Hybrid Cloud Platforms: Herding Cats and Dogs Together – Efficiently and Securely
SEE LAST PAGE OF THIS REPORT Paul Sagawa / Tejas Raut Dessai
FOR IMPORTANT DISCLOSURES 203.901.1633 /.1634
psagawa@ / firstname.lastname@example.org
February 21, 2019
Hybrid Cloud Platforms: Herding Cats and Dogs Together – Efficiently and Securely
The cloud era brings enormous potential for enterprise computing – much lower costs, access to cutting edge technologies (i.e. AI, blockchain, etc.), greater flexibility/agility, better service to users, etc. However, with those opportunities comes complexity (e.g. hybrid cloud, multi-cloud, microservices, containerization, etc.). Vendors have begun to develop platform solutions to help enterprises with these challenges, addressing a market that could top $400B annually and, importantly, increase the customer buy-in for IaaS and SaaS services up and down the stack. Competition will come from three main directions: 1. Cloud hosts (AMZN, MSFT, GOOGL) adding higher level services and multi-cloud/hybrid cloud support; 2. Traditional IT vendors (IBM, VMW, CSCO) offering infrastructure-independent integrated solutions; 3. “Tools” vendors with platform software elements (NTNX, PVTL, Docker, MDB, etc.) allowing enterprises to build their own platform layer. We expect customer platform choices to vary by their degree of heterogeneity, their reliance on a primary vendor, and their pace of change, dictated by their own IT capabilities, their concern for vendor lock-in, and the barriers to change for their existing systems. The top cloud hosts should see substantial growth and profitability from platform services, with MSFT and GOOGL gaining on AMZN, and all three looking to M&A and partnerships to beef up their offerings. IBM and VMW look ahead of the other traditional IT vendors on the platform layers, with IBM’s RHT deal a key move. We like NTNX and PVTL amongst the focused players, with both possible deal targets.
- Enterprises face opportunity and challenges in the cloud. Cloud hosting on hyperscale datacenters offers major advantages – e.g. 50%+ reduction in all-in costs, flexibility (capacity when/where needed), agility (respond to demand in real time), specialized resources (AI acceleration, specialized APIs, development support, etc.), regular updates, world-class security, etc. However, the transition will be challenging and expensive, as enterprises cope with complexity from hybrid public/private cloud, multi-cloud architectures utilizing new concepts like microservices and containerization to manage workflows.
- Platform solutions are a $400B market. To manage this complexity, enterprises are turning to software platforms that create a layer of abstraction between applications, data and infrastructure, such that performance can be optimized at minimal cost. The platform layer includes traditional “infrastructure” software like OS and DBMS, but must include newer functions like containerization, data streaming, AI acceleration, API management and others which enable enterprises to gain the greatest advantage from the cloud. Solutions for this abstraction layer are loosely termed “Hybrid Integration Platforms” and enable efficient use of resources across multiple clouds and private datacenters. All-in, we think cloud platform software and services addresses $400B in current IT spending with considerable room for years of rapid growth.
- IaaS hosts are moving up the stack to the platform. One set of solutions for the platform layer comes from the major hyperscale hosts. These companies emphasize a “cloud native” approach that adapts enterprise apps to the cloud architecture, while providing connectivity via APIs. However, the three main players have slightly different approaches. AMZN’s AWS focuses on a complete menu of atomized services, enabling customers to craft their own platform from multiple choices, both open source standards and AWS proprietary solutions. Partners like VMW provide the integration with on-premises resources. MSFT acts as its own integrator and offers its Azure platform stack as an on-premise as well as in-cloud solution. Its platform has great momentum in MSFT focused shops and offers strong open-source support. GOOGL emphasizes superior performance via proprietary configurations of open-source standards and value-added resources, like AI APIs. For all three, moving up the stack is a strategy to expand margins and add stickiness to their increasingly commoditized IaaS.
- Traditional IT vendors pitch cloud independence. Many enterprises are leery of being locked into IaaS host platform solutions, having spent the past decades suffering from their inescapable reliance on software vendors like ORCL and SAP and hoping to maintain some leverage over their cloud vendors. Traditional IT vendors like IBM, VMW and CSCO play to those concerns with HIP solutions that lean on open source software, preserve multi-cloud flexibility, and play nice with existing custom applications and in-house datacenter resources. While these companies often offer hosting services, the platform opportunity is first in their approach to the customer. Still, none of the companies yet offers a HIP that is truly “turnkey”, and enterprises must customize and pull in 3rd party tools to augment these solutions.
