0% completed
What a Solutions Architect Does
On This Page
What the Role Is
What the Architect Learned First
What the Architect Produced
How the Work Is Measured
What the Role Requires
Using This in an Interview
Key Takeaways
Practice Questions
The chief technology officer (CTO) of a home furniture retailer sends this message to a cloud provider:
"Our data center lease ends in 18 months, and we cannot renew it. During our last holiday sale, the online store was down for 3 hours on the first morning. We need to move, and we cannot lose the store during a sale again."
The retailer has 140 stores, an online store that brings in 30 percent of its sales, about 220 servers, and 12 infrastructure engineers who have never run production systems in the cloud. The message reaches a solutions architect at the cloud provider.
Fifteen months later, the last server in the old data center is switched off, and the online store has handled a holiday sale without an outage. This lesson explains what the solutions architect did during those months, and what they deliberately left to the retailer.
What the Role Is
A solutions architect (SA) helps a customer turn a business problem into a technical design on a vendor's platform, and then helps the customer put that design into production. The vendor is the company that sells the platform, such as a cloud provider or a software company. The customer is the organization that pays for it.
The SA works for the vendor, but the work succeeds only when the design is right for the customer. A design that sells more services but fails during the next sale loses the customer.
The SA does not operate the customer's systems and does not usually write the customer's production code. The retailer's 12 engineers build and run the new platform, while the SA designs it with them and proves that it works.
What the Architect Learned First
The CTO's message names two goals but no limits. The SA's first task is therefore discovery: a set of conversations that find the customer's goals, constraints, and current systems before anyone proposes a design.
Discovery with the retailer produced five facts:
- The 220 servers run about 60 applications, and many of them depend on each other.
- During the sale week, online orders rise to about 6 times the normal busiest hour, from about 400 to about 2,400 orders an hour.
- The online store runs on a fixed number of servers sized for normal traffic, and adding a server takes several days. That is why it failed during the last sale.
- Card payments go through an outside payment provider, so the retailer stores no card numbers. This made the expected compliance work much smaller.
- The 12 engineers know the current systems well, but they need training before they can run cloud systems on their own.
Each fact changes the design. The diagram below shows the decision that each one produced.
Look at the third row. Without discovery, the SA might have proposed larger servers, which would cost more for 51 weeks of the year and still fail if the sale grew again.
What the Architect Produced
An SA's output is a set of documents, decisions, and working proofs that let the customer act with confidence. For the retailer, there were five:
- An architecture proposal. The online store runs on servers that the platform adds and removes with traffic. Its managed database keeps a standby copy in a second availability zone, a separate data center in the same region.
- A migration plan. Migration means moving applications from one environment to another. The plan grouped the 60 applications into three waves, sets of applications moved together, with the online store first so that it would run on the new platform before the next sale.
- A proof of concept. A proof of concept is a small working version built to test the riskiest assumption. Here, it was a load test that sent 2,400 orders an hour to a copy of the online store.
- A cost estimate. It showed yearly running cost about 20 percent below the data center after the first year, mainly because the extra capacity is needed only in sale weeks.
- Training and reviews. The SA ran workshops for the 12 engineers and reviewed their work before each wave went live.
How the Work Is Measured
An SA is measured by the customer's results on the platform, not by the number of designs or meetings. Vendors use different names, but the measures usually fall into three groups. The table is this course's summary, not any one company's standard.
| Measure | What it means | Retailer result |
|---|---|---|
| Technical win | The customer chose the platform because the technical case convinced them | Signed in month 3, after the proof of concept |
| Workloads in production | The applications that the customer actually runs on the platform | All 60 applications by month 15 |
| Customer outcome | The business goal that the customer started with | No outage during the sale; data center left 3 months early |
A workload is an application together with the infrastructure it runs on.
What the Role Requires
The work draws on six skills, and the rest of this course develops each one:
- Technical breadth: compute, storage, databases, networking, security, and data services, known well enough to choose between them.
- Architecture design: combining those parts into a design that meets the requirements, with the trade-offs stated.
- Discovery: asking questions that reveal goals, constraints, and the systems already in place.
- Cost and risk judgment: estimating what a design costs and what can fail.
- Communication: explaining a design to engineers and a recommendation to executives.
- Customer trust: saying no when the platform is the wrong fit, and keeping every promise.
Two mistakes are common. An SA who presents a design before discovery often solves the wrong problem. An SA who agrees to every request loses the customer's trust when a promise fails in production.
Using This in an Interview
When you describe a customer project, start with the customer's goal and constraint, then give your decisions and the measured result.
"A retailer had 18 months before its data center lease ended and had lost its online store for 3 hours during a sale. Discovery showed that the sale peak was 6 times normal traffic and that card data was already outside their systems. I proposed a three-wave migration with the online store first, proved it with a load test at 2,400 orders an hour, and the store handled the next sale with no outage."
Expect two follow-up questions:
- "Why move the most important system first?" Its failure was the reason the customer called, and the next sale was the earliest deadline. The load test reduced the risk before the move.
- "What if the load test had failed?" Tell the customer at once, find the limit, and change the design before they decide. A failure found in a test costs far less than one found during a sale.
Key Takeaways
- An SA turns a customer's business problem into a design on a vendor's platform and helps the customer put it into production.
- Discovery comes before design, because the constraints decide the design.
- The main outputs are a proposal, a migration plan, a proof of concept, a cost estimate, and training.
- The work is measured by the customer's results on the platform.
The next lesson, Solutions Architects, Sales Engineers, and Forward Deployed Engineers, compares the SA with the other roles that work directly with customers.
Practice Questions
1. The CTO asks the SA to "move everything exactly as it is." Why should the SA still run discovery?
<details><summary>Show answer</summary>Because the request does not show what limits the solution. Moving servers unchanged suits many applications, but the online store would still fail at the sale peak. Discovery finds facts like that one.
</details>2. Why did the fact that the retailer stores no card numbers matter to the plan?
<details><summary>Show answer</summary>It removed most of the expected compliance work. Systems that store card numbers must meet strict security rules, which could have added months to the migration.
</details>3. A customer signs after an impressive design, but six months later only 5 of its 60 applications run on the platform. Was the engagement a success?
<details><summary>Show answer</summary>Not yet, because the customer has not reached its goal. A signature is a technical win, but the SA is measured by workloads in production and by the customer's outcome. The SA should find what is blocking the other 55 applications, such as missing skills or an unrealistic plan.
</details>Reading Progress
0%
On This Page
What the Role Is
What the Architect Learned First
What the Architect Produced
How the Work Is Measured
What the Role Requires
Using This in an Interview
Key Takeaways
Practice Questions