case

XENSOURCE (A)
Itís like David versus Goliath, except David is coming with an uzi.
óKevin Compton, venture capitalist and XenSource board member
In early 2007, Peter Levine climbed a steep hill on a bike ride through Portola Valley, California. Levine was the president and CEO of XenSource, a 70-person software start-up in the virtualization space offering a unique blend of open source and proprietary software. He and his team wanted to establish a strong second place position relative to VMware, the marketís established incumbent and a $1-billion company with a head start of almost a decade. XenSource had debated several ways it might compete and decided that working closely with OEMs and VARs1 seemed to be the best option. As Levine said, ìThe engineering process proved to be the model for our go-to-market plan. The open source community leveraged our product development with incredible speed and efficiency. We realized that leveraging partners and external resources for our marketing and sales strategies made just as much sense.î Levineís team knew their strategy was risky. Relying on partners meant engaging with the biggest players in the software and hardware industries, companies such as Microsoft, Intel, Dell, and HP. One could never be certain whether such behemoths would be friends or enemies, nor how the sands might shift. Because Levine and many of his executives were molded at those very same companies, they realized the breadth of the challenges: in addition to incorporating a publicly controlled software ìengineî into a for-profit product without alienating the open source community, they would have to navigate a daunting competitive and partner landscape. As Levine summited the hill, he could see that the next peakís climb would be just as tough. Intensely competitive, Levine knew that the success or failure of XenSourceís partnership decisions would be critical to the companyís long, uphill battle. 1 OEM (Original Equipment Manufacturer) and VARs (Value Added Resellers). Katherine Rudolph Bose prepared this case under the supervision of Mark Leslie, Lecturer in Business, as the basis for class discussion rather than to illustrate either effective or ineffective handling of an administrative situation. This case was made possible through the generous support of Mr. Alain S. Rossman. Copyright © 2008 by the Board of Trustees of the Leland Stanford Junior University. All rights reserved. To order copies or request permission to reproduce materials, e-mail the Case Writing Office at: cwo@gsb.stanford.edu or write: Case Writing Office, Stanford Graduate School of Business, 518 Memorial Way, Stanford University, Stanford, CA 94305-5015. No part of this publication may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted in any form or by any means ññ electronic, mechanical, photocopying, recording, or otherwise ññ without the permission of the Stanford Graduate School of Business. CASE: E-267 (A) DATE: 06/25/08 Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 2 INDUSTRY BACKGROUND Servers XenSourceís virtualization technology was focused on server computers. Servers are typically designated for running a dedicated function such as for the web, a specific application or a database. Because they are designed to handle heavy workloads, servers usually have specialized hardware features, operating systems, and ìmiddleware.î Hardware Some of the most successful server manufacturers in the world are: International Business Machines (IBM), Sun Microsystems, HP, Dell, and Fujitsu/Fujitsu Siemens. Most of the servers are based on processors and related hardware developed by Intel and Advanced Micro Devices (AMD). Some server manufacturers, such as Sun, develop custom processors for their products. Software The operating system manages the sharing of resources on a computer and provides an interface (to a user, administrator and / or programmer) to access those resources. Historically the most popular OSs for servers were derived from the UNIX operating system (e.g., FreeBSD, Sunís Solaris, and later Linux), because UNIX was viewed as a logical and efficient choice for servers. In 1993, Microsoft introduced its first server product, called Windows NT Advanced Server 3.1. The popularity of Microsoftís products steadily grew, and in 2000 Windows commanded about half of the server OS market share. Unix (Solaris from Sun) and NetWare (Novell) followed at about 17 percent each, and Linux had 10 percent. By 2005, Windows unseated Unix for the first time as the top server OS in terms of revenue from sales (Exhibit 2). By 2007, Windows controlled about 70 percent of server software market. Linux followed with 20 percent, Unix had less than 10 percent, and NetWare (Novell) was barely registering.2 (See Exhibit 3 for a graphical representation of ìSoftware Wars.î) Virtual Machines3 In recent years, virtualization technology gained adoption and significant attention, even though their origins could be traced back to the 1960s. A virtual machine (VM) monitor, or hypervisor, typically refers to a piece of software that sits between a computerís software and hardware, serving as an abstraction layer between them. The VM essentially decouples the hardware from the operating software, intercepts instructions that would normally go to the hardware, and then either passes them through or performs emulation depending on the specific instruction. The software simulates hardware devices such as input/output, buses, memory controllers, and so on (see Exhibits 4a and 4b). In the mid-1960s, engineers from IBM developed a way to partition the resources of a computer to study the effectiveness of certain new features without impacting the rest of the computer. Since then, software engineers have found many applications for the concept. For example 2 http://www.eweek.com/article2/0,1759,2207368,00.asp (December 17, 2007). 3 This section heavily references Stanford GSB Case E-149: ìVMware.î Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 3 virtualization allows users to run multiple OSs on a single hardware server. This enables ìsandboxî testing environments where many developers could test their software without compromising the rest of the system (instead of each developer requiring a dedicated system). Additionally, because VMs enabled independent management of software and hardware, IT professionals could deploy software services on a pool of machines without getting into the details of which software mapped to which machine and OS. Also, a common complaint from IT professionals had been the low utilization rates in their data centers, often only 10-15 percent per server.4 Low utilization coupled with the rising cost of power for running and cooling server farms offers a very attractive economic return on investment for virtualization. Finally, virtual machines can easily be moved from one physical server to another, facilitating such critical system functions as high availability (failing over virtual machine from failed hardware server). OPEN SOURCE The Xen product was the first open source offering in virtualization. Open source refers to software source code that has been developed as the result of public collaboration and made available to anyone for use or modification free of charge. According to the Open Source Initiative (OSI),5 to meet the standard set by the open source definition, such code had to comply with 10 criteria (see Exhibit 5).6 The standard was largely enforced by an unofficial system of checks and balances that included watchdogs in the open source developer community on one hand, and the desire of commercial companies to stay in compliance on the other. A New Paradigm Traditional software vendors saw their code as their principle intellectual property asset and took great care to patent what was appropriate and to keep the rest as trade secrets. Most importantly, traditional business models charged users for licensing their code. Open source on the other hand was public, free, and redistributable by anyone. One of the great benefits was that the wide and diverse user inspection and participation often flushed out bugs and contributed fixes much sooner than with proprietary software development, resulting in more stable and reliable code. The main advantage of open source, therefore, was the speed and efficiency with which it could penetrate the market, but the big challenge was how to make money without charging users license fees. In the late 1990s, open source business models were in their infancy and unproven. 4 According to Microsoft, ì5 percent [utilization] was not uncommon,î Alex Woodie, ìMicrosoft Unveils "Viridian" Hypervisor, Extends Virtualization Roadmap,î IT Jungle, http://www.itjungle.com/two/two052406-story01.html, May 24, 2006, (June 5, 2007). 5 OSI is a California 501(c) 3 corporation, founded in 1998. http://www.opensource.org/ (May 31, 2007). 6 Abridged version compiled from multiple sources, including OSI and Wikipedia. Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 4 Open Source Business Models Open source, particularly Linux,7 began to gain popularity and traction within businesses, because it is a less expensive way to develop higher quality products in a shorter time frame. With bright, talented and passionate software programmers working on the code all over the world at all hours of the day, even the large, established software vendors realized the benefits of time-to-market and product robustness that open source offered. Venture capitalists were not far behind in seeing the potential of the open source movement and invested heavily in Linux- themed companies in the late 1990s.8 According to Larry Augustin, former chairman of VA Software and a XenSource board member, ìSince 2000, open source start-ups have raised over$1.9 billion in financingî9 (Exhibit 6).
