In the previous post we talked about the Pure Storage FlashStack Converged Infrastructure (FlashStack CI for short) announcement and answered some basic questions. In this post we will focus on the Cisco UCS B-Series blade server architecture and some of the key concepts and features that makes it unique and differentiated in the busy server market space.
Cisco Unified Computing System (UCS) is an integrated Compute, Storage and Network system that can be configured with a embedded single point of management. The key differentiators include –
1. Service Profiles: Server provisioning via a programmable open XML interface using a centralized management tool. The process can help in automating the server provisioning process and help enterprise to build infrastructure as a service (IAAS). The ability to define policies and templates to deploy a service profile and apply it to a server to provision a server makes it fundamentally unique way of server provisioning.
2. Cisco SingleConnect Technology – Simplified connectivity with Unified Fabric (FCOE) as the backend protocol between the servers and the Cisco Fabric interconnect. This reduces the number of cables that runs from the back of the server chassis to the top of the rack switch.
3. UCS manager (UCS-M), and UCS Director, a Unified, model based management for configuring the server blades and rack mount servers by means of service profiles helps in drastically reducing the time to deployment times and makes server firmware management very easy.
4. Cisco Virtual Interface cards (VIC), converged network adapters (CNA) which lets UCS-M to configure as many NICs and FC ports on the card and provide as many as 512 – 10G, 40G ports to the servers.
Cisco UCS comprise of B-series (Blade servers) and Rack mount servers, and support Virtual Machine Fabric Extender (VM-FEX) virtualization support which helps deliver a very agile virtualized data center. Furthermore the benefits around reduction in cabling, power and cooling, ease of manageability makes it a idea server platform for building both private and public cloud infrastructure.
Detailed UCS Architecture
Lets take a closer look at the UCS Architecture and look at all the components that make up the B-series platform.
Picture showing Cisco UCS Blade server chassis with B-Series B200-M3 blades and Cisco UCS Fabric Interconnect
As shown in the picture above, the blade servers are housed in the Cisco UCS Chassis (can accommodate 8 half width blades or 4 Full width blades). There are redundant fans and power modules and the chassis has a inbuilt Cisco UCS Fabric extenders for upstream connectivity. The blade servers connect to the back plane on one end and the fabric extenders are connected to the other end of the back plane (think of the line card design in Cisco routers). The connectivity is via SFP+ cables connects the Cisco UCS Fabric Extender to the Cisco UCS Fabric Interconnect using Fiber Channel over Ethernet (FCOE) protocol. This simple wire-once infrastructure makes the platform very easy to configure and deploy (as opposed to having two or more cables for FC and Ethernet connectivity). The connections can be a 1 / 2/ 4 or 8 wires per the Cisco UCS Fabric Extenders(FEX) (assuming you are using Cisco UCS Fabric extender 2208XP) based on how much bandwidth you would need for the entire servers in the UCS Chassis. The FCOE traffic can be split into FC and Ethernet at the Cisco UCS Fabric Interconnect(FI) for upstream SAN and LAN connectivity. The newer Cisco UCS Fabric interconnect 6248/6296 UP series support direct connection of FC and iSCSI target ports and allows every port on the FI to be configured as any desired ports (hence it is called Unified Port switches).
Cisco UCS 6200UP series Fabric Interconnects
The Cisco UCS 6200UP series Fabric interconnects (FI) provides both the network connectivity and management capabilities for the Cisco UCS blade server or Cisco UCS Rack mount servers. The 6200UP series is the second-generation fabric connects that has an enhanced feature set such as unified ports; and line-rate, low-latency, lossless 10 Gigabit Ethernet, Fibre Channel over Ethernet (FCoE), and Fibre Channel functions.
The two operating modes supported on the fabric interconnect for Ethernet are:
• Ethernet end-host mode, sometimes referred to as Ethernet host virtualizer
• Traditional Ethernet switch mode
The two operating modes supported on the fabric interconnect for Fibre Channel are:
• End-host mode, sometimes referred to as NPV mode
• Fibre Channel switching mode
Note, for both Ethernet and FC, end-host mode is the default mode of operation. The Fabric Interconnect reboot is required to change the operating mode (from End Host to Ethernet Switching mode or vice versa).
Picture showing the Cisco UCS Fabric Interconnect Ethernet and FC modes (End Host is default)
If we dive deeper and look at the x-ray diagram to see how things are connected internally, we get a clear idea of the architecture.
Detailed picture showing the server to FEX to FI connectivity and upstream Fabric switch connections (Source: Slideshare.net)
One thing that is very clear in the picture above is the redundant connectivity of the every component that includes Adapters, Fabric Extenders, Fabric interconnects and the backplane connectivity.
Wrapping up on the topic of Cisco UCS architecture there are many ways we can connect a block storage device to the UCS platform (topic of the next blog) but any reference architecture will have to include the compute, network and storage design and the picture shown below is very typical configuration.
Picture showing a typical configuration of Cisco UCS with upstream Nexus 5500UP series switch and Pure Storage (Corrected on April 5th)
Hope we got a fairly good idea around the Cisco UCS architecture, in the next post we will talk more about Cisco UCS service profiles and how to configure them.