- Best-of-breed tools vendors serve both enterprises and integrators. The most sophisticated enterprises may build their own platform solutions, using tools offered by focused vendors. For example, NTNX offers tools to optimize multi-cloud and hybrid cloud operations, PVTL offers a platform to streamline the development and implementation of in-house software for cloud environments. MDB is an open source standard based, cloud-based database product that is gaining traction with leading edge IT organizations. Docker is an independent containerization solution. These, and other vendors, also partner with both cloud hosts and traditional IT companies to integrate their point products into broader solutions.
- IT sophistication and heterogeneity will drive platform choices. Most enterprises do not have the IT capability to implement a cloud platform strategy on their own. These customers will gravitate toward more integrated solutions offered by cloud hosts, their partners or by more independent traditional IT vendors. Their choices will be partly dictated by the complexity of their current IT operations (and their fear of “lock in”) – the greater the heterogeneity, the greater the impetus toward a more independent integrator.
- Room for multiple winners. We expect PaaS services to be a differentiator for IaaS hosts – a major advantage for MSFT, but perhaps an opportunity for GOOGL as well – and an opportunity to earn margin atop increasingly commoditized basic IaaS hosting. All three hosts will look to M&A and close partnerships to beef up their platform credentials, with AMZN and GOOGL most likely to make aggressive plays. We see IBM (with added strength via RHT) and VMW as leaders amongst the traditional IT side and expect VMW’s relationship with AMZN to continue to strengthen. Of the smaller players, we see NTNX and PVTL as the best placed, with acquisition by a cloud or traditional IT player very likely.
Putting it All Together
Most IT managers know that they are headed to the cloud. Its cheaper. Its more flexible. It is more powerful. It supports advanced capabilities, like AI accelerators and blockchain-style secure transactions. It gives access to world-class expertise in critical areas like security, deep learning, natural language processing, image recognition, big data analytics, and every other IT buzzword. Its just a matter of time.
And money. And headaches. For organizations that weren’t born web-native, the transition seems overwhelming. SaaS applications are the easy way. Replace your customer relationship management software with a subscription to Salesforce, start shifting from Office to Office 365, begin plans to downsize your datacenter and staff. However, most companies can only rely on SaaS for some of the change. Big, mission-critical applications, customized to the specific needs of the organization, tied to massive structured databases, running on half-depreciated in-house datacenters are the soul of the enterprise. The big cloud hosts – AMZN, MSFT and GOOGL – would love to see enterprises reconceive these applications as “cloud native”, but this is entirely impractical. How then, can enterprises take advantage of the cloud while keeping control of performance, costs, security, flexibility, user support, data assets, new development and other shared functions in a complex, heterogenous environment?
Historically, these controls resided in the middle of the classic software stack, in the “infrastructure” layers that included operating systems, database management systems, virtualization hypervisors, run-time platforms, development rubrics, and other tools. Looking forward, these solutions seem inadequate to cope with applications that may run on-premises AND in the cloud, that may run on multiple hosts at the same time. Moreover, new technologies – containers, APIs, AI development, mobile applications, etc. – add new demands. A category of new solutions, some adapting traditional IT approaches (Hybrid Integration Platforms) and some focusing on encapsulating classic datacenter tech within a “cloud native” platform, are emerging, addressing what we believe is a $400B market.
Competition is coming from three directions. The big cloud hosts – AMZN, MSFT and GOOGL – are anxious to move up the stack into platform functions, eying new higher margin business and increased stickiness for plain vanilla hosting services. AMZN has been the least adventurous, choosing partners like VMW for integration with on-premises systems and offering an exhaustive menu of self-serve hosted functions. MSFT can leverage its huge base of existing infrastructure software customers, folding in multi-cloud support to a very robust hybrid-cloud solution. GOOGL is playing catch-up and seems to be relying on high-performance turn-key configured services. It has partnered with CRM for distribution, but also has plans to aggressively build out its own enterprise salesforce.
Traditional IT suppliers covet this market, pitching flexibility and independence to enterprises fearful of being locked into an IaaS hosting vendor. We see IBM (with its recent deal for RHT) and VMW (partnered with AMZN) as the best positioned here, although CSCO is making an aggressive play. Most other traditional infrastructure software vendors will be threatened, including ORCL and SAP. There are also a flotilla of smaller players selling focused products aimed at pieces of the platform opportunity. These companies sell directly to very sophisticated end customers and in partnership with either hosts or integrators to the hoi polloi. We like NTNX and PVTL best of this group, which will also see M&A interest from the larger players.
