Scaling agents horizontally in Agentic AI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding horizontal scaling of agents

Imagine you have a team of helpers (agents) working on a task. What does it mean to scale these agents horizontally?

A Combining multiple agents into one bigger agent.

B Making each agent work faster by improving its code.

C Adding more agents to work in parallel on the task.

D Reducing the number of agents to save resources.

💡 Hint

Think about how adding more helpers affects the workload distribution.

Horizontal scaling means increasing the number of agents working side by side, so tasks can be done faster by sharing the load.

❓ Model Choice

intermediate

Choosing the right approach for horizontal scaling

You want to scale your AI agents horizontally to handle more user requests simultaneously. Which approach best supports this goal?

A Deploy multiple independent agent instances behind a load balancer.

B Optimize a single agent to use GPU acceleration.

C Reduce the agent's response time by simplifying its logic.

D Increase the memory of the machine running the agent.

💡 Hint

Think about how to handle many requests at once by using more agents.

Deploying multiple agent instances behind a load balancer allows requests to be distributed evenly, enabling horizontal scaling.

❓ Predict Output

advanced

Output of agent scaling simulation code

What is the output of the following Python code simulating agent task completion times when scaling horizontally?

Agentic AI

import random

def simulate_agents(num_agents, tasks_per_agent):
    times = []
    for _ in range(num_agents):
        agent_time = sum(random.uniform(0.8, 1.2) for _ in range(tasks_per_agent))
        times.append(agent_time)
    return max(times)

random.seed(0)
result = simulate_agents(3, 5)
print(round(result, 2))

A 6.00

B 5.00

C 4.75

D 5.51

💡 Hint

Run the code to see the maximum total time among 3 agents each doing 5 tasks with random times.

The code sums random times between 0.8 and 1.2 for 5 tasks per agent, then returns the longest total time among 3 agents. With seed 0, the output is 5.51.

❓ Metrics

advanced

Evaluating horizontal scaling efficiency

You measure the throughput (tasks per second) of your agent system as you add more agents horizontally. Which metric best shows how efficiently the system scales?

A Accuracy of each agent's predictions

B Speedup = Throughput with N agents / Throughput with 1 agent

C Latency of a single task processed by one agent

D Memory usage per agent

💡 Hint

Think about how adding agents affects total work done compared to one agent.

Speedup compares throughput with multiple agents to a single agent, showing how well horizontal scaling improves performance.

🔧 Debug

expert

Debugging horizontal scaling bottleneck

You horizontally scaled your agents by adding more instances, but the system's throughput did not improve as expected. Which issue below most likely causes this bottleneck?

A A shared resource like a database is limiting concurrent access.

B Each agent instance has too much CPU power.

C Agents are running on separate machines with network isolation.

D The number of agents is too high, causing perfect scaling.

💡 Hint

Consider what shared parts might slow down many agents working together.

If agents share a resource that cannot handle many requests at once, it becomes a bottleneck preventing throughput improvement despite more agents.