Server SAN 2012-2026

Wikibon Analysts:
David Floyer
Stu Miniman
Ralph Finos

Executive Summary Server SAN

Enterprise Server SAN has revenues greater than $1 billion, with an annual current growth rate of greater than 150%. This research looks at why Server SAN is growing so fast, who are the main players, and what is the likely impact on current Enterprise networked storage.

Wikibon’s extended forecasts are designed to paint a picture of the profound changes roiling IT; the exact figures in the outer years are less important than the shape of the curves, reflecting the strong market forces at play. In this research paper, Wikibon reviews the introductory Server SAN research, adjusts the Server SAN definition to include System Drag (see System Drag below for definition), and increases the speed of adoption of Enterprise Server SAN based on the very fast adoption from 2012 to 2014. Wikibon has also integrated our Cloud research into the growth of Hyperscale Server SAN, and split out SaaS, IaaS, and PaaS as components. Based on our cloud research, Wikibon has increased our outlook on the proportion of spend on Hyperscale Server SAN, and lowered the later years of Enterprise Server SAN.

The results of this revision is shown in Figure 1. The overall growth of Server SAN is projected to be about a 23% CAGR from 2012 to 2026, with a faster growth from 2014 to 2020 of 38%. The total Server SAN market is projected to grow to over $50 billion by 2026, split 44% for Hyperscale Server SAN and 56% for Enterprise Server SAN. Wikibon projects that Server SAN will replace most of the Traditional Enterprise Storage which is projected to decline to about 12% of current revenue by 2026 (see Figure 6 below). The traditional enterprise storage market is projected to decline by -16% CAGR.  Overall growth in storage spend is expected to grow at 3.1% CAGR through 2026.

Wikibon expects that distributed edge Server SAN will grow to support the adoption of the Internet of Things and individual decision making, and will interconnect to mega-datacenters that are using an increasing amount of Hyperscale Server SAN. These mega-datacenters will connect cloud services and collocated Enterprise Server SAN with fast secure  hyper-networks. In turn, the mega-data centers will connect to Enterprise private/hybrid clouds to support local enterprise storage and processing.

The adoption of Server SAN will be driven by much lower cost, increased functionality, flexible performance and improved interconnectivity. The increasing percentage of flash storage will allow the deployment of very low latency (less than 50 microseconds) and high bandwidth storage required by hyper-databases within this interconnected mesh. 2015 will see the first adoption of this technology with the expected shipment of EMC’s DSSD, supporting Flash as Memory Extension (FaME). The traditional storage model based on islands of proprietary code supporting proprietary data services and proprietary management tools will not be a significant part of this interconnected computing beyond 2020.

Wikibon strongly recommends that CTOs & CIOs initiate Server SAN pilot projects in 2015, particularly for applications where either low cost or high performance is required.

Server SAN Update

Wikibon defines Server SAN as a combination of compute, pooled storage resource, local networking and distributed software. The compute comprises multiple servers (more than one). The storage comprises more than one storage device directly attached (e.g., SAS, PCIe) to separate multiple servers. Local communications between the direct attached storage (DAS) and the servers is enabled by high speed interconnects (e.g., InfiniBand, Low Latency Ethernet, PCIe switches, etc.). Coherency, security, availability and data management are managed by the software of the solution. Server SAN storage can utilize both spinning disk and flash storage. The applications can run concurrently on the Server SAN servers using virtualization and/or containers. Wikibon defines two major Server SAN segments, Enterprise Server SAN and Hyperscale Server SAN (mainly used by cloud service providers and large single application providers such as Facebook and Google search). Server SAN is sometimes called Hyperscale, although Server SAN also includes very small configurations that can be found in edge servers for the internet of things.

Converged infrastructure is a combination of servers, storage networking and other technologies delivered a single package and SKU. The benefits of converged infrastructure are greater the more that the components are integrated. Wikibon has written extensively about this, and has shown how minimizing the number of SME (Single Managed Entities) increases the value. In Wikibon research entitled “The Business Value of Converged Infrastructure Topologies“, Wikibon emphasizes that the ideal converged infrastructure is a single SME with a Single Source (SKU), Single update, Single handshake and Single throat to choke. Example of converged infrastructures include EMC’s Vblock, NetApp’s FlexPod, and Oracle’s Exadata. Converged infrastructure can use traditional SAN storage, NAS storage and Server SAN. EMC’s Vblock is used to deliver a Server SAN solution using ScaleIO. Wikibon believes that using Server SAN will lower converged infrastructure costs and improve performance, and over time will become the de facto standard for converged infrastructure.

