08-27-2004 05:45 AM
Greatly appreciate any assistance you could give.
We have 4 Brocade Silkworm switches (1 off 2800 and 3 off 3800) fully interconnected to each other in a mesh configuration. All 4 switches are configured to be in one fabric and we've designated 2 switches as Edge and 2 as Core. We have 1 JBOD (HP VA7410) and 1 tape silo (STK L700) connected to the core switches and hosts (mixture of W2003, Linux & HP-UX) connected to the Edges. All devices (JBODs, hosts and tape silos) are connected to the fabric via 2 paths i.e. for hosts there is 1 path to each Edge switch and for the JBOD and tape silo one path to each core. So physically the configuration is fully resilient.
A new project has commenced with the intent to add a new JBOD (HP EVA5000), six additional hosts & two additional 3800 switches to our existing fabric. The installation engineer has advised that the existing fabric needs to be split into two separate fabrics if we require full resiliency i.e. "currently if one switch were to fail there is a potential for the whole fabric to become unusable. The configuration is not resilient to a single point of failure."
- is our single fabric prone to failure if one of the Brocade switches fails?
- if so will we need to implement a two fabric design?
I'll post a Visio schematic showing current and proposed re-design if I can find out how.