaging, software still needs hardware to run, the packaging allows for better QA, better support, and simpler purchasing and consumption
4. Distributed shared nothing architecture with no single point of failure – cluster wide data and services
5. Self-healing system, always on, non-disruptively upgradable, fault isolation and distributed recovery, reduce impact of failure, predictably recovery from component failure
6. API Driven Automation and Rich Analytics – Operations that would traditionally have been done manually are automated and metrics and analytics can be easily visualized and it can all be exposed via API’s, this is partly how these systems are able to scale and how the productivity of IT personnel can be greatly improved
7. Run multiple applications with diverse requirements simultaneously

Item 7 is the primary divergence between Web Scale IT infrastructure and the Web Scale or hyper scale companies that pioneered the technologies, usually the Web Scale and hyper scale companies only had one application, but enterprise IT has many. Any successful Web Scale IT solution must be able to support diverse application requirements, while taking the design patterns and learnings from the pioneers.

Some of the additional benefits that aren’t strictly part of the Web Scale IT definition but are critically important IMHO:

The ability to have a known quantity of capacity and performance per unit of growth. This gives you predictability of growth, so you know how much performance or users or whatever metric when you scale. You can incrementally grow and shrink your environment non-disruptively as business needs change. Having a known quantity per unit of growth also makes the whole architecture design process dead simple, less costly, and far less risky. If you make a small miscalculation then the consequences are far less than if you did the same with a legacy architecture, or converged (PO) system.

Commoditization and standardization are key aspects of scalability and they are also key aspects that allow the costs to be kept to a minimum, while maximizing value. Using standards based x86 architecture, and being software defined, allows you to achieve much lower TCO, with lower CAPEX and OPEX expenses. Your expenses start small and grow with demand, rather than purchasing everything up front and having it mostly idle for a few years until demand picks up. I hear all the Cloud and Managed Services Providers out there wanting this, especially as it allows their expenses (capital and operational) to match their revenue and end user demands. It’s so important to have this ability when demand is unpredictable.

Respond faster to business needs (goes back to more time on value creation and less time on feeding and watering infratructure). From purchase to running VM’s in less than 14 days with Nutanix. This significantly simplifies capacity planning, as does the ability to non-disruptively grow the environment at the click of a button and have the capacity and performance immediately available.

Take advantage of the latest in performance and technology when it’s available.  Moore’s law is still in action and technology is advancing in performance and capacity and doubling about every 18 months. Why would you want to wait 3 to 5 years to take advantage of the latest advancements? With a Web Scale IT approach and with the Nutanix platforms you can gain the advantages of the latest technology when it becomes available, without waiting, and without disruption to your existing environment. New components can slot in and live quite happily next to the older components, and because the platform is virtualized it’s just a matter of live migrating your systems over to the new nodes to immediately take advantage of the performance improvements, the capacity is already available to the rest of the environment seamlessly. This means you completely and forever avoid any need to do a forklift upgrade, to either a new platform because of age, or because you’ve run out of capacity or performance. You just add nodes and scale out the platform.

Choice of hypervisor. Although there is really only one dominant hypervisor today (VMware vSphere), and it’s because it’s been the best option that fit the most requirements, doesn’t mean that it will always be the case. The worst thing that could happen is if there was no competition and the dominant player become complacent and stopped innovating. The reality is that customers want choice. That is why Nutanix supports three hypervisors today, VMware vSphere, Microsoft Hyper-V and Centos KVM, and may add more in future based on demand. Even if you don’t want to make a change today, having the ability to do it in the future is valuable. It will keep everyone honest and keep everyone innovating, hopefully all delivering ever more value to the customers.

Web Scale IT Sounds Great, What Doesn’t Fit? 

Web Scale IT systems, and the Nutanix Virtual Computing platform, are built to scale out. The unit of scale can be quite large, but there will be some applications or workloads that don’t fit within the performance capability of a single node or the single unit of scale. 90% of workloads work fine out of the box and there is nothing more to do. 5% will work great with some minor tuning, and another 5% aren’t yet candidates. This will change over time as technology advances.

The types of workloads that aren’t candidates are anything that needs more performance than can be delivered, such as a single very large database that needs many tens of thousands of IO’s and many GB/s of throughput, that can’t be broken down into smaller units. I recently had a customer asking about virtualizing a single Oracle Database that was 45TB in size and routinely did between 41,000 IOPS and 174000 IOPS and 1300MB/s – 5500MB/s throughput. This was a single database system that couldn’t be broken down into a smaller unit and is an example of something that is not a good candidate at the moment for the Nutanix platform. Another customer asked about migrating an Oracle RAC 10G database (Only Oracle RAC 11g R2 11.2.0.2 and above is supported by Oracle and Nutanix) that was running a critical (and unsupported by the app vendor) application, my advice was not to do it, it’s not a good candidate. Another customer wanted to virtualize four or five 6TB – 12TB databases on a single host for licensing reasons, the combined workload profile of these databases and the constraints around licensing meant it wasn’t a good fit, but it was very close and this year we’ll likely have a solution that would support this type of scenario.