In 2014, Wikibon defined and measured the Server SAN market and projected the growth through 2027. In this research, Wikibon enhances the definition of Enterprise Server SAN Revenue to include Systems Drag (see next section for more discussion on Systems Drag). Enterprise Server SAN grew from a restated $370 million in 2013 to $1.1 billion in 2014, at a growth rate of 187%. Figure 2 below shows this rapid growth, and the contribution of the System Drag component. Server SAN Vendor only Revenue grew from $268 million to $708 million, at a growth rate of 164%. The Server SAN Systems Drag grew from $102 million in 2013 to $354 million in 2014, at a growth rate of 249%.

In 2013, most of the revenue came from appliances, which included both the hardware and software. 2014 has seen a significant increase in software-only sales, especially from VMware’s VSAN product. The reason for including System Drag is to allow a consistent comparison with traditional SAN markets, and a more accurate comparison of market position between appliance and software vendors.

Enterprise Server SAN is one component of the overall Server SAN market, which includes Hyperscale Server SAN as the other component. With the addition of Systems Drag to Enterprise Server SAN, and an update of the Hyperscale Server SAN, the Wikibon Server SAN Total Market for 2013 is restated as $2,554 million, up from the original estimate of  $1,717 million in the Wikibon research “The Rise of Server SAN”. The Enterprise Server SAN proportion of the market is 14% in 2013.

In 2014, Enterprise Server SAN grew 187% to $1,061 million, breaking the $1 billion mark. The restated Hyperscale Server SAN grew 50% to reach $3,285 million. The Total Server SAN revenue grew 70% in 2014 to $4,347.

These updates are shown in Figure 3 below.

Systems Drag

Systems Drag is revenue from additional hardware and software not provided by the Server SAN vendor, but necessary to enable a Server SAN. For example,  the majority of revenue from Nutanix comes from appliances supplied directly by Nutanix, but additional Server SAN drag comes from Nutanix sales of software-only to system vendors such as Dell, as discussed in a Wikibon post entitled “Dell and Nutanix Introduce a Converged VDI Application Appliance”. Systems Drag was not included in the original definition of Enterprise Server SAN laid out in “The Rise of Server SAN”.

Wikibon has included System Drag in order to normalize software-only Server SAN vendors such as VMware from mainly appliance vendors such as Nutanix. The Systems Drag includes storage hardware, server hardware to drive the storage and other software required to implement Server SAN. This definition will allow more relevant storage forecasting as Server SAN grows over the next decade to be a dominant storage model.

A general Wikibon model for System Drag for software-only sales is 36% Enterprise Server SAN software revenue and 64% additional non-vendor revenue. The additional revenue includes storage hardware, server hardware and additional software. Systems Drag represents 27% of the Enterprise Server SAN market in 2013, and 33% in 2014.

Enterprise Server SAN Vendor Revenue & Participation

Contributors to the very high growth of Enterprise Server SAN include the maturing of software, the significant growth of  vendors already in the market, and the very strong growth of new entrants such as VMware’s VSAN. The original estimate made by Wikibon in 2014 was that the Enterprise Server SAN market would grow by 77%. The actual growth in 2014 was 187%, as shown in Table 1.

Table 1: Wikibon Enterprise Server SAN 2014 Update
Source: © Wikibon Server SAN Research Project 2015

The Enterprise Server SAN Revenue by vendor is shown in Figure 4. In this figure, the System Drag is not allocated by vendor, and is included in Figure 4 as an overall single column of $354 million. The first five positions in Server SAN vendor revenue are held by Nutanix, HP ESA, SympliVity, VMware and Pivot3.

The Enterprise Server SAN Vendor Revenue Participation is shown in Figure 5 below. In this Figure, the System Drag is allocated by vendor, and is shown as a red bar on top of the blue revenue bar. The largest contribution to Server Drag is $86 million from VMware. The first five positions in Server SAN vendor participation (including System Drag) are held by Nutanix, HP ESA, VMware, SympliVity,  and StorMagic.

Enterprise Server SAN Vendor Observations

Figure 6 below shows the 2013 and 2014 percentage market shares of the top 70% of leading Enterprise Server SAN vendors in the Enterprise Server SAN market. These percentages include System Drag.

The Server SAN storage services stack is still maturing. The stack is more mature in early vendors such as Nutanix, HP ESA, SimpliVity and Pivot3. For more analysis on the hyperconverged vendors, see Server SAN Enabling a Modern Operational Model of IT.

