VCDX – Think like a Panelist

The Panelists are trained to ask questions without giving the answer away.  When you first hear the question, they will appear “vague” and “soft”.  So you need to train yourself to look beyond the generic question and extrapolate the answer from the possibilities contained within.  This includes clarifying the question with the Panel and demonstrating your knowledge.  Always be mindful of point scoring opportunities.  If you do not know the answer, try working through the question out loud and arrive at your answer dynamically.  This proves that you can think on your feet.

List of articles in my VCDX Deep-Dive series (more than 70 posts)

For example: Panelist asks, “How did FC Switching impact your Storage design?”  You think for a moment and respond with, “Are you referring to the redundant switch fabrics or the single-initiator-multi-target zoning?” The Panelist says, “Yes, I am interested in zoning.” Then you respond, “Because this design is for a Cloud Provider, we could not implement single-initiator-single-target due to the administrative overhead of so many permutations and combinations, therefore I was forced to select Single-Multi.  Single-Multi increases the audience of RSCN broadcasts but it does reduce the operational risk of zoning mistakes.  That is the trade-off.”

The quickest way to develop this mindset is to peer review as many designs as you can.  Reviewing your own design is not enough, you need to hone your skills on designs that are not yours, it removes emotion and ego from the process.  We are naturally protective of our designs and have “blind spots” for the weaknesses of what we create.  This is why Study Group is so important, you pressure test each other to become better architects.  As Michael Webster commented below, you need to focus the beam of your “Critical Thinking” on your design; find the weaknesses and fix them, pre-empt the questions you will be asked and develop responses.

Listed below is a sample set of questions that I developed for VCDX Mock Panel Defences (they are in no particular order, I typed them as they occurred to me):

• You have five 9s of availability (or four or three), how was that achieved?
• So where are you tracking and measuring that SLA?
• So what else do you need to achieve five 9s and how have you tested it to prove you can actually achieve it?
• So what are the other aspects outside of technology to meet five 9s?
• So did you do that?
• So how long does it take to switch from site to site?
• So how many outages a year can you tolerate?
• So how do you prove that the DR Automation scripts work before trying for real?
• You have two Windows VMs running with the same IP Address. What happens?  What if portgroup security is enabled?
• What if they are Linux VMs?
• Storage DRS I/O Metrics and SIOC Congestion Threshold have what relationship?
• Why would you failover to DR? What would have to occur to failover to DR?
• Advantages/Limitations of single target zoning?
• What is the advantage of using beacon probing?
• Requirements for beacon probing?
• Advantage of physical NIC load balancing?
• Biggest constraint of blade servers?
• Technical issues using blade servers?
• If you were not constrained by the blade servers, what would you implement?
• Are there technical advantages?
• Business Critical Apps: What was the I/O profile? read/write ratio? How was it determined?
• Advantages of Powerpath versus Roundrobin?
• Storage Auto-tiering impact on the vSphere layer?
• So you are using Storage I/O Control then? How does SIOC interact with FAST-VP?
• Lets say you have your clusters split between separate chassis and you have Exchange DAG clusters, what stops those local copies from being in the same rack?
• What are some of the downsides of using FCoE?
• Have you used Jumbo frames on your Virtual Machines?
• Just back to storage DRS, what are the settings and how do they impact your design?
• How does VASA work?
• What is RSCN?
• What are the implications of locking down the host from the DCUI?
• What happens if you manage to corrupt your vCenter or the vCenter becomes unavailable?
• What guidance or advice did you give around operational processes such as capacity planning and things like that?  How does capacity planning impact on your design?