Cisco's UCS explained

By Shamus McGillicuddy
Wednesday, 18 March, 2009


Cisco Systems yesterday (US time) unveiled its long-rumored entry into the blade server market with its new Unified Computing System, which it says will pave the way for cloud computing. If Cisco's strategy with this new technology works, networking professionals will see a lot of organizational change in data centers. IT organizations will have to blow up the silos that exist among networking, server and storage teams.

Rumors about the server product, code-named "California," have been circulating for months. Despite all the hype about Cisco moving into the server market, the company played down the introduction of the server technology. Instead, the networking vendor and its star-studded cast of partners -- including Microsoft, BMC Software, EMC, NetApp, VMware, and Accenture -- focused on the concept of Unified Computing, which Cisco believes will open the door for cloud computing at many high-end enterprise and service provider data centers.

Cisco's Unified Computing System (UCS) vision for data centers proposes a world where servers, networks, storage and virtualization are tightly integrated into one architecture that is optimized for delivering a scalable virtual infrastructure. Server virtualization has created headaches for networking professionals as virtual machines have begun migrating from one physical server to another. Making sure that the proper storage access, policy and security settings, and network connections are in place as those virtual machines migrate across the data center presents a major management headache for many enterprises.

"For the network guy, what this means is the industry is going to spend a lot more time looking at how you manage and provision virtual machines in the data center," said Abner Germanow, director of IDC's enterprise networks services. "And if you look at the complexity that is created when you connect servers to everything they connect to, [with virtualization] what you get is a level of complexity that increases very rapidly."

Germanow said the provisioning and management of both physical and virtual servers will be a much higher priority for the IT industry from now on as vendors invest in products that allow the process to occur faster, more efficiently and more securely.

Cisco's UCS is made up of a variety of hardware components and management software. Everything is built upon Cisco's new UCS B-Series Blade Servers, based on new Intel Xeon processors. Each server uses network adaptors to access a unified fabric and uses what Cisco is calling "memory-expansion" technology to increase the amount of memory that servers have available for dealing with large-dataset workloads.

The blades will fit into the UCS 5100 Series Blade Server Chassis, which will hold up to eight blades and two fabric extenders in a 6RU enclosure. The chassis will connect to a UCS 6100 Series Fabric Interconnect switch, which is a line-rate, low-latency, lossless 10 Gbps Ethernet and FCoE interconnect switch that consolidates server I/O within the system. Both 20-port and 40-port versions are available.

The Cisco UCS 2100 Series Fabric Extenders will carry unified fabric into the blade server chassis, providing up to 10 Gbps connections each between each blade and the fabric interconnect.

All of this will be managed by Cisco's UCS Manager, embedded software that streamlines all the UCS components into a single system. The software, in turn, has open APIs that will allow enterprises to integrate the UCS technology into their larger systems management software platforms. BMC, one of Cisco's UCS partners, has already done the work to integrate UCS into its management software, and Cisco hopes that other systems management vendors, such as IBM Tivoli and HP OpenView, will follow suit.

UCS Manager will offer role-based views so that network administrators can manage the data center network through it, while server administrators use it to manage servers, and storage administrators use it to manage storage systems.

Cisco said that the UCS system is optimized to plug into its Nexus family of end-of-row or core switches, which are enabled to support a data center built on a unified fabric, where storage and servers are all connected by Ethernet. However, the uplinks that come out of UCS will also connect with Cisco's Catalyst switches or switches from any other vendor.

As Cisco rolls this technology to the market with its partners, it will be trying to engage IT organizations rather than just networking professionals. Such a move could force some organizational changes with IT departments.

"One of the things that is fascinating about this announcement is that [Cisco] doesn't intend on selling it to the network guy," Germanow said. "If servers are more or less efficient, the network guy doesn't really care. Cisco is clearly looking to open up the conversations that they have across other parts of the IT organization. Cisco generally believes that the silos which exist between computing, networking and storage organizations largely get in the way and that there is a need for organizational change on that front."

Rob Lloyd, Cisco's executive vice president of worldwide operations, made this clear while unveiling UCS.

"The money that we hope to unlock with UCS is in the cracks between silos, in the buying centers of the enterprise," Lloyd said. "It will be the CIO that we have conversations with. It will be an executive engagement. That's how we've decided to go to market. I've already seen customers begin to realign to do that."

Glen Keels, director of product marketing for Cisco Unified Computing, said UCS should make life simpler for network administrators who have been struggling within their silos to provision networks for virtualized servers. "This will allow them to focus on more strategic issues in the data center," he said.

Cisco is already a strategic vendor in the data center, Germanow said. UCS will resonate with enterprises whose data centers already have thousands of virtual machines running on their servers, but which are looking to scale up to having tens of thousands of virtual machines over the next two to three years.

Other networking vendors, such as Brocade/Foundry and Juniper, will approach the problem of scaling upward with virtualization in a variety of ways, Germanow said. "The question will be: How much of a lead does Cisco have?" he said. "We'll see how long it takes before we start seeing some really interesting product announcements and partner announcements from other networking companies – and server and storage vendors."

Related Articles

How to prepare for the AI future (that isn't here yet)

Something big is indeed coming, but the revolution is not here yet.

Storage strategy in the multicloud era

Data has become the essential raw material in the strategic orientation of business, making data...

Private AI models: redefining data privacy and customisation

Private AI signifies a critical step forward towards a more secure, personalised and efficient...


  • All content Copyright © 2024 Westwick-Farrow Pty Ltd