While open sourceís benefits were becoming clear, it still had its challenges. First, people needed to know who was running which version of the code. Even if a company could track and manage that, there was the pervasive belief that everyoneódevelopers and usersócould benefit from running a single, standardized version. The need for a clearing house was born. But how would the middle man make money? Red Hat emerged as the preferred provider by rightly predicting that companies would pay for a gold-standard version that was nicely packaged and well- supported by some of the best minds and developers in the Linux community. (See Exhibit 7 for additional, generic business risks for a company looking to incorporate open source code into its product offering.) As such, Red Hatís strategy exemplified the first of the four open source business models. Other vendors who opted for this model typically charged an annual fee for support, per-student fees for training, and per-project fees for consulting engagements.
The second model was that of a dual-license, where code was simultaneously published under both a traditional open source license and a commercial license. Vendors such as XenSource and MySQL typically charged a perpetual license fee for additional closed-source features, supplementary documentation, testing, and quality, as well as intellectual property indemnification to protect the purchaser from legal liability.
The third model was called ìsoftware as a serviceî or SaaS, and was employed by companies like SugarCRM. The vendor typically charged a monthly subscription fee for use of its service or access to its hosted applications. It did not charge for the software itself, because no software was ever shipped to or installed at a customer site. SaaS was an exceptionally cost-effective way for customers to use various software packages without having to make upfront investments in software and dedicated hardware.
Finally, the functional encapsulation business model featured an open source framework or library, which was installed on a user's computer separately from the commercial product, with the commercial product using the open source functionality in an "arm's length" way (based on
7 Linux is a Unix-like operating system and the most widely known example of open source.
8 Red Hat [NYSE:RHT], Cobalt [Nasdaq: SUNW] and VA Software (aka VA Linux) [Nasdaq: LNUX], which became Sourceforge in 2007, had some of the biggest IPOs in history, posting jumps of 272 percent, 482 percent and 689 percent respectively on their first days of trading (various public sources).
9 Larry Augustin, ìOpen Source Venture Investment Through Q1 of 2007,î Larry Augustinís Weblog, http://lmaugustin.typepad.com/lma/2007/05/open_source_ven.html (June 5, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A) p. 5
the argument that the commercial product was shipped without the open source library, even though it used it). Vendors, such as Technalign, typically charged a perpetual license fee for the functionality that they provided under closed source, as they usually did not provide services or other direct value for the open source elements.
XENSOURCE COMPANY HISTORY
At around the same time that Linux was emerging as a commercially viable solution for enterprises, an ambitious research project called XenoServers was underway in the Computer Laboratory at Cambridge University. Begun in 1999 by senior lecturer Ian Pratt and a handful of others (see Exhibit 8 for a complete list of XenSource executive team biographies), XenoServers was designed to be a global platform for computation. Pratt explained:
We started to build a service where someone could buy fractions of a server instead of the whole thing. To do that, we had to carve up a machine efficiently and securely. Efficiently, because there was value in allocating resources to maximize computing power for the least cost. And securely, because we might be hosting Pepsiís data on one half of the machine and Cokeís on the other, and the two could not co-mingle for obvious reasons.
From the XenoServers project came Xen, open source software to create a hypervisor, or ìa virtual machine monitor (VMM) for x8610 that supported execution of multiple guest operating systems with unprecedented levels of performance.î11 The focus on Xen alone began in 2002.
Pratt had been a longtime friend of Simon Crosby, a fellow lecturer at Cambridge. A few years earlier, Crosby had left academia to work for several different start-ups before joining Intel, where he ìspent a lot of time pounding away at management to do something with virtualization.î Frustrated, Crosby left Intel and reconnected with Pratt, determined to start a company focused on the technology. The two then met up with Nick Gault, who was at the time president and CEO of a network management company. Gault knew investors at Sevin Rosen Funds, a U.S.-based venture capital firm to whom he introduced the Xen team. Pratt then met Marshe Barr, an open source veteran who had relationships with Wall Street firms that wanted a technology like Xen. Gault became the companyís first CEO and Barr the CTO. Three senior engineers rounded out the founding team of seven people.
Funding
XenSourceís talent and technology attracted some of the most respected venture capital firms in the industry, including ---- . In addition to Sevin Rosen, the company received funding from Kleiner Perkins Caufield & Byers, Accel Partners, NEA and Ignition Venture. Before the end of 2006, the company had raised a total of $41 million, placing its valuation at$137 million. (See Exhibit 9 for a summary of the companyís funding history.)
10 ìX86 is the generic name for the series of Intel microprocessor families.î (Source: Stanford GSB case E-149 ìVMWareî.)
11 University of Cambridge, SRG, http://www.cl.cam.ac.uk/research/srg/netos/xen/ (May 24, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A) p. 6
Peter Levine was a general partner at the venture firm Mayfield Fund, when XenSource began raising its first round of investment in late 2004. As he recounted:
I remember when the deal first came through. It came and went so fast that we missed it. It caught my eye, though, because I had a lot of background in virtualization from my days at Veritas, which had made an investment in VMware in the late 1990s. VMware at the time was a tiny, little, nothing company. I was on the advisory board and developed heavy technical knowledge in the space. It was square up my alley.
When Mayfield did not participate in the XenSource opportunity, Levine more or less forgot about it. A few months later, at the end of 2005, Levineís friend and former colleague from Veritas, Frank Artale, called him. Artale had been consulting at XenSource and thought that Levine was uniquely qualified to help the start-up. Levine performed some due diligence on XenSource, and realized that it was a company in trouble, but also that he might be able to help.
Focus and Change
In late 2005, XenSource had an immature code base and large, demanding customers. As Crosby remembered it, ìWe realized that, while we had enjoyed some success with building pilot projects for Fortune 10 companies, they were not going to bet their entire infrastructures on our brand-new product.î With Gault still at the helm, the young company had bitten off a lot in terms of product scope, building the hypervisor as well as the management tools that would go along with it. Individually, those development projects were challenging, but collectively they were overwhelming the new company and slowing down the teamís time-to-market, and making the company ìschizophrenic,î as Crosby said. The board of directors felt that XenSource needed focus, which Levine echoed when he took a closer look:
They were trying to boil the ocean, having overcommitted as a result of being really hot and believing their own hype. Unless they picked one of their products to nail firstóand I felt it should be the hypervisor⎯they would be mediocre on everything and have no competitive differentiation. XenSource needed to be a platform company and then worry about developing management tools for their infrastructure later. They also needed a tangible goal, and in my view chasing VMwareís taillights was it.
Crosby, Pratt and the board agreed with Levineís assessment and saw his ambitious target as the rallying point that XenSource desperately needed. However, pursuing such a course necessitated forging partnerships with large OEMs and VARs. While the indirect sales route played to Levineís proven strengths from his days at Veritas, it was a big bet for the tiny company and would put enormous pressure on each deal, because almost every target partner would lean in one of two waysócooperation or competition. Further, the success of future deals could largely depend on whether XenSource could lock in early agreements. The partnerships would involve embedding XenSourceís technology into other software and hardware vendors, such as Red Hat,
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A) p. 7 Novell, NEC, HP and IBM, any of which could pass on XenSourceís offer and then compete
with them or agree and insist on taxing demands or terms.
