As Cisco UCS enters a new phase of growth with Cisco’s entry into the top 5-server vendors list, let’s step back and take a look at how we manage infrastructure in the world of Unified Computing. UCS Management is a key ingredient of Cisco’s Unified Data Center, which unifies physical and virtual infrastructures across data centers. The architecture delivers IT services through a holistic fabric-based approach, and has proven to lower the cost of computing and enable real IT innovation. Customers are reporting an 84% reduction in provisioning times and 61% reduction of ongoing administrative/management costs. In this post we’ll examine some of the drivers for that. More recently, UCS Director has bolstered the portfolio and provided significant capabilities to reduce data center operational expenses. Since there are many questions about how the portfolio products complement each other, I hope to clarify it.
The UCS management portfolio includes:
UCS Manager for individual UCS domains
Provides embedded, unified, policy based management of Cisco UCS blade and rack servers and all fabric components within the system Read More »
A few weeks ago, I was at Cisco Open Source Conference 2013 -- a conference hosted by Cisco where we had speakers from IBM, Canonical, Red Hat and Rackspace, among others. I learned a lot, specifically about the evolution of Hadoop and the OpenStack project. As a follow on, I collated different activities around Cisco UCS and OpenStack, which I will share in this blog.
Dr. Dan Frye, Vice President, Open Systems Development, IBM, head of the IBM Linux Technology Center (LTC) gave the keynote address at the conference. It was nostalgic considering the fact that I sat in the same aisle as some of the LTC team members in the IBM facility in Austin, a few years ago. His talk included some fascinating historical anecdotes and three lessons IBM learned about open source software development-
“Develop in the open” (Don’t try to contribute finished software products, heed to feedback)
“Don’t reinvent the penguin” (“Scratch your own itch” – interesting phrase to explain the behavior of communities which want to solve the problems at hand and not those perceived to be problems by external entities)
“Work with the process” (The community process is usually an agile methodology with no assumptions on roadmaps and delivery dates)
These lessons are invaluable in light of the open source projects such as OpenDaylight (no pun intended) and OpenStack that Cisco is now an integral part of. According to Dr. Frye, these newer open source consortiums have the following characteristics:
Larger number of initial members
Relatively large initial budgets
Often require the commitment of a specified level of FTEs
Chris Wright from Red Hat expanded upon the principles and ethos of open source projects including release early, release often, iterative development and the culture of giving back. He contrasted the Linux kernel development project with the OpenStack project showing the relative speed of projects with the number of developers and commits by release. He gave a fantastic overview of the various Openstack component projects. He also identified two newly graduated projects namely, Ceilometer and Heat in the Grizzly release. I gave a talk on the requirements for the Ceilometer project, and you can find the slide deck on slideshare.
Capital cost savings with infrastructure consolidation
Lower operating costs with automation
Speed of implementation and infrastructure deployment
Better SLAs with faster recovery or migrations
Let’s dig a bit deeper and start by looking at the difference between a FBI server and a run of the mill server. FBI essentially lets us define the profile of a server in software. The profile here refers to as many as 120 attributes of a physical server stored as meta data in a profile. These attributes include BIOS version, LAN connection parameters, SAN connection parameters, UUID, MAC Address etc.
In the case of run of the mill servers some of these attributes remain the same throughout the life of the physical server. You may be able to alter other attributes with manual operations through proprietary user interfaces. As shown in the figure above, the server identity (service profile) of a FBI server is abstracted from the physical server.
Tune in to the webcast, this Thursday, Nov 8, which specifically addresses large-scale fabric computing to find out more. Research firm, Gartner, defines fabric computing as “A set of computing, storage, memory and I/O components joined through a fabric interconnect, and the software to configure and manage them”. In a study on fabric computing adopters earlier this year, Gartner researchers called out the following three major impacts:
External service providers justify fabric-based infrastructure (FBI) based on operating cost savings and density (for greater revenue per square foot), while enterprises base their FBI acquisitions primarily on capital cost savings.
Gartner clients found that FBI’s use of templates and profiles improves resiliency because, in the event of infrastructure failure, they can recreate servers in minutes.
Virtually all clients with FBI in production found a reduction in time to provision from two to three months to a few hours to three days.
Cisco Unified Computing System (UCS) is leading this industry transition to fabric computing, and with Cisco UCS Central, catapulting it to an unprecedented scale. In his blog, Todd Brannon, Unified Computing Product Marketing Senior Manager, explains UCS Central in a nutshell. Cisco UCS Central lays the foundation for disaster recovery by providing the ability to recreate the infrastructure environment in a different data center. With Cisco UCS Central, customers can manage dynamic environments efficiently without higher-level software and complex setups. With an open API, UCS Central allows users to retain existing data center processes and tools. It also provides role-based administration to support collaboration across disciplines and to accommodate necessary organizational changes.
Cisco Live in San Diego with a record breaking 17000 attendees was a blast, as you can tell from the video montage below. I unwittingly ended up in the video for a brief second.
The tag line “What you make possible” was prominently displayed at the conference. For me it was a great opportunity to talk to Cisco UCS users and partners who make a whole lot possible in their respective organizations. The keynote by Cisco CEO, John Chambers featured moving a computation workload from a Cisco UCS server in the data center to an E-Series server in a branch office and it created a buzz.
In a customer appreciation event, I got to talk to a prominent service provider and Cisco customer who had been a user of the Cisco UCS when it was just introduced in the market three years ago. We talked at length about how they wanted to set up a common server Read More »