Nutanix grew its overall market participation/penetration from 17% in 2013 to 22% in 2014, growing from $61 million to $236 million (286%). Nutanix has a strong channel program, and a clear value proposition, as the Wikibon research on the partnership with Dell illustrates.

HP ESA grew from $100 million to $169 million (69%) from 2013 to 2014, but declined in market share to 16%.  HP will need to make their product more available through their channels, and their server channels in particular.

VMware VSAN has had great success in 2014 growing from 0% to 13% market share ($169 million) , and is poised for strong growth in 2015 and beyond within the VMware ESX marketplace. VMware VSAN has been held back because VMware/EMC focussed on supporting traditional storage arrays. VMware will need to accelerate the completion of an effective storage services stack to take advantage of its market position. In addition, enterprise customers are increasingly using a mixture of hypervisors within the datacenters. Wikibon believes that VMware will need to be able to support and manage across hypervisors and containers in order to increase their TAM (total available market). VMware has started to reach out to manage other platforms with the Cloud/Realize products. VMware will need to decide between a high value VMware-only strategy for VSAN, or a broader high-volume general hypervisor with minimal tie-in to enterprise editions.  This is an important strategic decision for VMware, and history is consistently on the side of winners going for volume.

SympliVity has had an outstanding year, growing from $15 million to $75 million (400%), and growing market share from 4% to 7%.

Pivot3 has had a very successful Server SAN it has delivered to the video surveillance industry, and has an opportunity to promulgate a very efficient technology into the broader market.

Having initially put ScaleIO on ice after the acquisition, EMC has decided to move more aggressively in 2014 and 2015 to market ScaleIO in competition with its own storage arrays. EMC has also announced a VCE-based Server SAN appliance. This is an important development, because converged solutions can be built with much greater flexibility and lower cost than with traditional SAN-based systems, and with easier long-term integration of the ScaleIO and VCE management systems. To be competitive, the software uplift on the storage hardware revenue will be much lower than the traditional EMC SAN storage uplifts of 10x to 20x the cost of the storage drive.

An even bigger opportunity for EMC is with the DSSD acquisition made for $1billion in 2014 (see Footnotes-4). This very low latency (50 microsecond) flash storage uses Flash as Memory Extension (FaME), with direct PCIe connections to servers through PCIe switches. Wikibon expects DSSD and other similar products to make strong inroads into the high-value high-end and higher margin segment of the Server SAN marketplace.

Perhaps the major disappointment for 2014 was the low marketing and technical emphasis placed by Microsoft on Storage Spaces.

Additional comments on leading Server SAN vendor strategies can be found on Wikibon Premium at “Server San Enabling a Modern Operational Model of IT”

Enterprise Server SAN Functionality & Investment

Table 2 below shows a high-level functionality assessment, and investments by vendor.

The functionality matrix is an indication of the total market TAM that can be reached. The total functionality reach is the sum of the “Y”s in the functionality column. Datacore’s SanSymphony, the Citrix Sanbolic and Nexenta platforms have the highest functionality scores (7/7), with EMC ScaleIO close behind (6/7).

The growth in Server SAN has been funded by significant investment. Investment in Enterprise Server SAN grew 100% from approximately $625 Million during and before 2013, and rose to a total of $2,262 Million during and before 2014. In 2014, Sanbolic was acquired by Citrix, and Ceph Inktank has been acquired by Red Hat for $175M (to add to Gluster acquired by Red Hat for $136 million in 2011).  DSSD was acquired by EMC in 2014 for $1 billion, including earlier investments (see Footnotes-4), and is expected to start delivering products in late 2015. No Enterprise Server SAN vendors went public in 2014.

Key IT Trends

There has never been so many simultaneous and interconnecting trends in IT and the business use of IT technologies. Some of the important trends include:

• Converged infrastructure, the continuous integration of infrastructure into larger components that are managed as a single component;
• The rapid growth of software services as a cloud offering (SaaS), and the consequent lowering of maintenance and service costs, and the increase in
• The development techniques that allow a move away from software versions every year or so to continuous improvements with new versions every day if necessary;
• The continued growth of virtualization and containerization to lower hardware dependencies;
• The rapid growth of cloud infrastructure delivered as services (IaaS) that allow faster deployment and lower initial capital costs for IT projects;
• The rapid increase in data volumes from mobile devices and the expected huge increase of data from a nascent internet of things;
• The growth of mega-datacenters that house clouds, data aggregation and collocated enterprise computing, and provide ultrafast secure networks to minimize data movement and time to data;
• The growth of distributed edge servers to manage the internet of moving and static things, and migrate what is important to the cloud data aggregator services;
• The rapid migration to flash as the dominant storage medium for active data;
• The ability of new architectures such as FaME to provide several orders of magnitude improved levels of consistent low latency and higher bandwidth for applications;
• The tremendous advances in systems of intelligence that can utilize the vast amounts of data and allow near-real-time analytics;
• The ability to combine transactional and analytic systems to create true real-time businesses driven by real-time automated processes and software;
• The ability of these technologies severally and in combination to add enormous business value from personnel efficiency, reduced business cycle times, reduced time to adapt and change, and reduced risk;
• The movement of proprietary solutions from IT vendors towards open ecosystems of system and software value with support services from IT vendors;
• The movement of IT vendor value-add from hardware optimization for IT to software delivery of value to the business.

Hyperscale and Enterprise Server SAN are technologies that enable more efficient  and more open converged infrastructure of servers, storage and networks, greater speed and flexibility in integrating new technologies such as FaME, and significantly lower acquisition and operational costs for cloud services and enterprise cloud computing. Hyperscale Server SAN is a key enabler of cloud services such as infrastructure as a service (IaaS) and a prerequisite for efficient Software as a Service (SaaS) and Platform as a Service (PaaS). The general movement again is towards increasing emphasis on software and platform. This trend will become clear in the Cloud and Hyperscale Server SAN storage projections in the next section.

Table 3 shows the differences between SaaS, PaaS and IaaS. SaaS provides the application and everything underneath, and the user provides the data and is ultimately responsible for security and compliance. PaaS provides a total set of micro-services to support the user development of applications. These applications can be delivered on top of a IaaS cloud (PaaS on IaaS), or on top of SaaS applications. Haas provides virtualized hardware and increasingly OS and middleware services, but the user is responsible for integrating and maintaining the pieces. The amount of software is highest on PaaS, next on SaaS, and lowest on IaaS. Overall the percentage of added value through software and software services is increasing.

Cloud Vendor Revenue Projections 2012-2026

One of the major disruptive forces is the movement of work from traditional enterprise datacenters to the cloud. Wikibon has done extensive work on the definition and growth of Cloud computing, and the breakout of  SaaS, PaaS and IaaS.  The scope definitions are shown in Table 3 above.

The Table Footnotes-2 is based on Wikibon cloud research that is just about to be published. This tables shows the actual cloud vendor revenue for 2012-2014 and the first half of 2015. The data is split between PaaS, SaaS and IaaS.

The data is derived from detailed analysis of financial statements of over 150 cloud vendors and service providers. The model projects the future revenue of each cloud player until the end of 2016. The data is completed with estimations of the “other” category.

This data is used as the basis for the long-term projections using Wikibon adoption curve technology. The key IT trends are applied to each segment, and an overall projection constructed out until 2026.

Figure 7 shows the vendor revenue projections for the SaaS cloud component, the IaaS cloud component and the PaaS Cloud component. SaaS has continued projected adoption growth to about $300 billion in 2026, which is a significant proportion of the total IT spend. IaaS has a larger hardware component, and projected to grow to about $100 billion by 2026. Wikibon believes that PaaS will take take longer to mature, but will grow strongly in the outer years.

Figure 8 shows the cloud vendor projects by component from 2012 to 2026. PasS has an overall growth rate of 32% CAGR from 2014-2026, with faster growth projected to start further out. Both of software-heavy cloud components together, Saas and PaaS, are projected to grow at 20% CAGR, and the growth of the IaaS cloud is projected to grow at 16%.

Hyperscale Server SAN Projections

The Hyperscale Server SAN storage spend for 2012 and 2016 is calculated in Table Footnote-2 in the Footnotes below. The key assumptions are taken from the first three rows of column f, which show the percentage of software assumed with IaaS (40%), SaaS (85%) and PaaS (90%). What is not software is assumed to be hardware, and 19.4% of this cost is assumed to be storage.

The Wikibon research for IaaS, SaaS and PaaS is used to project the Hyperscale cloud storage by component, as shown in Figure 9. IaaS shows the greatest amount of storage spend, and PaaS the fastest growing further out. The overall growth of Hyperscale Server SAN is projected to grow at an overall 22% CAGR 2014-2026. . The growth rate for the period 2014-2020 is projected to be 34%. Compared to the previous research note in “The rise of Server SAN”, Hyperscale Server SAN is projected to be a slightly larger proportion of the Server SAN spend by 2026 (56%). This data is shown in more detail in Figure 10 below.