The new focus would also require changes to the executive team. Gault had not been the one to articulate this new vision, and it was a significant departure from his experience and skill set. As a result, Gault left XenSource in February 2006, and Levine assumed the CEO role (Exhibit 10).
MARKET OPPORTUNITY AND RISKS
Levine was attracted to the virtualization space for a number of reasonsóit was large, growing and dominated by a single company that was raking in revenue with very little direct competition. Almost 10 years after its founding, VMware was considered ìthe fastest-growing large software firm on the planet,î with a growth rate that was actually increasing as it got bigger. VMware was expected to generate well over $1 billion of revenue in 2007, up from$704 million in 2006, $387 million in 2005, and$219 million in 2004. Industry analysts and experts were especially excited by the fact that, as of 2006, only about 5 percent of shipping servers were virtualized.12 So, even with VMwareís impressive revenue numbers, XenSource executives estimated that VMware owned 80 percent of the tiny 5 percent install base, a base which was expected to grow to between 20-40 percent of servers within three to five years based on the estimates of research analysts such as Gartner and Thomas Weisel Partners.13 Microsoft controlled 10-12 percent of the market, and the remaining percentage points were split across XenSource, Virtual Iron and a few others.
Virtualization looked promising in terms of macro-economics and hardware trends as well. The rising cost of oil in 2006 and 2007 meant that companies needed to use less power, and therefore, fewer machines to do the same or more work. Fewer machines also meant fewer endpoints to administer and manage, so a solution that could reduce the number of servers from five to one as well as achieving much higher rates of utility from each node was doubly appealing to CIOs. Companies with huge data centers, like Yahoo and Google, were already using virtualization while simultaneously making large investments in new locations that could leverage lower costs.
Virtualization Disrupts the Balance of Power
As attractive as virtualization was to many, its role in the computing ecosystem bled into previously well-defined territory. In 2005, Computer Associates honored VMware with its Riding the Crest Award for ìbest-selling operating system and utility productî for the VMware Workstation, a move which highlighted the threat posed to OSsóVMs were figuratively and literally coming between hardware and software makers. One analyst downgraded Red Hat as a
12 The vast majority of virtualized servers were deployed in test and development environments, as opposed to production or ìliveî environments. Test and development servers accounted for 20-25 percent of all servers in 2007. 13 A report by Thomas Weisel Partners predicted 20 percent growth, ìwhich would equate to an addressable market of $3 billion.î Tim Klasell and Jeffrey Peck, ìThe Rise of the Virtual Machine and the Real Impact It Will Have: A White Paper on the Infrastructure Software Industry,î February 3, 2006, http://www.xensource.com/files/VM_Whitepaper_TWP.pdf, (June 21, 2007). Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 8 direct result of VMwareís increased success, ìbecause companies that previously cut costs by using cheaper Linux servers are now using virtualization software to run the same programs.î14 Prior to the VM disruption, for years, hardware and software makers had worked together, either in harmony or begrudgingly, to get their products to market. A few hard and fast ìrulesî had emerged, rules which virtualization threatened to rewrite. The Nature of Development Regardless of the innovation and speed of the latest chipset, the operating system would ship whenever the software vendor determined. Often a major release of an OS took years, and the requirements list was sealed at the very beginning before any coding began. Therefore, many of the improvements made by the chipmakers could not be taken advantage of even with the ìlatestî versions of software, which created a tension between OS and chip-makers. For example, Microsoft and Intel had a powerful partnership, but they were not exempt from this tension. Levine felt that XenSource could take advantage of the situation by courting Intel with the promise that Xen could quickly and directly leverage the chip-makerís innovations, because VMs rode on the ìbare metalî of the hardware, and therefore did not have to wait for the ìslow and clumsyî Microsoft OS releases. While XenSource had worked with Intel closely from the beginning in terms of Xenís product development and performance optimization, would the start-up need to make more overt gestures from a business strategy perspective? Did it make sense to do the same with AMD? Human Nature Complicating these corporate relationships were the personalities behind them. Some of the original pioneers and architects of the ecosystem in which XenSource found itself were still at the helms (or within close reach) of the companies, and many were fiercely competitive, if not openly antagonistic. A famous example was Scott McNealy of Sun and Bill Gates of Microsoft. On numerous occasions McNealy publicly disparaged Microsoft as a monopolistic threat to Sun and other companies, ìThe only thing that I'd rather own than Windows is English,î said McNealy, ìbecause then I could charge you two hundred and forty-nine dollars for the right to speak it.î15 The personalities frequently influenced whether partnership agreements were struck or not. While Levine knew how to build channels, he was not a self-described diplomat, and he wondered if he was entering a new realm of corporate deal-making. (Un)natural Partnerships To combat real and perceived threats from various players, companies showed fickle loyalties to each other. For example, a complicated series of events involving Apple, Intel, Dell and AMD resulted in the May 2006 announcement that Dell, a long-time Intel loyalist, would use AMD chips in some of its multi-processor servers.16 Eric Lundquist, the editor in chief of eWEEK, was cynical of the partnership: 14 ìAnalyst: VMware digging into Red Hat's market share,î Triangle Business Journal, December 10, 2007, http://www.bizjournals.com/triangle/stories/2007/12/10/daily1.html?surround=lfn, (December 17, 2007). 15 ìScott McNealy's Top Ten Barbs vs. Microsoft: Should Old Acquaintance Be Forgot,î http://www.sys- con.tv/read/44347.htm, (December 5, 2007). 16 http://www.theregister.co.uk/2006/05/18/dell_picks_amd/ (December 5, 2007). Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 9 It would be nice if vendors made product and component decisions based on customer wishes and needsÖ, but it isn't the reality under which the technology industry Ö operates. Supplier decisions are often based less on technology than on co-marketing agreements, product placement agreements and exclusivity deals. These agreements are usually invisible. Vendors would like you to think they are the champions of users' needs. The invisible usually doesn't become visible unless one supplier approaches monopoly status and the resulting courtroom fight reveals previously private agreements, which become public as part of legal proceedings. 17 Another example was the surprising Linux and Intel partnership forged in 2004, which had Linux adopting Intelís x86 technology, because it was inexpensive and widely used, as opposed to technically superior. No thanks to ìLintel,î the once cozy relationship between Microsoft and Intel (Wintel) had begun to sour by the time XenSource came along. Wintel felt additional strain when Microsoft began using other chip makers for its pocket PCs and games consoles. The Unique Case of Microsoft Among all of the players in the landscape, the most important one was undoubtedly Microsoft. In 2003, the software giant made an investment in virtualization through a small acquisition, yet the jury was still out on its official internal stance toward the technology. In conjunction with its next major release, Vista, would Microsoft make a concerted effort to charge headfirst into the virtualization market or was it too afraid that VMware et al would undermine the value of its operating system franchise and therefore do everything in its power to diminish the market size and crush the players therein? (See Exhibit 2.) THE STRATEGY The Business Model XenSource had decided to focus on the hypervisor and deliver it through OEMs and other partners, but Levine still had to sort out the details of the strategy and business model. Crosby remembered, ìWe had to make Xen work in our system and deliver it to our customers, but we were also giving it to anyone who wanted to bundle it into their products.î XenSource decided to give the basic Xen hypervisor away for free while building capabilities on top (e.g., additional concurrent VMs, CPUs and physical RAM), which could be turned on with a simple license key. The enhanced versions could, in turn, be bundled with OSs and resold through various channels, such as Red Hat, Novell, HP and IBM. XenSource would add management tools to their offering later on, as the company gained credibility and a foothold in the industry. Taking from the Xen Pool XenSource described the Xen hypervisor as raw virtualization technology or the ìengine,î merely a code base that needed to be combined with other components before a system 17 Eric Lundquist, ìMactel Inspires Other Strange Bedfellows,î eWEEK, http://www.eweek.com/article2/0,1895,1913634,00.asp January 23, 2006, (December 5, 2007). Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 10 administrator could create, run or manage a virtual machine. Whatever ìcarî a developer or company built around that Xen engine was up to them. In addition to Red Hat and Novell, XenSource was also building its own car around the Xen engine18 (Exhibit 11). At least 75 percent of the code that was checked in to the Xen pool was developed by XenSource employees. The remainder came from the developer community and the companies that used Xen. While the pool was an open source environment, XenSource maintained some control in that its engineers had the most intimate knowledge of the code and could make it difficult for other developers to make changes to the common code base. XenSource engineers knew which features were customer-ready and which still needed further testing. The fact that anyone else who used the Xen code did not have as good an understanding came back to bite XenSource in at least two ways. The first arose if Red Hat, for example, wanted to take a risk on unstable code, thinking that they had a lot of smart engineers within their own company who could make it work. If Red Hat then released that version of Xen and it crashed, customers would complain to Red Hat, who would in turn blame Xen, which reflected poorly on XenSource. The second challenge would come when Novell, for example, took a chance on a new feature and released it in its own product, a feature which XenSource had not yet incorporated into its latest release. Customers in the Xen community would question XenSource and demand that they play catch-up. Freely admitting the challenges with the model, Levine said: This whole open source thing is a means to end, not an end in and of itself. We are not about free hippy love over here. We are all about making money for hard work and great products. The open source challenges are many, and they frustrate me, but there is no way we would be where we are today in terms of product quality and name brand recognition without it. Licenses Open source created its own unique challenges with regard to licensing too. A developer could access the Xen hypervisor under the terms of the GPL (General Purpose License),19 which permitted anyone ìto copy, modify and redistribute any portion of the software,î but licensees were not permitted to use the Xen (or any XenSource) trademark. Exceptions to this restriction included ìvendors in the Xen ecosystem,î who were selling Official Versions of the Xen hypervisor and agreed to the terms of the Xen trademark license. The license to use the ìIncludes Xenî logo was royalty-free. This decision was seen as critical to the success of the company, given the importance of brand in open source. Products and Pricing Once management agreed how to license and offer the product, they needed to address its pricing. The next major release of the hypervisor (v. 4.0) was scheduled to ship in August, and 18 This was different from the arrangement with Microsoft, which refused to use the open source ìGeneral Public Licenseî (GPL) code. XenSource and Microsoft were working together to build a hypervisor for Microsoft which leveraged Xen, but was developed from scratch. 19 The GPL was created by the GNU Project, http://www.gnu.org/copyleft/gpl.html (June 5, 2007). Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 11 Levine claimed it would have 75 percent of the functionality that VMware offered, but with significantly better performance.20 It was a key bet, which Levine explained further: To get a product out the door, we need to promote performance, which is inherent in our design. Adding new features would take time. We may never catch up with VMware in terms of product functionality, so we need to have rocket-like performance, making the decision to buy from us a no-brainer. We donít have very many customers and very few using it in production,21 so weíre eager to generate feedback from real customer experiences as soon as possible. To accomplish that goal, the company felt that it needed to seed the market with its technology quickly if it was to have a fighting chance against VMware. The logic followed, therefore, that the price had to be really low, a fraction of VMwareís. Artale had estimated somewhere between one-third and one-half of the incumbentís price for an ESX Server seat, the companyís high-end data center product, against which XenSource planned to compete directly (Exhibit 12). While there were plenty of enterprise software veterans within XenSource who would have preferred a higher price point, the majority of the team liked the ubiquity option, which segmented the market into three categories: (1) developers who wanted to ìkick the tiresî of the XenSource offering, (2) Windows-only IT professionals, and (3) enterprises and data centers looking to consolidate their Windows and Linux deployments. People in the first category, who wanted the starter package, could download XenExpressô for free. Those interested in Windows-only deployments could buy XenServerô for$99. Finally, enterprises that wanted to manage multiple operating systems on multiple servers could buy XenEnterpriseô for $750. Artale said, ìThere is not a lot of magic in how we came up with those prices. We just want to keep the cost down to encourage trial and adoption of our technology.î Partnerships The Microsoft Deal Roughly 70 percent of servers in production worldwide ran Windows, and so to quickly tap into that user base, XenSource needed a partnership with Microsoft. Even though the software giant was openly and actively building its own hypervisor, which would compete with XenSourceís, Levine and his team felt they had a great head start and some of the best engineers in the field. In July 2006, the two companies signed a collaborative development agreement to produce ìinteroperable virtualization solutions between Xen-enabled Linux guest operating systems and the new Microsoft hypervisor-based Windows Server virtualization.î Due to Microsoftís self- imposed rules to not license open source code, this agreement required XenSource to provide Microsoft with its own flavor of Xen (i.e., not taken from the generic Xen pool). Such a 20 Performance, but not feature set parity. XenSource executives openly acknowledged that VMware outpaced them with respect to number of features, by a ratio of about 25 to 4. XenSource maintained, however, that customers only used 4 or 5 of the 25 features from VMware. The start-up was, therefore, in no rush to add 20 more features into its own offering. 21 In other words, the product was being actively used, not simply deployed in testing environments. Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 12 partnership was a feat for a company of XenSourceís size, especially given Microsoftís preference for ìnot putting anything in writingî after its antitrust lawsuit settlement in 2004. The irony that XenSource was working with Microsoft on a directly competitive product was not lost on the start-up. Levine justified the near-term relationship with the Redmond-based behemoth by pointing to his belief in XenSourceís long-term strategy of building management tools once hypervisors became commodities in the marketplace: Ultimately our hypervisor, at all levels, will be free. And eventually VMware will do the same thing. Microsoft will get theirs to market at some point for a cheap price that might as well be free. While it wonít be better, because smaller, more nimble companies like ours will always innovate and have better technologies, there are so many Microsoft loyalists out there that we will have to support the their technology. So the logical next phase of this company has to be building management tools that go on top of everyoneís hypervisors. Crosby did not agree completely, saying, ìThe hypervisor is really tough to build. Weíve had the best people in the world working on this for years in Cambridge.22 Management tools are easy, though: all you need is 10 guys in India or Russia to develop those in a matter of months.î Levine pushed back by pointing to XenSourceís engineering talent, saying that Crosby and Prattís teams ìcould build superior management tools that were optimized for the Xen platform. Owning our own hypervisor will enable us to create the best integrated product with those tools.î Whoís Next? To achieve its ubiquity goal, XenSource needed to sign more OEM agreements with systems companies both domestically and abroad. Levine was in discussions with NEC and HP, and while he was tight-lipped about additional companies they were targeting, he was clear that they were large, fiercely competitive with Microsoft, and looking for any area to launch attacks on the behemoth. According to marketing vice president John Bara, ìIn this business, the enemy of my enemy is my friend, and for a lot of people, including us and VMware, Microsoft is the enemy.î Still, pursuing that path risked angering the giantóand being squashed. ìIs there an argument for exclusivity with Microsoft and playing it safe?î Levine asked. ìSure, but that doesnít sound like very much fun to me.î COMPETITION Microsoft In 2002, Microsoft had been working with VMware as a partner, but soon realized the emerging strategic importance of virtualization. Microsoft made an offer to buy VMware but was unsuccessful in closing the deal due to disagreements about the price.23 Instead, the following year Microsoft acquired Connectix, a much smaller virtualization competitor, for a markedly 22 XenSource maintained its core Xen development at an office in Cambridge, England. It also had a satellite office in Redmond, Washington. Company headquarters were in Palo Alto, California. 23 VMware was estimated to have about$50 million in revenue at the time.