The Wild Blue Yonder
The cloud era is upon us. The hyperscale datacenters operated by the top cloud hosts offer huge benefits to enterprises. We have estimated that the top hosts run at all-in costs as much as 90% below traditional enterprise datacenters, and even on a marginal basis could offer better than 50% savings (Exhibit 1). The public cloud host also bundle world-class security and offers superior up-time performance and recovery. The cloud is flexible, allowing enterprises to dial up temporary capacity as needed and to access specialized resources, like GPUs or FPGAs configured to accelerate AI training. Almost all new applications are now developed to make use of the public cloud and longstanding enterprise software leaders have made extending their franchises into a public cloud era a priority.
Exh 1: SSR Cloud Cost Comparison for On-premise and Public Cloud Services
We believe that the enterprise transition to the cloud is proceeding more quickly than had been expected. The growth of annual spending on SaaS applications accelerated to better than 30% in 2018 (Exhibit 2). IaaS cloud hosting is also growing at a 30% annual clip, with the top players – Amazon, Microsoft and Google – all continuing to take market share. This shift is driving benefits to the enterprises that are move expeditiously, but there are substantial obstacles for most companies. Some applications are better candidates for the cloud than others, meaning enterprise IT gets more complex long before it can start getting simpler and making it hard to eliminate fixed costs needed to support the older applications. That
Exh 2: Global Public Cloud Spending in all 3 Categories Forecast, 2017 – 2021E
Exh 3: Snapshot of major computing eras and a Hybrid and Multi-cloud future
complexity must be monitored and managed. For applications that move to the cloud, there are transition costs – implementation work, users and staff to train, redundant operations to maintain for a while (Exhibit 3). Reliance on outside vendors can bring lock-in, bringing higher prices, reduced flexibility and loss of control – IT managers who have been under Oracle’s thumb for years may be leery of beginning another dependency.
Exh 4: Independent surveys suggest high number of large enterprises are adopting complex multi-cloud strategy
This means most organizations will be operating in a “hybrid cloud” mode – some applications entirely run on public clouds, some run entirely in-house, and some straddling the two domains (Exhibit 4). Moreover, most will also operate on multiple clouds, both to avoid lock-in and because application needs dictate it. This will strain the capabilities of enterprise IT departments, which will require powerful software tools to manage connectivity, efficiency and security as they enable interactions between disparate hardware and software architectures.
Perhaps the most important technology behind “hybrid cloud” is Containerization (also referred to by the open source standard Kubernetes), an idea first developed within Google but later contributed to open source. Containers break application run-times into small, discreet, secure pieces that can be executed separately, with the results subsequently combined. This gives enterprises the flexibility to assign individual containers to specific processing resources depending on cost and performance. However, containers require an “orchestrator”, essentially a software platform that manages the processes of breaking computing tasks into executable fragments, assigning those fragments to servers, and retrieving and combining the executed containers, while monitoring the whole operation for performance, security and cost. In this, the container orchestrator is akin to the virtualization hypervisors that gave IT the ability to assign individual compute jobs to compartmentalized “virtual” servers running simultaneously on shared datacenter server
Exh 5: Platform layer acts as a glue integrating Apps at top, Infrastructure at bottom, legacy systems and other supporting middleware
clusters. Containerization takes this a giant step forward by abstracting the compute job down to discrete calculations (Exhibit 5, 6, 7).
Another important role in the platform layers is managing application interaction with data. Traditionally, Data Base Management Systems (DBMS) like Oracle or SQL handled this job, but in a hybrid cloud world where data flexibility is valued, where analytics and AI demand non-traditional access to data, and where volumes of important data are collected and processed in real time, a classic DBMS becomes a major obstacle. New cloud native databases are the long-term solution, but the costs of disrupting mission critical applications that rely on existing DBMS are very high. In the intermediate term, solutions that facilitate the sharing of captured data through high-performance APIs have merit. In addition, Apache Kafka is an open source standard for building real-time data pipelines and defining the way in which individual applications can access them. Kafka is becoming widely adopted as a piece of HIP solutions. At a more basic level, the platform layer must also assure that communications that run between programs either adhere to a single agreed mode of communications or that necessary translations are performed. More involved transactions between disparate applications may require customized APIs, which must be developed, implemented and managed from the platform layer.
