The Lippis Report

It’s been over a week since Cisco launched its Unified Computing System (UCS) initiative. I’ve reviewed all the presentations, documents, financial analyst research notes, twitter and Facebook chatter as well as talked to a dozen or so IT leaders and in this Lippis Report Research Note I provide my assessment. There’s so much to consider with the UCS launch, the technology, the data center value, go to market strategy, competitive responses, etc. But I have a one important observation. This is the first time in the history of IT that a networking company has entered the computing industry. My first job in 1984 was at Digital Equipment Corporation, where networking was always an accessory, or stepchild, to computing. And yes, a lot has transpired since then but the sheer fact that a networking company has grown to the size of Cisco and that it can stand up and say “Hey IBM and HP, I’m entering your core market and there’s nothing you can do about it” is a remarkable occurrence. Anyone who has spent his or her career in networking and has worked within DEC, IBM or HP networking groups knows how it feels to be tangential to corporate strategy. For this group, UCS was a moment that tipped networking technology and networking professionals to a higher level of importance and influence in the IT industry. In this Lippis Report Research Note we review UCS and Cisco’s opportunities and challenges.

Let’s start with what the problem is that Cisco is attempting to solve with UCS. The key issue UCS addresses is scaling virtualized data centers. While there has been much hype associated with virtualized data centers, its market penetration is only 15% among the 28 million x86 world-wide installed servers, thanks in large part to the lack of support for legacy applications to be hosted on virtualized machines (VM). As VMware represents some 70% of the virtualized data center market it only makes sense to focus there when asking why more applications have not been virtualized. The absence of widespread virtualization adoption includes the lack of a true real-time failover, immature management tools, a lack of VM visibility and security questions about virtual environments. VMware’s vSphere (VI 4.0) scheduled for release 2Q09 does much to address many of these existing issues and will be a catalyst for increased virtualization projects as legacy applications will become increasingly virtualized.

vSphere will span up to eight cores from 4, address up to 256 GB of memory, allow users to change the amount of RAM allocated to VMs without rebooting and increase its maximum RAM limit to 1 TB. vSphere will now be able to virtualize larger, more mission-critical applications, such as Oracle, SAP and large Microsoft Exchange implementations. Some have estimated that VMware will be able to virtualize more than 99% of applications in existence with this release.

What Cisco sees is a market transition to massive virtualization implementations that current data center providers do not address. UCS does not go after the non-virtualized blade servers but participates in the growth market of virtualization that will lead to private cloud implementations, and then inter-cloud (private/public) projects.

Why Now?

Why is the network so important now? In an ideal virtualized data center IT resources, those being compute, storage and networking, would be pooled, with services dynamically drawing from the pools to meet demand. Virtualization techniques have enabled processes to be moved between computers, but network constraints create barriers that prevent elasticity such as static network assignment via VLANs, ACLs, broadcast domains, fragmentation of resources thanks to load balancers, service-specific network engineering and poor server-to-server connections, etc. In addition virtualization adds significant load to data center networks and this load is only going to increase significantly. Networks traditionally move data between client, server and storage devices. But as applications become more mobile-enabled by vSphere, networks will need to move entire application, OS and driver bundles if applications are going to spin up and down based upon demand. The additional network load is huge as the file size of a single VM image is typically 16 GB, but can be up to 2 TB, which will move over the network for “just in time” application provisioning.

It seems that Cisco seeks to eliminate network barriers to virtualization, increase data center network bandwidth and compute resources while simplifying data center architecture and its associated cost. Key to the last point is consolidated I/O thanks to Fibre Channel over Ethernet (FCoE) support within the Nexus 5000 which Cisco believes will require fewer switches, adapters and cables, reducing management points and thus streamlining operations. Its not-yet-available VN-Link or Nexus 1000v hopes to enable management of the virtual environment in the same way as the physical environment and break the boundary between server and network. Its recently introduced fabric extender in the Nexus 2000 promised to scale server connections without increasing management points and streamlines cabling.