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A) p. 13
lower price. Microsoft then repackaged the Connectix product and called it Windows Virtual Server, which allowed multiple OS images, including Windows, Novellís Netware, and Linux, on a single server. Priced at $199, the Enterprise Edition (which supported more than four processors) was seen as an ìOKî product by industry analysts.24 Microsoft had acquired the technology for the purpose of incorporating it into its Windows Server product line, not to offer as a stand-alone product. However, their development project did not seem to be going well, and Microsoft said that it would need to cut key features as well as delay the release of its first public beta of Windows Server virtualization technology, code- named Viridian, until the second half of 2007.25 The hypervisor product was supposed to be a key feature of Vista, Microsoftís major release of its Windows OS and the successor to XP. The company went so far as to commit to shipping Viridian within 180 days of shipping Vista. The Microsoft delay was a phenomenally positive piece of news for XenSource, buying the company untold time to seed the market with its technology. However, Microsoftís delay also benefited VMware, further solidifying its products in its customersí computing infrastructures. Virtual Iron Founded in 2003, Virtual Iron originally tried to develop its own hypervisor technology that would allow ìa single virtual server to run across multiple physical servers.î Various factors, including ìthe pace of processor development and the progress of clustering technologies,î made this approach somewhat obsolete. After two years of development the company turned to Xen. The result, Virtual Iron 3.X, was a product similar to XenSourceís XenEnterprise in that it used the open source Xen to compete with VMware and at a much lower price point,$499 per socket.26 XenSource believed that the Xen brand and connection to the company gave it a clear competitive advantage over Virtual Iron.
VMware
Founded in 1998, VMware sought ìto bring mainframe-class virtual machine technology to industry-standard computers.î27 In 1999 the company introduced its first product, VMware Workstation, which allowed desktop users to run multiple OSs on a single workstation.28 Two years later, it began adding products for servers, including the GSX Server29 and ESX Server. The server software was suited for quick, efficient and more-affordable application deployment for enterprise environments and server hosting facilities (Exhibit 13).
24 Jason Brooks, ìVirtual Server 2005 R2 Is Solid Solution for Windows,î eWeek, http://www.eweek.com/article2/0,1895,1927284,00.asp, February 20, 2006, (June 5, 2007).
25 John Fontana, ìMicrosoftís Server Virtualization Delay,î Network World, http://www.networkworld.com/news/2007/041207-microsoft-delays-release-of-key.html, April 12, 2007, (June 5, 2007).
26 ìA socket is the physical component on a motherboard in which one or more cores reside.î Source: Virtual Iron website, http://virtualiron.com/products/purchase.cfm#socket (June 5, 2007).
27 Stanford GSB Case E-149: ìVMware.î
28 Ibid.
29 GSX Server was renamed VMware Server and became a free product in 2006, http://www.vmware.com/company/news/releases/server_beta.html, (July 16, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A) p. 14
Most of VMwareís products required a host OS to function. ìThe product acts as both a virtual machine monitor [interacting directly with the hardware, such as Intel or AMD chips], and as an application that runs on top of the host operating system.î30 The dual purpose yielded two distinct advantages. First, having a direct relationship with the hardware enabled optimization of performance, portability, ease of use and other benefits. Second, running on top of the OS ìfrees the VMM from having to deal with the large number of devices available on the PCs.î31
A key difference between the technologies of VMware and XenSource was that VMwareís approach ìcompletely simulates a machine, which theoretically allows any operating system to run unmodified on a virtual machine.î32 XenSource employed ëparavirtualization,í which did not simulate the entire machine, which implied faster performance but required that the OS be modified to run (Exhibit 4b), a modification which was made by the OS vendors before they shipped the software to the end user. Unlike VMwareís other products, ESX was a bare metal product, like XenSourceís, but it did not use paravirtualization. XenSource executives speculated that the ESX product was responsible for about 75 percent of VMwareís licensing revenue.
While VMware had not formally acknowledged XenSource as a viable contender, analysts and reporters suggested that when VMware entered the ìfree productî market in February 2006 it was implicit admission of that. Renaming its mid-range GSX Server the ëVMware Server,í the company described it as ìa free new entry-level hosted virtualization product for Linux and Windows servers.î33 XenSource believed that the VMware Server product relied on the same original GSX architecture, and that roughly 65 percent of VMwareís install base was still using the GSX / VMware Server. Because GSX was a hosted architecture, it required the OS first to be installed. According to Levine, ìUpgrading from their free product to ESX is a substantial change for the customer. By comparison, our architecture is the same from Express to Enterprise.î
Pricing
VMware had a family of products that ranged from free to very expensive. The enterprise product, ESX Server, was priced between $4,000 and$8,000 per server license depending on the level of functionality. Before becoming the free VMware Server product, GSX Server had been priced at $1,400 for dual-processor servers and$2,800 for more powerful machines.
Sales
VMware was reported to have 1,100 employees in its sales and marketing functions. XenSource believed VMware had at least 160 direct sales representatives and the same number of inside sales people and field engineers. Several years after its founding, the larger company decided to complement its direct sales team with resellers, reaching 4,000 partners. VMwareís sales force was achieving great success. The companyís sales representatives were assigned average quotas
30 Amit Singh, ìAn Introduction to Virtualization: Itís Hot Yet Again,î kernalthread.com, http://www.kernelthread.com/publications/virtualization/ (May 29, 2007).
31 Ibid.
32 Stephen Shankland, ìXen Lures Big-Name Endorsements,î http://news.com.com/Xen+lures+big- name+endorsements/2100-7344_3-5581484.html, February 18, 2005 (May 31, 2007).
