Cisco UCS – A game changer

Hello everyone, and welcome to DCV CloudTech. In my last blog, we saw about EMC’s latest cloud offering VSPEX, its features and weather it provides any value add to the already cramped cloud services market we have in place. This time, we are going to have a look at Cisco UCS server offering, its features and what the latest release has in store for us. The reason, I am discussing UCS is due to the fact that some of the major cloud in a box offerings available in the market have UCS for its server component.

Unified computing system aka UCS was released around in 2009 by Cisco to compete with the already crowded blade server market which was dominated by HP and IBM. I have to agree that Cisco has done a really good job in pushing UCS through that crowded space and now, it’s on the top three in most countries. What makes UCS special? Let’s discuss some features that immediately come to my mind. We can always take each component and analyse them separately in my other blogs.

1. The first thing what Cisco has done to make UCS blade servers       special, is that they have taken the brain component that controls the blades out of the chassis. Instead of each chassis having its own brain, the brain is now a separate entity which is totally independent of the chassis. This enables us to control multiple chassis at the same time without the need to do any additional configuration. 

2. The connection from the brain to the chassis and blades is now done through IO modules or remote line cards. These remote line cards are controlled directly by the brain which helps to reduce the number of cables running from each server to the brain.

From the picture, it is clearly visible that the amount of cabling has been reduced by around 40%.

    3. Cisco has used FCOE as the main protocol to transfer data and management packets from the blades, till the Fabric Interconnects (the Brain). Now, people might think, what’s great in that? Since, the Fabric Interconnects are separate from the chassis, we can actually connect FCOE capable storage devices, to the FI’s directly and all of the resources on the FCOE storage devices can now be shared by all the chassis controlled by the FI’s. The FI’s are configured as a High Availability pair for improved redundancy. With the latest release of UCS, Cisco has also introduces support for multi-hop FCOE from the FI’s to an upstream FCF switch such as the Nexus 5500. By doing this, Cisco has introduced a solution which incorporates end to end FCOE.

      4. Extended Memory Technology is one more feature that stands out and it’s safe to say that Cisco is the only company that provides this solution in the market today. EMT is a Cisco exclusive technology in which additional memory slots are provided per socket (upto 24 slots per socket). This helps in reducing the cost of memory to a great extent by using smaller DIMM’s when compared to competitors who have 6 or 8 DIMM slots per socket and have to use large DIMM’s thereby resulting in increased cost.

      5. One of the major issues people had with regards to migrating their data centres to new hardware was that do they have to buy blade servers all the time, even if their requirements are small. Cisco made life easier for these people by introducing a rack server variant of their UCS system.  The rack server from Cisco offers the most of the features as that a blade server, but it does not need the Fabric Interconnect to be functional. But, what if I don’t need most of the features? What if I need all the features? The only way to do that  is to make the FI’s talk to the rack servers also. Cisco has been smart enough to understand that and have come up with a solution where the rack servers can be integrated with the FI’s, which is all the while managing the blade servers.

This management is done through special remote line cards call the nexus 2k. The key element and differentiating factor is that UCS Manager (which runs on the FI) creates a SINGLE MANAGEMENT DOMAIN FOR ALL SERVER PLATFORMS whether its blade or rack mounted. It is the only architecture that does that.  It also allows for seamless transtition of servers through stateless computing, allowing bare metal operating systems to be moved from blade to rack mounted server hardware without impacting any performance or software requirements.Again, more of this can be discussed in separate topics.

As we can see, Cisco UCS offers lot of features that are either not fully supported with other vendors or not fully functional. This makes UCS more of a game changer in the server industry. Again, I have discussed only few features and not all, as I want this to be more of a teaser. So until next time, see you in the cloud.

For DCV

K