Cisco DCID 300-610 study materials
Cisco DCID 300-610 study resources

Cisco DCID 300-610: Data Center Compute Design

Cisco DCID 300-610 · Work through your exam, one topic at a time.

All guide sections

Data Center Compute Design

Data Center Compute Design covers roughly a quarter of 300-610 and tests whether you can design Cisco UCS domains-B-series with fabric interconnects, C-series rack servers, and Intersight-managed environments-so that fabric redundancy, port roles, and service-profile abstraction all hold up.

Rules that decide the answer

  • Two disjoint fabrics. Fabric A and Fabric B are separate FI/VSAN domains. No LAN or SAN traffic crosses between them; the only shared elements are the UCSM cluster links and management.
  • One role per FI port. A port is unconfigured, server, uplink, FCoE uplink, FCoE storage, appliance, or (on 6300 unified ports) FC. Change role only while unconfigured; a port already set to uplink cannot become a server port without resetting it.
  • Chassis connectivity. Each chassis has two IOMs; IOM A connects only to FI A, IOM B only to FI B. Server ports to an IOM form plain port channels-no LACP.
  • Uplinks. LAN uplinks use LACP port channels to the upstream switch; FCoE uplinks need DCB/PFC end to end.
  • Abstraction. A service profile = pools (MAC, WWNN, WWPN, UUID, IQN/IP) + policies (boot order, BIOS, vNIC/vHBA templates, QoS), which is what makes replacements stateless.
  • Fabric failover on vNICs only works when fabrics are disjoint; vHBAs rely on multipath in the OS.

Worked example

Four B200 M5 chassis with 8 servers each, VIC 1385 (2x25G), four 100G LAN uplinks per FI and two 32G FC uplinks per FI. Host bandwidth per fabric = 4 chassis x 8 servers x 25G = 800G. Uplink capacity per fabric = 4 x 100G = 400G, giving 2:1 oversubscription-acceptable for a virtualized workload with a documented ratio. Server ports per FI: 4 chassis x 2 links per IOM = 8. SAN: 2 x 32G = 64G per fabric for boot and storage I/O, which must be checked against boot LUN queue depth, not just bandwidth.

Traps

  • Trying to change a port role without first resetting the port to unconfigured.
  • Assuming fabric failover protects FC-it does not.
  • Mixing FI generations in one cluster.
  • Forgetting that WWPN/WWNN pools must be unique per fabric.
  • Designing uplink oversubscription without accounting for vMotion and backup bursts.