33 http://www.vmware.com/company/news/releases/server_beta.html, (July 16, 2007). Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013

XenSource E-267 (A) p. 15
of $6 million per year, more than double the industry average, and it was rumored that 90 percent of VMwareís sales force was meeting or exceeding their quota, something unheard-of in the enterprise software industry. Big Install Base VMware boasted a large installed customer base, including 100 percent of the Fortune 100 and over 84 percent of the Fortune 1,000. Its server solutions were deployed in over 20,000 organizations worldwide. According to an independent survey34 of over 800 IT decision-makers, with a heavy skew towards large enterprises (1,000 to 10,000 employees), 40 percent stated that they had either deployed virtualization technologies or planned to do so within the next 12 to 18 months. ìSpecific to VMware, 58 percent of those surveyed had or will deploy VMware GSX server, and 50 percent would look to utilize ESX server.î35 Bought by EMC In the second half of 2003, EMC, a$6.2 billion information storage and management hardware, software, and services company, went on a buying spree. After acquiring Legato and Documentum, for $1.3 billion and$1.7 billion respectively, EMC announced its plans to buy VMware in December for $635 million. Partners and Position in the Ecosystem As a subsidiary of EMC, VMware was often considered within the context of the enterprise storage industry. More specifically, analysts compared EMC against the big enterprise hardware names: Dell, HP, IBM, Sun, NetApp, and Hitachi. In addition to their competitive relationships with EMC, most of those vendors also had (pre-existing) relationships with VMwareómainly interoperability and/or reseller agreements. VMwareís position against Microsoft was complicated, because while the two were somewhat recently directly competitive in terms of delivering virtualization services, VMware had to service a largely Windows-dependent client base, which necessitated some warmth between the companiesí engineers. VMware also had a nuanced relationship with Intel. Seeing an x86 virtualization vendor36 succeed was good for Intel, because as its chips became more powerful, it could help the virtualization market take off and, in turn, Intel could gain market share from chips and servers from IBM, HP and Sun. Intel also indicated that it was interested in owning part of VMware when the latter went public. XenSource felt that it could compete with VMware in a multitude of ways, including getting its hypervisor prepackaged on many of the big hardware vendorsí servers, including those of IBM, HP and Dell. VARs and systems integrators would respond favorably, XenSource argued, because then those companies and their customers would have a choice vis-‡-vis VMware. 34 Aaron C. Rakers, CFA, ìEMC ñ Increased CQ1í07 Comfort; Maintain Positive Long-Term Thesis, IBM P6 Transition Challenges = Possible Emulex Headwind?, New FCoE Standard a Positive, Micron Highlights Vista- Related PC Dynamics, & Much MoreÖ,î A.G. Edwards & Sons, Inc., April 9, 2007. 35 Aaron C. Rakers, op. cit. 36 x86 indicated Intel or AMD, as opposed to chips made by IBM, HP or Sun (SPARC). Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 16 XenSource also had to consider what relationships it should have with Intel and/or AMD, and how its partnerships with hardware and software vendors might play a role there. VMware IPO In late 2006, the EMC subsidiary hinted at plans to issue 10 percent of its stock on the open market. The SEC filing reported information that EMC had previously kept under wraps, such as profitability, annual revenue and net income figures. Both EMC President Joe Tucci and industry analysts talked about ìunlocking the valueî that the EMC-VMware combination had created since 2003. Tucci believed the VMware division could be worth$6 billion, which, if true, meant the EMC investment had grown 10 times its purchase price in value in three years. Industry analysts were even more bullishóestimating the IPO at $10 billion. XenSource investor and director Kevin Compton viewed the VMware IPO as great for XenSource and perhaps problematic for VMware. ìGoing public is a terrible strategy with a lower-cost competitor right on their heels. They will be in no position to drop their pricesóa move they would have been forced to make as a private company to respond to XenSource, but as a public company it will be a real challenge for them to execute on that.î Compton continued, ìItís easy not to like VMware. With their monopoly position and predatory pricing, we are not concerned about companies rallying around us to topple them.î EARLY RESULTS Channel Conflict? XenSourceís sales team was eager to rise to the challenge and sign up resellers and distributors around the world, but the companyís size and low price point would have to be addressed. According to VP of sales for the Americas region, Rafael Santini: We already get a lot of attention, because people have heard of Xen and XenSource, but that will only get us in the door. At a fraction of the price of VMware, what will be the incentive for resellers to push our products? To overcome that objection, our job will be to convince them that they need to be competent across all technologies and products out thereóthat they need to be ëvirtualization experts,í in order to be credible to their customers. While Santiniís team was busy signing up partners in North America, Artale was working on the large OEMs. One of the challenges with being a small start-up was leverage, or lack thereof, with big, established companies. Levine commented: In unknown, land-grab environments like this one, whatís fair and who should get what margins are guesswork. Signing a major partner like NEC or HP is tricky, and they are used to huge margins, so Artale might start out his negotiation with them by giving away 40 percent or moreóhe has to. Meanwhile, Santini is down the street with a smaller partner, and heís holding firm with them at 20 percent. Well, people talk of course, and discrepancies like that can kill a partnership fast. Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 17 Making Elephants Dance For the first half of 2007, XenSource was on track to acquire about 500 unique customers. The ratio of server licenses to customers was low, however, at best 2 to 1, given the trial nature of the software and the developers testing it. The sales team expected to generate a few million dollars by yearís end, assuming the majority of the installations continued to be XenEnterpriseô, the most expensive product. Santini was optimistic and actually gave credit to VMware, ìThanks to them, the market is educated about virtualization. We donít have to explain to prospects what it is and why they need it. We can focus on the job of explaining why we are better, which is better for us, because we are.î Compton agreed that the market was nicely primed for XenSource. He felt that the challenge was really more how they priced the product as opposed to the price itself: ìCompanies are realizing that there is no way we could charge the full value that they will realize from our product. Weíre telling them that weíll deliver the value of 10 servers and 10 OSs in one. While they agree, theyíre not going to pay us the equivalent of those hardware and software costs. So, weíre wondering what we should take away from that?î Levine had communicated to the team that the first half of 2007 would be all about charting user adoption of XenSourceís products. In the second half of the year, he would start measuring revenue growth. Although everyone expected revenue to be modest, the energy and buzz around the company was palpable. Levine had already turned away suitors who wanted to buy the company, and talks of an IPO were underway. Company executives thought that a 24-month horizon for such an event was likely. While Levine was comfortable with a certain level of ambiguity, as he closed in on the next two key industry partner agreements he was starting to feel more uncomfortable with ìmaking the elephants dance.î He was eager to be out in front of the market and not let his partners dictate XenSourceís future and next moves. But in an environment where corporate loyalties changed faster than versions of software, the question was⎯how? Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 18 Exhibit 1 Top 5 Corporate Family, Worldwide Server Systems Factory Revenue, Second Quarter of 2007 (in millions of dollars)37 Vendor 2Q07 Revenue Market Share 2Q06 Revenue Market Share 2Q07/2Q06 Revenue Growth IBM $4,069
31%
$3,824 30.9% 6.4% HP $3,707
28.2%
$3,434 27.8% 8.0% Sun $1,711
13.0%
$1,620 13.1% 5.6% Dell $1,526
11.6%
$1,270 10.3% 20.2% Fujitsu/Fujitsu Siemens $542
4.1%
$554 4.5% -2.3% Others $1,582
12.0%
$1,659 13.4% -4.7% All Vendors $13,137
100%
$12,361 100% 6.3% Exhibit 2 Microsoftís Revenue by Business Segment (in millions of dollars) Year Ended June 30 2007 2006 2005 Client $14, 812
$13,040 $11,901
Server and Tools
$11,182 $9,680
$8,314 Online Services $2,474
$2,306 $2,353
Microsoft Business
$16,381 $14,516
$13,298 Entertainment $6,132
$4,808 $3,485
Unallocated
$141 -$68
$437 Consolidated $51,122
$44,282 $39,788
37 ìServer Market Accelerates as New Workloads and a Strong Refresh Cycle Drive Server Demand in the Enterprise, According to IDC,î http://www.idc.com/getdoc.jsp?containerId=prUS20840807, August 23, 2007 (December 5, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 19
Exhibit 3 Software Wars38
38 Copyright 2006 ñ Steven Hilton
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 20
Exhibit 4a39
Virtualized and Non-Virtualized Environments
Exhibit 4b40
Para-Virtualized and Hosted or ìPureî Virtualized Environments
39 Tim Klasell and Jeffrey Peck, ìThe Rise of the Virtual Machine and the Real Impact It Will Have: A White Paper on the Infrastructure Software Industry,î Thomas Weisel Partners LLC, February 3, 2006.