Databases or applications that do up to ~4K IOPS each, and have memory and CPU requirements that fit into the appliance resource capabilities though work perfectly well, and by adding more workloads and more nodes to the system you can scale out performance and capacity linearly. A 4 node Nutanix system (in 2RU) can achieve up to 100K random 4KB read IOPS and 70K random 4KB write IOPS (NOS 4.0 and above), performance varies by IO size, randomness and read write mix. You can mix and match nodes to fit the proportions of compute (CPU and RAM) and storage (Capacity and Performance) that you require, so that you can achieve an efficient balance of resources.

The Nutanix platforms available today are perfect for 90-95% of workloads that run in business environments. Java systems, app servers, web servers, mail servers, small to medium sized database servers (up to few TB of data, 16 – 20 CPU’s, 256GB RAM etc), distributed big data systems, VDI, dev test workloads, the list goes on. There is an easier way to cater to the requirements of these systems and it is available today, and it’s available from Nutanix and by implementing a Web Scale IT infrastructure. For the 5% of systems that don’t fit, the monstrous database servers, and any single instance workload that doesn’t fit the performance profile, they can get better performance from your existing investments in your traditional architecture if you migrate off workloads that are a good fit for Web Scale IT.

You can keep your existing investments in traditional IT architecture for the application that don’t fit, until you’re ready to retire them. This is what happened to Mainframes, this is what happened to Unix systems. They are still around, but they are only handling a very small percentage of enterprise IT workloads and only for specialised and niche use cases. Meeting use cases and requirements that are becoming fewer and fewer all the time with the growth in capability and reliability of standard x86 architecture.

Final Word

Web Scale IT isn’t a fad, it isn’t pure marketecture, it is real, it is here, thanks to Nutanix (and the web scale / hyper scale pioneers), and it is applicable to business of all sizes and for diverse workload requirements. It’s not just one technology implementation, but a set of defined capabilities and ways of looking at an overall IT system. Nutanix has taken the Web Scale technologies, methodologies, lessons learned and benefits from the pioneers and distilled them down and formed them into a package that benefit any organisation.  This breaks down silos and removes the necessity for specialised skills just to manage a platform (there will be much more value in being a generalist in the future). This directly upsets the status quo, which has created the unnecessary complexity that we have today, but it is the future. You can find out more about Web Scale IT by viewing the Web Scale Wednesday On Demand webinar, which includes leading web scale companies such as Twitter, former Facebook engineer, Dell, Nutanix and Citrix.

All of this technology stuff and talk of Web Scale IT is all well and good, but it’s not why I decided to get involved with Nutanix. I saw the writing on the wall for the traditional enterprise IT infrastructure architectures (it won’t happen over night, but it will happen eventually), just like I did with traditional Unix systems before that. I was easily able to architect fantastic solutions for large scale, but became frustrated as it often wouldn’t be useful at smaller scale or cost effective. Smaller businesses were locked out of all the benefits that large business could take advantage of. I was also frustrated by what I saw as unnecessary and unjustified complexity, I thought there must be an easier way. Lastly I was also becoming frustrated by what I perceived is a lack of customer focus, I always try to put customers requirements first, regardless of who I work for or what technology is being used (there is no one size fits all). I’ve found in Nutanix a company that is focused on delivering great customer experiences in every aspect of the technology and business, and a company that has taken incredible complexity and made it consumer simple enough for anyone to use. Now my job is to take this and make business critical applications and monster VM’s more simple. Delivering on the promise of virtualization, cloud computing and software defined everything, no matter if the customer choses private, public, or a hybrid deployment solution.

A lot of the problems with enterprise IT and the big vendors that help create the complexity (and sometimes mess) are very well described in Robert X. Cringely’s book – The Decline and Fall of IBM.  I see parallels and lessons in this applicable to many other organisations and signs of what to look out for and avoid. It’s worth reading.

Your feedback and comments are always appreciated.

—

This post first appeared on the Long White Virtual Clouds blog at longwhiteclouds.com. By Michael Webster +. Copyright © 2012 – 2014 – IT Solutions 2000 Ltd and Michael Webster +. All rights reserved. Not to be reproduced for commercial purposes without written permission.