Update: FlexPod Entry Level Announcement

After learning some new information about the Entry Level FlexPod design I felt the need to put an update together on some of the questions I posed in an earlier blog post seen here.  First a quick recap, Cisco & NetApp announced a new FlexPod design for smaller workloads.  The new design consists of the NetApp FAS2240 storage array, Cisco Unified Computing system, and Cisco Nexus Series switches.  Below is a quick diagram I put together with all the parts & pieces and how they will interconnect.

What Protocols will be supported?

One of the questions I posed when the entry FlexPod was announced was which protocols and configurations would be supported. As you can see in the diagram above it is 10Gb – supporting iSCSI, NFS, and CIFS protocols.  Considering this is meant for smaller workloads it is not a huge deal that it will not support native Fiber Channel or FCoE (due to the mezzanine card is not a unified adapter) since this is meant for smaller workloads.  If by chance you outgrow either due to capacity or I/O load – you always have the option of upgrading to a FAS3200 series and turning the 2240 into a disk shelf for that FAS system.

Cost Comparison & Scale:

Prior to UCSM 2.0(2m) to manage the C series rack-mount servers you were required to connect them via a Cisco 2248 Fabric Extender for Management and the P81 to the Fabric Interconnect seen below.  This obviously doesn’t scale nearly as well and adds cost as you were purchasing additional port licenses for the Fabric Interconnect.  The ability to use the 2232 Fabric Extender will definitely make this a lot more cost effective as you can connect up to 16 rack-mount servers to a single pair of 2232 Fabric Extenders.

From a compute standpoint you can scale either through additional C series rack-mounts or through B series blade servers within the pod.  You heard that right, the FAS2240 will be supported in a B Series blade environment!

From a storage standpoint, the FAS2240 supports up to 5 additional disk shelves beyond the 24 drives within the disk controller itself.  If you manage to out grow that you can use the 2240 as a disk shelf to a 32XX or 62XX series system.  You could also add an additional controllers although you will not have the capability for Cluster Mode in the new code – the requirement of 10Gb inter cluster links leaves you with 1Gb connectivity for server/client traffic.

Cabling and Config Considerations

This should not be anything new since the 2208 I/O Module has been shipping for a while now – but its worth bring up – cabling for the 2232 Fabric Extender.  With the 6248 Fabric Extender there are six sets of 8 contiguous ports – with each set of ports manage by a single chip.  If you will be cabling up all 8 uplink ports from the 2232 you will want to maximize your VIF configuration and connect them to a complete set of 8 ports in the fabric interconnect.  More information on this can be found here.

From a configuration standpoint you will want to choose how your uplinks will function – either through Hard-pinning or port-channel within the FEX discovery policy in UCSM.

Advertisements

3 responses to “Update: FlexPod Entry Level Announcement

  1. Brian,

    Why would you put a 2232 behind a 6248? Are you suggesting building an entire data center behind a single port-channel off the 5000? In your diagram, if you add 2232, here is capacity off that single port-channel. 16xUCS per 2232×16 per 6200FI=256 UCS. <- That's bigger than most data center. At 10 VM per UCS, 2560 VM Capacity. 2×5000 can host additional 16 set of that. 4096 UCS on those 2×5000.
    Why would you even suggest a design with additional oversubscription point at 2232, potential multiple hop for data flow because of nature of FEX, when 2x6200FI already can provide full capacity of most data center still leaving the entire data center on a single port-channel.
    Such design is only suited for Service Provider who offer discounts on oversubscribed rack connections. But most Data Center can afford 2x6200FI simply because amount of important business that will be running behind it.. Flexpod is a HIGHLY scalable design and there really is no room for additional oversubsrcription with 2232.
    Any comments will be appreciated.

  2. Jon,

    Entry-level FlexPod calls for the 2232 due to the VIC in the C-Series server. To manage a C-Series UCS server it is a requirement in UCSM 2.1 to use a 2232 FEX for connectivity between the Fabric Interconnect and the C-Series VIC. I agree with the oversubscription however the Entry Level FlexPod is not meant for the scale you mention. Hope this helps.

    • Thanks. Ah, the VICs on C-series are what requires the 2232. What UCS with Nexus seems to offer is a stunning level of scalability and concentration on 5k’s 960Gbps backplane.. I don’t think most of IT appreciate the breadth of what that means. I can put 2 billion dollar business on 2×5000. Even worse it will even fit in a single FlexPod 2×6200 with a core. Scary thought.. But tempting..

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s