Pub/sub vs streams comparison in Redis - Performance Comparison

Time Complexity: Pub/sub vs streams comparison

O(s)

Understanding Time Complexity

When using Redis for messaging, it is important to understand how the time to process messages grows as more messages or subscribers are involved.

We want to see how the cost changes when using Pub/Sub versus Streams as the number of messages or consumers increases.

Scenario Under Consideration

Analyze the time complexity of these Redis commands for message delivery.

# Pub/Sub publish
PUBLISH channel message

# Streams add and read
XADD mystream * message value
XREAD COUNT 10 STREAMS mystream last_id
    

Pub/Sub sends messages directly to all subscribers. Streams store messages and let consumers read at their own pace.

Identify Repeating Operations

Look at what repeats when messages are sent or read.

• Primary operation: For Pub/Sub, sending message to each subscriber; for Streams, appending message and reading batches.
• How many times: Pub/Sub sends once per subscriber; Streams append once and consumers read independently.

How Execution Grows With Input

Consider how work grows as subscribers or messages increase.

Input Size (n)	Pub/Sub Operations	Streams Operations
10 subscribers	10 sends per message	1 append + 10 independent reads
100 subscribers	100 sends per message	1 append + 100 independent reads
1000 subscribers	1000 sends per message	1 append + 1000 independent reads

Pub/Sub cost grows directly with number of subscribers per message. Streams cost grows with number of consumers reading messages separately.

Final Time Complexity

Time Complexity: O(s) where s is number of subscribers or consumers.

This means the work to deliver messages grows linearly with how many clients receive or read them.

Common Mistake

[X] Wrong: "Pub/Sub and Streams have the same cost regardless of subscribers."

[OK] Correct: Pub/Sub sends messages to every subscriber immediately, so cost grows with subscribers. Streams store once but each consumer reads separately, so cost depends on consumers reading.

Interview Connect

Understanding how message delivery scales helps you choose the right Redis feature and explain your reasoning clearly in conversations.

Self-Check

What if we batch multiple messages in one XADD call for Streams? How would that affect the time complexity?

Practice

1. Which Redis feature is best suited for instant message delivery without storing messages?

easy

A. Hashes

B. Streams

C. Pub/Sub

D. Sorted Sets

Solution

1. Step 1: Understand Pub/Sub purpose

   Pub/Sub delivers messages instantly to subscribers without saving them.

2. Step 2: Compare with Streams

   Streams store messages for later processing, unlike Pub/Sub.

3. Final Answer:

   Pub/Sub -> Option C

4. Quick Check:

   Instant delivery without storage = Pub/Sub [OK]

Hint: Instant delivery without storage means Pub/Sub [OK]

Common Mistakes:

• Confusing Streams with Pub/Sub for instant delivery
• Thinking Streams do not store messages
• Choosing data structures unrelated to messaging

2. Which of the following is the correct Redis command to publish a message to a channel named news?

easy

A. PUBLISH news "Hello"

B. XADD news * message "Hello"

C. SUBSCRIBE news

D. GET news

Solution

1. Step 1: Identify publish command syntax

   The PUBLISH command sends a message to a channel immediately.

2. Step 2: Check other commands

   XADD adds to streams, SUBSCRIBE listens, GET retrieves keys, so only PUBLISH fits.

3. Final Answer:

   PUBLISH news "Hello" -> Option A

4. Quick Check:

   Publish message = PUBLISH command [OK]

Hint: Use PUBLISH to send messages to channels [OK]

Common Mistakes:

• Using XADD instead of PUBLISH for Pub/Sub
• Confusing SUBSCRIBE with PUBLISH
• Using GET which is for keys, not messages

3. Given the following Redis commands, what will be the output of XREAD COUNT 2 STREAMS mystream 0 after adding two entries with XADD mystream * field1 val1 and XADD mystream * field2 val2?

medium

A. An error because COUNT is invalid here

B. Two entries with their IDs and fields returned

C. No entries returned because 0 means no messages

D. Only the first entry returned ignoring COUNT

Solution

1. Step 1: Understand XADD and XREAD

   XADD adds entries to a stream with unique IDs. XREAD reads entries from streams.

2. Step 2: Analyze XREAD COUNT 2 STREAMS mystream 0

   This reads up to 2 entries from 'mystream' starting from ID 0 (the beginning).

3. Final Answer:

   Two entries with their IDs and fields returned -> Option B

4. Quick Check:

   XREAD with COUNT 2 returns up to 2 entries [OK]

Hint: XREAD with COUNT returns that many entries from stream start [OK]

Common Mistakes:

• Thinking COUNT is invalid in XREAD
• Assuming 0 means no messages returned
• Expecting only one entry despite COUNT 2

4. You wrote this Redis command to subscribe to a channel: SUBSCRIBE mychannel. But your client never receives messages published with PUBLISH mychannel "Hello". What is the likely problem?

medium

A. The client must use XREAD to receive Pub/Sub messages

B. PUBLISH command syntax is wrong

C. Streams must be used instead of Pub/Sub for messages

D. The client is not listening because it did not call SUBSCRIBE correctly

Solution

1. Step 1: Verify SUBSCRIBE usage

   SUBSCRIBE must be called on the client to listen to messages; if not, no messages arrive.

2. Step 2: Check other options

   PUBLISH syntax is correct; Streams and XREAD are unrelated to Pub/Sub listening.

3. Final Answer:

   The client is not listening because it did not call SUBSCRIBE correctly -> Option D

4. Quick Check:

   Client must SUBSCRIBE to receive messages [OK]

Hint: Client must SUBSCRIBE to receive published messages [OK]

Common Mistakes:

• Thinking PUBLISH syntax is wrong
• Confusing Streams with Pub/Sub for message delivery
• Using XREAD to get Pub/Sub messages

5. You want to build a task queue in Redis where tasks must be stored reliably and processed even if the worker restarts. Which Redis feature should you use and why?

hard

A. Use Streams because they store messages and support replaying tasks

B. Use Sorted Sets because they keep tasks ordered by score

C. Use Pub/Sub because it delivers messages instantly without storage

D. Use Hashes because they store key-value pairs efficiently

Solution

1. Step 1: Identify requirements for task queue

   Tasks must be stored reliably and processed even after worker restarts.

2. Step 2: Compare Redis features

   Pub/Sub does not store messages, so tasks can be lost. Streams store messages and allow replay.

3. Final Answer:

   Use Streams because they store messages and support replaying tasks -> Option A

4. Quick Check:

   Reliable storage and replay = Streams [OK]

Hint: Reliable task queues need Streams, not Pub/Sub [OK]

Common Mistakes:

• Choosing Pub/Sub for reliable storage
• Confusing Sorted Sets with message storage
• Using Hashes which are not for message queues