We are just starting a push to convert all of our remaining ESCON devices to FICON. As part of this, the design of the fabric itself is the one thing I am struggling with the most - primarily because once we choose a direction, that is the path we will be on for the next x years.
Our ESCON fabric is configured in a 'mesh' style, with every CPU connecting to multiple ESCON directors, each of which provide connections to the various peripheral devices. This can be said to ensure path redundancy or remove single points of failure, depending on whether you are a glass half full or half empty person. This has served us well for a long time, and I am sure would continue to work effectively in our new environment. Certainly I have put together a few fibre-channel SAN's using this methodology. The other way I am thinking about involves the use of cascading within a single data centre. All hosts connect into a pair (or more) of directors, which are then connected to another pair (or more) of 'secondary directors' via ISL's. These secondary directors would provide ALL of the connectivity to the disk and tape devices. I guess you could loosely refer to this as a form of 'core-edge' methodology if you are a SAN design reader, but given there are directors in place at each level I see it as more of a 'core-core' structure. The benefits of this second mechanism would be a major simplification of the cabling environment - and therefore (I hope) any troubleshooting activity. On the down-side, there is the complexity of the cascading config work, the ISL traffic would need to be managed (buffer credit issues??) etc etc. Should I be thinking about this, or is sticking with the current mesh- style design still the best way to go?? ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
