Yesterday I ran some face to face mocks at Dell EMC’s offices in Brentford UK and during part of those at the end we discussed the design scenario and what i recommend people practice. I have partly mentioned some of it already in a posting i did after the last face to face mocks here. Additional to that I was showing those people that were going to defend this coming week the plan I had around the things I wanted to ensure I asked and also what I felt I needed to keep in mind during the design scenario. Now i have to make it clear I am not a panellist nor have i seen nor know the rubric the scenario is scored on but what worked for me at least are the following:
Question any information that doesn’t make sense!!!!
1. NUMA aware applications/OS’s?
2. If not how wide is the VM that is not NUMA aware?
3. Amount of growth?
4. Total vCPUs? Divide that by 5
5. Peak GHz for CPU?
6. Peak RAM for Memory?
7. PCIE connected hardware or USB dongles?
8. Rack or Blade? If blade are they half or full and logical diagram spreading across chassis. FDM mention
9. What kinds of applications are on these servers?
10. Scale up or scale out? N+ value
11. Right size VM’s after conversion
12. Add N+ value
2. Existing Tiers of Storage?
3. Amount of Free space? Is this dedicated to the project? How long is the support for this?
4. PEAK IOPS
5. Average IO size
6. Read/write ratios
7. SP speeds? Active passive,active active, ALUA?
8. Standard access switch and core setup?
9. Speed of existing FC switches if applicable?
10. Existing HBA’s? speeds of these HBA’s?
11. Software or hardware initiators?
13. IOPS, Amount of required storage?
14. Amount of growth?
1. Traditional or CLOS?
2. North/south or East/West traffic?
3. vSS,vDS on N1KV?
4. Hybrid or vDS only?
5. 1GbE or 10GbE?
6. Converged Network Adapter or 10GbE?
7. Peak network throughput required?
8. Speeds of Network Switches
9. MTU for Jumbo frames 9000
1. HA and DRS? (limit 32 hosts and 3000 VM’s)
2. Dedicated management cluster?
3. Virtual vCenter or VCSA?
4. Linked Mode?
5. vCenter heartbeat
6. VUM and UMDS?
8. vCenter design to separate vCenter from DB
I also did a second prep list that is very much alike to the one above but some slight addtional parts and for the cenceptual were the questions i had made a sentence out of to remember to ask them as i felt they gave me a good start and it was what i asked myself during each practice of the design scenario so when i did it in the defence it just felt like another practice and so calmed my nerves
- Pick out the main objective of the project (design a cluster for the migrated workloads)
- Write down all other requirements ,assumptions ,risks and constraints
- Availability Requirements? <- if over 99.9 then warn of additional costs
- N+ requirements
- If existing hardware then how old? Is it dedicated to the project? What speeds and free space does it have? Processor family? NIC speeds? Room for growth? Anything missing that is required (HBA cards or 10GB NICS)
- Consolidation and containment seem to always come up so ask about the steps for migrating the servers to vm’s and minimizing the risk of server sprawl….
- BC/DR? <- How soon is the requirement for the failover site and is the site even built? Distances and latency if already built
- Total number of CPU’s <- Question/highlight if really high or low
- Are applications/OS’s NUMA aware? What is the largest non numa aware vm?
- Calculate number of required hosts
- Very quick mention that could work it out with Peak CPU and Peak RAM also but DON’T DO THIS CALC
- Add N+ to the total number and make sure the cluster doesn’t go over 32 or 3k vm’s (This was applicable to my vSphere 5.0 design where the limit was 32)
- Right size VM’s after conversion
- Peak Storage ,Peak IOPS,Average IO,Read/write? <- Question/highlight if really high or low
- Storage Protocol?
- Active/Active, Active/Passive,ALUA?
- Different workloads require different tiers of storage?
- Allocate RAID levels to different tiers
- Traditional Core access switch setup?
- Total Peak throughput <- Question/highlight if really high or low
- North/south and east/west
- 1GB or 10GB
- CNA or 10GB
- vSS,Hybrid or VDS based on licencing question and on amount of required throughput and storage protocol
- HA and DRS with the N+1 portion from earlier included
- If avail was high and they keep to it then vCenter heartbeat
- Dedicated Management Cluster
- Separation of database and vCenter for resiliency
- vSphere licences from conceptual questions