Enterprise Server SAN & Overall Storage Projections

Enterprise Server SAN exploded in 2014 with a 187% increase over 2013, and is projected to grow at a 44% CAGR through 2026. Server SAN is a lower cost solution that brings the storage closer to the server, and builds an ecosystem that is more open, lower cost and flexible. It is a much better fit for distributed edge storage for the internet of things, and much faster and closer to compute for high performance workloads. The growth rate for the period 2014 through 2020 is projected to be 62% CAGR. Figure 11 shows Enterprise Server SAN growing rapidly from a low base, compared with Hyperscale Server SAN growing from a higher initial base. The overall storage is projected to grow at about 3% CAGR, in line with general IT growth.

Traditional SAN/NAS/DAS, with its generally proprietary software in the storage controllers is projected to decline by 16% per year. Flash storage arrays will start to take a high proportion of this spend after 2016. During the period 2014 to 2020, the traditional enterprise market is projected to decline by 14% CAGR.

Figure 12 below shows the interaction between traditional enterprise storage, Enterprise Server SAN and Hyperscale Server SAN. Traditional enterprise storage is being squeezed in a vice between a superior, lower cost and more flexible storage model with Enterprise Server SAN, and the migration of IT towards  cloud computing and Hyperscale Server SAN deployments.

The one area where traditional storage systems can continue to serve is the long-term data retention segment. This usually requires infrequent access to the data, but more frequent access to metadata about the archived data. Wikibon has discussed this segment in a research paper entitled “The Emergence of a New Architecture for Long-term Data Retention.” In this segment, object based architectures can distribute data across multiple sites using highly efficient erasure coding. Because of the fast data transfer character of tape, “Flape” (Flash & Tape) systems could be ideal for this kind to system. Companies like DDN with WOS, Cleversafe and Scality are pioneering new ways of drastically reducing the cost of long-term data retention.

Wikibon expects this segment to remain in the traditional SAN/NAS/DAS systems, and for long-term data retention to be a major part of the outer years of traditional storage.

Action Item

Action Item: Enterprise Server SAN is a rapidly evolving lower cost model for storage, and is a simpler and lower cost fit for converged infrastructure than traditional storage arrays. Like many other entries into the storage market, Server SAN is being initially implemented in applications that are particularly sensitive to storage costs, such as VDI and video surveillance.  Wikibon strongly recommends that CTOs & CIOs initiate Server SAN pilot projects, particularly for applications where either low cost or high performance is required. 

Footnotes:

Footnote 1: Additional research and perspectives can be found on Wikibon Premium at “Server San Enabling a Modern Operational Model of IT“

Footnote 2:  Wikibon Enterprise Server SAN and Hyperscale Server SAN Estimates for 2013 & 2014. This table is the source of all the Figures and Tables in this research.

Footnote 3:  Wikibon Hyperscale Server SAN Cloud Storage Spend Estimates for 2013 & 2014, and part of 2015. All the cloud revenue numbers are based on the analysis of public company financial statements, with additional information survey data and some company interviews. The data for remainder of  2015 and 2016 is based on expected performance by the 150 business that make the portfolio of cloud companies that are followed in detail by Wikibon. These performance estimates are based on historical performance, public statement, interviews of Wikibon practitioners, and other sources of data.

The data in Table Footnotes-2 is a summary of the detailed data model. The first three rows give the IaaS, SaaS and PaaS past and expected revenues from 2002 to 2016. To look at hardware data, the software component is extracted.  Column “f” of Rows 1-3 shows the percentage of revenue that is based on software for IaaS, SaaS and PaaS. Pass is estimated to only have 10% of revenue that is driven by hardware, and only applications and data are brought to the PaaS platform; the rest is software. That data is used in row five to give and estimate of the cloud hardware component. From previous research, an estimate of the amount of hardware cloud revenue is just less that 20%. The total amount of IaaS, SaaS and PaaS storage is shown in rows 7-9. Row 10 give the total of hyperscale Server SAN (Cloud Storage) in each type of cloud (XaaS) by year. Row 11 shows that about 5% of total cloud revenue supports Hyperscale Server SAN.

Footnotes-4: The EMC DSSD information was downloaded as a WSGR.PDF file on July 5 2015 at 1:34pm PT from https://www.wsgr.com/WSGR/DBIndex.aspx?SectionName=attorneys/BIOS/2997_2.htm. 

WSGR