← Adobe Interview InsightsSoftware Engineer

Adobe·Software Engineer·Onsite - System Design / Architecture·Senior

SeniorPrefer not to say
Jun 2026

Summary

Adobe system design round for a software engineer role. One question, but it kept branching into harder territory every time I thought I had a handle on it. Not a bad experience, just dense.

Questions Asked (1)

Q1

Design an API that returns a unique random integer from a range of 1 to 33 trillion, guaranteed never to repeat, at 300,000 requests per second. Walk through your data structure choices, how you'd partition the keyspace across nodes, how you'd ensure durability and exactly-once issuance, how you'd handle coordination and leader election, and how you'd extend the design if the keyspace doubled to 66 trillion.

System DesignAlgorithms & Data StructuresTechnical Trade-offs
Author's notes

I started with a bitmap and immediately realized I was describing something that would need terabytes of RAM to hold in memory.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying requirements and constraints, then propose a partitioned counter-based approach using a distributed database or a range-based allocation service. Walk through the design choices for data structures, partitioning, durability, and coordination, and explain how to scale when the keyspace doubles.

Pro tip: Emphasize that the system must be horizontally scalable and fault-tolerant; use a proven distributed coordination service like etcd or ZooKeeper for leader election and metadata management, and consider using a segmented counter approach to avoid hot spots.

1. Clarify Requirements and Constraints

Confirm the exact requirements: 300k RPS, unique random integers from 1 to 33 trillion, no repeats, durability, exactly-once issuance. Discuss trade-offs between randomness and sequential allocation.

2. Choose Data Structures and Partitioning

Propose a partitioned counter or range-based allocation. Use a distributed database like Cassandra or a custom service with sharded counters. Partition the keyspace into ranges and assign to nodes.

3. Ensure Durability and Exactly-Once Issuance

Use write-ahead logging, replication, and idempotent operations. Implement a two-phase commit or consensus protocol (e.g., Raft) to guarantee exactly-once issuance.

4. Handle Coordination and Leader Election

Use a coordination service like etcd or ZooKeeper for leader election and metadata management. Ensure high availability and fault tolerance.

5. Scale for Doubled Keyspace

Explain how to extend the design: add more partitions, rebalance ranges, and possibly use a hierarchical allocation scheme to handle 66 trillion keys.

Key Points to Mention

  • Partitioning strategies (range-based, hash-based) and their trade-offs
  • Use of distributed counters or sequence generators (e.g., Snowflake IDs, segmented counters)
  • Durability mechanisms: replication, write-ahead logs, consensus protocols (Raft, Paxos)
  • Exactly-once semantics via idempotency and transactional guarantees
  • Coordination services (etcd, ZooKeeper) for leader election and configuration
  • Scalability: adding nodes, rebalancing, and handling increased keyspace

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.