Hardware Capacity Planning for Part-2 Exam 1Z0-865 Deployment Diagram

[Mithun Ms]

Dear All,
I have a question regarding the Deployment Diagram for Part-2 Exam (1Z0-865).
Are we required to calculate the required number of web & application server cores and Hardware profile (CPU, RAM, disc memory, network bandwidth requirements) for the target enterprise application? If yes, is it possible to calculate it using only the following information?
1. Total number of users (not concurrent users)
2. Expected growth of total users in next 1 year
3. Min required Speed of web page response/performance.

Based on my searches on the internet, I did not find any way to find this out before building the application. All approaches suggested to build the application, test it, profile it and then calculate the hardware requirements.

Can any ranchers who successfully passed the part-2 assignment share details on the following:
1. whether you calculated the hardware requirements and then used it in the deployment diagram?
2. If Yes to point 1, then please share any pointers to such a material/ read-up.
3. Or did you just use a minimum number of server nodes for implementing High Availability?

[K. Tsang]

Hardware profiles in the deployment diagram is a recommended thing to have. How accurate such profiles is without any stress/load testing is anyone's guess.

Unless your assignment specify those things you mentioned (# of users, response time, growth etc) a decent server configuration will do. Of course some servers may need more CPU or RAM given their function.

For HA, usually have 2 of everything is basically the minimal requirement. Server clustering is another option.

Check out this link for insight what should be considered when doing deployment diagram.

[Mithun Ms]

Hi Tsang,
Thanks for your quick response. I also went through the link you provided.
My assignment mentions following NFRs:
1. Total number of users
2. Expected growth of total users in next 1 year
3. Min required Speed of web page response/performance.
4. Expected up time/availability.

As you mentioned: "Unless your assignment specify those things you mentioned (# of users, response time, growth etc) a decent server configuration will do. Of course some servers may need more CPU or RAM given their function. ", these points have been specified in my assignment. How do I find the appropriate hardware configuration (CPU, RAM, memory etc) to meet these NFRs? I want to learn how to estimate this. Can you please share pointers to any materials in this regard?

[K. Tsang]

For hardware profile: First you look at what a typical (web/app/db) server comes with on the net to get an idea. Then determine if any servers are CPU/memory intensive etc and adjust accordingly.

Do mention something like without stress/load testing such profiles are for reference only. The testing will determine the accuracy of resources.

Given those NFRs, the uptime availability is the simplest to meet. I hinted that in my previous post.

Total number of users and expected growth is talking scalability. Say you have server with X RAM, X CPU core etc, if there is sudden surge of users will the server crash or perform slower than expected?

The network response time is talking about time from user click to all those servers and back. Assuming the bandwidth is big enough is one thing, but the API calls can impact such network response time.

There isn't a fool-proof formula per se for capacity planning. It's an art in itself. If you want, you can check out the book "Art of Capacity Planning" by John Allspaw from O'Reilly

[Mithun Ms]

Thanks for the details, Tsang. I will work on it accordingly.