40 Ibid.
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 21
Exhibit 5
Abridged Open Source Definition (OSD)41
1. Free Redistribution: the software can be freely given away or sold. (This was intended to expand sharing and use of the software on a legal basis.)
2. Source Code: the source code must either be included or freely obtainable. (Without source code, making changes or modifications can be impossible.)
3. Derived Works: redistribution of modifications must be allowed. (To allow legal sharing and to permit new features or repairs.)
4. Integrity of the Author's Source Code: licenses may require that modifications are redistributed only as patches.
5. No Discrimination Against Persons or Groups: no one can be locked out.
6. No Discrimination Against Fields of Endeavor: commercial users cannot be excluded.
7. Distribution of License: the rights attached to the program must apply to all to whom the program is redistributed without the need for execution of an additional license by those parties.
8. License Must Not Be Specific to a Product: the program cannot be licensed only as part of a larger distribution.
9. License Must Not Restrict Other Software: the license cannot insist that any other software it is distributed with must also be open source.
10. License Must Be Technology-Neutral: no click-wrap licenses or other medium-specific ways of accepting the license must be required.
41 Source: http://www.opensource.org/docs/definition.php (May 31, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013

XenSource E-267 (A)
p. 22
Exhibit 6
VC Investment in Open Source Since 2000 42
Year
$Raised Number of Open Source Start-ups 2006 $481 M
48
2005
$306 M 41 2004 $298 M
36
42 Graph created by Larry Augustine, XenSource board member and former chairman of VA Software (also known as VA Linux), http://lmaugustin.typepad.com/lma/open_source/index.html (June 19, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 23
Exhibit 7
Generic Business Risks of Open Source43
1. The development community does not drive core product development, which means that the company will still need to be in charge of most of the engineering efforts. Further, few customers actually modify source code, so a company should not depend on that feedback loop for major enhancements.
2. A company needs resources to ìmanageî the community, which otherwise can become a burden to customer support and engineering.
3. Companies should expect that competitors will inevitably read through their code. According to Fountain, ìOthers will benefit from the ideas, if not their embodiment.î
4. The companyís product road map, traditionally held close to the vest, generally needs to be disclosed when open source is a part of it. This ensures ongoing dialogue with the developer community, a level of transparency that is commonly accepted as necessary.
5. Having open source as a key piece of a companyís product strategy may limit the number of would-be acquirers.
6. A competitor might 'fork' a companyís project and take its code base to compete against it. Open source product companies generally derive their market leverage from (explicit or implicit) control of one or more projects, which are, by definition, available under an open source license, which in turn opens the door for another start-up to fork the source project. The new derivative could capture a following of some of the old community and/or build an entirely new community. The risk can be partially mitigated through the companyís licensing choice*, such as a GPL. However, such a choice can also constrain the business model options.
* For example, if a company made its project available under GPL (General Public License), then any would-be competitor would need to also make its linked source code available under GPL, a licensing scheme which often constrains the business model options. As a result, a company has hampered any future competition and partially mitigated the risk.
Even with this restriction another company could still fork the project and build its business model within the constraints of the license. For example, a company may elect to use a modified MPL (Mozilla Public License) instead, which allows code under the MPL to be combined in a program with proprietary files.44 Such a license could have a requirement to include a "Powered by XYZ" logo and link to the companyís website in derivative works. This would increase the friction for a competitor and make it less likely (though not impossible) to fork a companyís code base.
43 Tom Fountain, ìEvaluating Open Source: License & Business Models,î Mayfield Fund, May 31, 2006. 44 Wikpedia.org (July 9, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 24
Peter Levine, President & CEO
Exhibit 8 Executive Team Bios45
Peter Levine has over twenty years experience in the software industry, having held both venture capital and operating positions in engineering, marketing, sales, and executive management. Prior to XenSourceô, Levine was a Managing Director at the Mayfield Fund, where he spent three years investing in a variety of software companies. His notable investments include Mendocino Software, Centrify, Zenprise, TrueDemand, OuterBay (recently acquired by HP), and Actona, (acquired by Cisco). Before joining Mayfield, Levine was an early employee of VERITAS software and during his eleven year tenure with the company, helped to grow the organization from no revenue to over $1.5B and twenty employees to over 6,000. When he left VERITAS in 2001, Levine served as the Executive Vice President for Strategic and Platform Operations. Levine had responsibility for world-wide marketing, OEM sales, business development (M&A, Venture fund), and several new product and platform development groups. Prior to VERITAS, Levine was a software engineer at MITís Project Athena, where he worked on a variety of distributed computing projects. Levine holds a BS degree in engineering from Boston University and attended the Sloan School of Management at MIT. Ian Pratt, Xenô Project Lead, Founder and Office of the President Ian Pratt is the leader and chief architect of the Xen project, which he founded in 2001 with the aim of making virtualization ubiquitous on scale-out hardware. Pratt has played a key role in both the architecture of Xen and formation of industry partnerships that led to the emergence of Xen as the open source virtualization technology. Pratt is a member of senior faculty at the Computer Laboratory of Cambridge University, UK, where he has led Systems Research for 7 years. He holds a PhD in Computer Science, and was elected a Fellow of Kings College in 1996. Pratt was a founder of Nemesys Research, acquired by FORE Systems, and has consulted widely in the technology industry. Simon Crosby, CTO and Founder An industry evangelist for the Xenô open source hypervisor, Simon Crosby is CTO of XenSource. In this position, he is responsible for XenEnterprise R&D, technology leadership and product management, and maintaining a close affiliation to the Xen project run by Ian Pratt, the founder of XenSource. Prior to XenSource, Crosby was a principal engineer at Intel where he led strategic research in distributed autonomic computing, platform security and trust. Before Intel, Crosby founded CPlane Inc., a network optimization software vendor, and held a variety of executive roles while there including president & CEO, chairman and CTO. Prior to CPlane, Crosby was a tenured faculty member at the University of Cambridge, UK, where he led research on network performance and control, and multimedia operating systems. He is author of over 35 research papers and patents on a number of datacenter and networking topics including security, network and server virtualization, resource optimization and performance. Crosby is a frequent speaker on the topic of enterprise-grade virtualization with open source technologies, and has most recently been a presenter at such well-known industry events as LinuxWorld, Interop and the Server Blade Summit. Frank Artale, VP of Business Development Frank Artale joins XenSource from Accel Partners, where he was most recently a member of the venture development team and worked as an advisor to such companies as Mendocino Software and Centrify Software. Prior to Accel, Artale was a founder and CEO of Consera Software, a management software company, which was acquired by HP in 2004. Before founding Consera, Artale was vice president of the Windows Solutions Group for VERITAS Software, where he oversaw overall product strategy of the Windows targeted solutions. Prior to this position, he was a general manager in the Windows 2000 group at Microsoft Corporation, one of the many and varied product management and sales positions he held during his nine-year career with the company. 