Exh 6: Quick Comparison of Virtualization and Containerization Technologies
Exh 7: Comparison Between Virtualization and Containerization Architectures
Security is another key factor for the hybrid cloud. Authentication must carry across public and private datacenter venues facilitating valid access without impediment but also restricting unauthorized users and assuring the integrity of enterprise data. The platform layer must also offer a robust environment for development, particularly given new technology like machine learning, blockchain, IoT, etc. which may demand specialized infrastructure or access to unusual expertise (Exhibit 8). While these resources may reside with a cloud provider, they may be incorporated into applications that will need to run locally. The platform layer should make this integration easy.
Exh 8: The platform layer combines critical functions of cloud native development
IBM CEO Ginny Rometty pegs the hybrid cloud as “a trillion-dollar market”. While there is more than a little sales hyperbole in her assessment, the opportunity is indeed very large. Start with the current commercial market for infrastructure software – operating systems a la Microsoft Windows, DBMS like Oracle, virtualization platforms like VMWare and other big platform layer functions – which currently represents $240B in annual spending worldwide. Add a kicker for some datacenter infrastructure that could be downsized, likewise for internal IT personnel. Consider also that a decent part of the $165B spent on IT consulting also derives from infrastructure issues. Putting it together suggests an addressable market that is likely north of $400B, not quite a trillion but still a considerable opportunity (Exhibit 9).
Exh 9: SSR Estimate of Total Addressable Market for Cloud Native PaaS Solutions
Lined up against this market are three different sorts of competitors approaching it from different directions. In one corner, the hyperscale datacenter IaaS hosts Amazon, Microsoft and Google (Alibaba too if you happen to come from outside the US). The big cloud players see the writing on the wall. Containerization and other hybrid cloud technologies threaten to commoditize basic computing and storage by making it easy to direct volume to different clouds based on price and execution. Moving up the stack into platform layer functionality adds differentiated value-added services that can carry higher margins, and conceivably, could make the use of the underlying infrastructure “stickier” (Exhibit 10).
Exh 10: Snapshot of Cloud Native Middleware and PaaS Companies
Exh 11: Traditional IT firms looking to advance with “Hybrid Cloud” Approach
Traditional information technology vendors, largely burned by the rapid rise and scale advantages of the big three hosting companies, see the hybrid cloud as an opportunity to defend their turf while working to relegate the IaaS hosts into that commodity role (Exhibit 11). Often with specialized knowledge about legacy systems and applications, and with perceived independence from the cloud hosts, these longstanding suppliers can tout incrementalism and independence as meaningful benefits. Support for open source standards, like Kubernetes and Kafka, give IT managers confidence in the future portability of the solutions, even if a bit of platform lock-in is inevitable.
Finally, a horde of startups has been attracted to the space (Exhibit 12), offering piece part tools that an ambitious enterprise could use to build a hybrid integration platform of their own in combination with open source software that is freely available. A flexible database product here, a container orchestrator there, a data streaming solution, and so on. Some solutions may integrate several functions, but all typically support open source standards and promise infrastructure independence.
Exh 12: Snapshot of Cloud Native Middleware and PaaS Private Companies
Platform as a Service
Amazon Web Services is the big gorilla of cloud hosting, built on a philosophy of giving customers what they wanted (Exhibit 13). At first, this meant serving web-based businesses like Netflix, which built their applications as web native from the ground up and seamlessly integrated cloud hosting with their own datacenter assets. AWS was set up as an extensive self-serve menu, with new services added in response to the expressed demands of the customer. Over time, more traditional enterprises entered the mix, first using AWS as an offline development environment, then as flex capacity for applications that had been revised to make use of the service. These customers had different demands and AWS was responsive to them, adding support for classic infrastructure software like Oracle databases and VMWare hypervisors in addition to the open source standards favored by the web-native crowd.
The rise of Microsoft’s Azure platform and Google Cloud Platform as cloud alternatives has placed some pressure on AWS to find ways to make its relationship with its customers stickier and to add higher margin services in the platform layers of the software stack. Still, Amazon’s offerings are still more responsive than innovative, with me-too AI APIs, container orchestration, and other value-added offerings. AWS is also less proactive in defining configured integration platforms for hybrid services, relying instead on 3rd party integrators. VMWare is a particularly close partner, hosting its VMWare Cloud solution directly on AWS and using it to help customers from its dominant virtualization franchise transition to a more flexible container driven approach.
Microsoft Azure is AWS’s most potent rival. Its close integration of its cloud software stack and its on-premises infrastructure software give it a powerful advantage for the many clients who already rely on its solutions. Its solutions are much more configured than Amazon’s offering a more turnkey integration
Exh 13: Comparison of Big Cloud Positioning in Platform Layer