Unified Computing System

But UCS is a system with integrated management, networking and consolidated storage/network access. UCS unites network, compute, and virtualization resources into a single system consisting of:

UCS Manager: Integrated system-level device management

UCS Fabric Interconnect: Line-rate 10 GbE, DCE & FCoE fabric

UCS Fabric Extender: I/O fabric extension, cut-through architecture

UCS Blade Server Enclosure: Optimized for energy efficiency

UCS Blade Server: Patented memory expansion, based upon new Intel Xeons large memory capacity unified fabric

UCS Virtual Adapter: Scalable virtual HBA & NIC resources

Support for the following 3rd party software solutions were announced on March 16th:

Hypervisor: VMware ESX & Microsoft Hyper-V

OS: Microsoft Windows Server, RHEL & SLES

Management: BMC BladeLogic, Microsoft System Center & EMC Smarts

Application: Microsoft SQL Server, Oracle DB & Oracle RAC

UCS is based on a standard set of components with which most IT staff are familiar. The intelligence for managing the overall system is based on a Nuova-class processor that Cisco has embedded in the UCS Fabric Interconnect. The UCS Manager software that manages the entire system communicates with firmware embedded in every device in the system. Note that UCS provides three adapters: a standard 10GB Ethernet adapter, a 10GB FCoE adapter, and most important, the adapter which supports the virtualization of network connections.

UCS looks like a standard rack of servers, but the wiring of the components is much simpler and easier to manage, track and deploy. This approach creates a cleaner, simpler model for managing data center assets that stands in sharp contrast to existing PC architectures. As a result, IT organizations should find it easier to set and maintain hardware policies. They will also be able to better physically secure the environment. The overall environment supports 63 percent more airflow than traditional servers, which leads to substantially less heating and cooling costs. UCS Manager can manage each component of the system to the point where specific power and cooling thresholds can be set. UCS also provides easier access to disks residing on each server blade. Power supplies are hot swappable. Each UCS should have two Fabric Extenders to maximize availability. Taken all together, the UCS should greatly reduce all the points of management compared to traditional blade server environments.

Cisco believes that it can drastically reduce data center total cost of ownership with UCS thanks to its ability to consolidate data center infrastructure through fewer switches, servers and management points, driving down both cap and op ex cost while delivering faster time to new server deployment.

Cisco’s Opportunities

UCS is innovative on multiple levels including leveraging Intel’s next generation of silicon, significantly expanding available memory, timing UCS to be ready for VMware’s vSphere VI 4.0, simpler cabling, easier management and provisioning and consolidating I/O. As networking speeds increase to 10Gb, 40Gb, 100Gb and beyond, computing will increasingly become hollowed and distributed through the network. By integrating computing, networking and virtualization Cisco has developed a platform that enables this re-distribution of computing and applications to occur along with its associated benefits of lower cost and increased utility of IT resources.

At the same time UCS is offered, the world economy has fallen into recession. One impact of this has been that the barriers between IT administrative domains of storage, applications, networking and compute have fallen too. The requirement to focus on service delivery and cost efficiency within IT has been dramatic over the past 9 months with much of the previous IT organizational distractions and friction being swept away. As a result, Cisco will find a welcome reception within executive IT to learn about UCS and how it may benefit their corporation.

Cisco’s Challenges

Cisco will find this adjacent market entry much more difficult than its previous entries into IBM’s SNA, voice communications, security, etc. In short it’s competing with a different class of competitors, going head on with HP and IBM, who have expected calendar year 2009 revenues of $110B and $100B. HP in particular is not a stranger to virtualization as it’s the number one provider of such services and products to market, providing VMWare and Microsoft virtualization software. Cisco hopes that it will be able to ride the virtualization transition much faster than HP and IBM. And while this was true for Cisco in the past, there is no evidence that IBM or HP will be slow to respond to the virtualization and cloud computing industry initiatives. Further, most IT leaders buy based upon market share. Cisco enjoys this preference in the networking space while HP and IBM do the same in the blade server market. Cisco will have to overcome this preference and time will tell if UCS innovation tips the preference scale.

As organizations look to transform data centers, strategic system integration may be a key area of focus. IBM and HP derive approximately $58B and $37B in revenues from IT services; that is one of the largest challenges Cisco faces. HP recently acquired EDS while IBM global services are massive. Cisco has a partnership with Accenture and Tata. While there is no rule that Cisco needs to have a large professional services organization, it clearly has served HP and IBM well.

Margins will decide if UCS is successful. As I wrote in Lippis Report 119 HP operates happily on 25-30% margins while Cisco enjoys 65-75% margins. Even if Cisco does generate $1B at 30% GM that’s $300M profit, but they could lose $1B and $650M of profit too in switching. It’s unclear what the margin contribution of UCS is and until that is known, it’s not possible to judge if Cisco’s bet will pay off.