45 Source: XenSource company website: http://www.xensource.com/company/management.html (June 4, 2007). Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013  XenSource E-267 (A) p. 25 John Bara, VP of Marketing Exhibit 8 (continued) Executive Team Bios John Bara joined XenSource from Interwoven, where he served as SVP Marketing. Previously, he was VP Marketing at Genesys Telecommunications Labs, Inc. At Intel, he was a management team member for the Pentium group, financial controller for the microprocessor group, and held responsibility for large OEM customers including HP. A former banking executive with the Bank of Boston, and Citibank Tokyo, Bara holds an MBA degree from Harvard Business School, and a BA from Oberlin College. Earl Charles, CFO Prior to joining XenSource, Earl Charles was senior vice president and CFO at Cupertino Electric, Inc., where he developed programs to manage risk and decrease operating costs. Previously, he was CFO of mycfo.com, and the COO and CFO of Youth Sports Network. Charles has also held CFO positions with Prism Solutions, Inc., R2 Technology, Inc., and Strategic Mapping, Inc. Charles was previously a partner and 20-year veteran at Deloitte & Touche LLP. He holds a BBA from the University of Notre Dame. John Glendenning, VP Worldwide Sales John Glendenning joined XenSource from Platform Computing where he headed sales for Northern Europe and the European channel. He has also held a number of sales leadership roles at Citrix Systems within Europe, North America and Asia as well as at Compaq Computer. Glendenning's enterprise systems and software sales management experience has always included a highly leveraged channel and industry partner model and this is a philosophy being driven through XenSource's entire sales processes and field operations. Gordon Mangione, Senior VP Product Operations Gordon Mangione comes to XenSource from Ignition Partners, a Seattle-based venture capital firm where he was an executive in residence. Prior to Ignition, he served as corporate vice president of the Security Business and Technology Unit (SBTU) at Microsoft Corporation. Mangione also served at Microsoft as corporate vice president of both the SQL Server Team and the Exchange team, and as a product unit manager in the Internet Services business and the development manager for SNA Server, one of the first servers developed for the Microsoft Windows NT operating system. Mangione joined Microsoft in 1991 after working as a developer at Bell Northern Research, where he focused on PC/telephony integration. He has an engineering degree from the University of Waterloo in Ontario, Canada. Rafael Santini, Sales Americas With over 18 years in technology sales for both early stage start-ups and large technology ISVís, Rafael Santini brings an in depth understanding of how to identify a solutions primary value proposition and build a scalable repeatable sales process and team to meet the companies goals. Prior to joining XenSource, Santini spent 8 years at Marimba most recently as the VP of Americas Sales. His leadership and sales contributions help ensure a successful company IPO in 1999 and eventual acquisition by BMC in 2004. Prior to Marimba Santini held a number of sales leadership roles at Compuware Corporation, Ecosystems, Demax Software and The Wollongong group, covering core technology sectors such as Networking, Systems Management and Security. Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 26 Exhibit 9 XenSource Funding History Series Date Investment Post-Money Valuation Investors Lead Investor A 01/2005 $6 M
$12.8 M • Sevin Rosen Funds, • Kleiner Perkins Caufield & Byers • Accel Partners Sevin Rosen and Kleiner B 09/2005 $18 M
$57 M All previous plus NEA NEA C 10/2006 $17 M
$137 M All previous plus Ignition Venture Partners Ignition Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013 XenSource E-267 (A) p. 27 Exhibit 10 XenSource Timeline Year Quarter Funding Round 1 Round 2 Round 3 (planned ) Products v3.0 Linux-only commercial product v3.1 Windows support lands v3.2 planned updated release (Q2/07) v4.0 planned (Q3/07) Executives Gault Barr Crosby Pratt Levine Artale Bara Charles Glendenning Mangione Santini X X 2002 1234 2003 1234 2004 1234 2005 1234 $6 M
$18 M 2006 1234$17 M
Jan
Jan
Feb
July
Jan
Dec
May
Sept
Dec
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 28
Exhibit 11
Xen Open Source Code Access Model
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 29
Exhibit 12
XenSource Product and Pricing Menu
Intent
User profile Concurrent VMs
Physical RAM
RAM per Windows virtual machine
Physical CPUs Physical-to-virtual conversion
(P2V)
Windows guest support
Linux guest support Guest VLAN support
Resource management and prioritization (CPU, disk, network) Live Migration
Shared storage
Administrator Console
Price
Maintenance (software updates)
Maintenance Entitlement Technical Support
XenExpressô
XenServerô
XenEnterpriseô
Free Starter Package Developers, testers, support, IT enthusiasts
Up to 4
Up to 4GB
Up to 4GB
Up to 2 sockets
Linux P2V included; Windows P2V via technology partner offers (optional)
Windows Server 2003; Windows XP; Windows 2000 Server
Red Hat EL 3.6, 3.7, 3.8, 4.1, 4.2, 4.3, 4.4, 5.0; SUSE SLES 9.2, 9.3, 10.1; Debian Sarge
N/A
N/A
N/A
N/A
Single server at a time
Free
N/A (free)
N/A (free)
Available for purchase
Consolidate Windows Servers
Windows IT professionals Up to 8
Up to 8GB
Up to 4GB
Up to 2 sockets
Windows P2V via technology partner offers (optional)
Windows Server 2003; Windows XP; Windows 2000 Server
N/A (Windows guests support only)
N/A
N/A
N/A
N/A Multi-server
$99 annual subscription Included in subscription All service packs, minor ("dot") releases, and major (whole number) upgrades for same product Available for purchase Consolidate Windows and Linux Enterprise IT, systems integrators Unlimited Unlimited by license (16GB in current release) Unlimited by license Up to 32 sockets Linux P2V included; Windows P2V via technology partner offers (free starter) Windows Server 2003; Windows XP; Windows 2000 Server Red Hat EL 3.6, 3.7, 3.8, 4.1, 4.2, 4.3, 4.4, 5.0; SUSE SLES 9.2, 9.3, 10.1; Debian Sarge Standard Standard Mid-2007 Mid-2007 Multi-server$750 perpetual per dual socket; $488 annual subscription per dual socket
20% of MSRP for perpetual license; included in subscription All service packs, minor ("dot") releases, and major (whole number) upgrades for same product
Available for purchase
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013
XenSource E-267 (A)
p. 30
Exhibit 13 VMwareís Product Portfolio46
46 VMware Form S-1, http://sec.gov/Archives/edgar/data/1124610/000119312507090409/ds1.htm#rom83668_8 April 26, 2007 (July 3, 2007).
Purchased by: Zheng Xiang realerickxiang@gmail.com on September 13, 2013