What is Tiered Storage for Hot and Cold Data?
Learn how tiered storage automatically moves hot and cold data between fast and cheap storage tiers to cut cost.
Expected Interview Answer
Tiered storage places data on different classes of storage medium based on how frequently it is accessed — hot data (frequently read/written) sits on fast, expensive storage like NVMe SSDs, while cold data (rarely accessed) is moved to slower, cheaper storage like HDDs or object storage, cutting overall cost without sacrificing performance where it matters.
Most datasets follow an access pattern where a small fraction of rows (recent orders, active user sessions) are read constantly while the majority (records from years ago) are almost never touched. A tiering policy monitors access frequency or recency and automatically migrates data between tiers — for example, moving a partition of orders older than 90 days from SSD to an object-storage tier — so the expensive fast tier only holds data that actually benefits from its speed. Queries against cold data still work, just with higher latency, since the system transparently fetches from the slower tier when needed.
- Cuts storage cost by keeping only actively used data on expensive fast media
- Preserves fast performance for the hot working set
- Scales cheaply for cold/archival data using object storage
- Automates migration based on access recency or frequency, no manual intervention
AI Mentor Explanation
A cricket club keeps this season’s match videos on a fast local drive at the ground so coaches can pull up any recent game instantly during analysis sessions, but videos from five seasons ago sit in a cheap offsite archive that takes a day to retrieve. Nobody wants to pay for expensive fast storage for footage nobody watches anymore. Tiered storage applies this same logic to a database: recently accessed "hot" data stays on fast, costly storage, while old "cold" data moves to cheap, slower storage.
Step-by-Step Explanation
Step 1
Classify access patterns
Track how frequently and recently each partition or row range is read or written.
Step 2
Define tiering policy
Set thresholds, e.g. data older than 90 days with no recent reads is eligible for demotion.
Step 3
Migrate data between tiers
Automatically move cold partitions from fast SSD storage to cheaper HDD or object storage.
Step 4
Serve queries transparently
Queries against cold data still succeed, fetched from the slower tier with higher latency as needed.
What Interviewer Expects
- Clear hot vs cold definition based on access recency/frequency
- Understanding that tiering is a cost-vs-latency trade-off, not a correctness feature
- Awareness of automated migration policies (e.g. time-based partition aging)
- Mention that cold data remains queryable, just slower
Common Mistakes
- Assuming cold data becomes inaccessible rather than just slower
- Confusing tiered storage with sharding or replication
- Ignoring the need for a policy/threshold to trigger migration
- Not considering that access patterns can shift, requiring re-promotion of data
Best Answer (HR Friendly)
“Tiered storage means keeping frequently accessed "hot" data on fast, expensive storage and moving rarely accessed "cold" data to cheaper, slower storage. It is an automated cost-optimization strategy — data that nobody queries anymore does not need to sit on premium infrastructure, but it is still available, just retrieved a bit more slowly when needed.”
Code Example
-- Partition orders by month, then assign old partitions to a cheap tablespace
CREATE TABLE orders (
order_id BIGINT,
order_date DATE,
total NUMERIC
) PARTITION BY RANGE (order_date);
CREATE TABLE orders_2026_01 PARTITION OF orders
FOR VALUES FROM ('2026-01-01') TO ('2026-02-01')
TABLESPACE fast_ssd;
CREATE TABLE orders_2019_01 PARTITION OF orders
FOR VALUES FROM ('2019-01-01') TO ('2019-02-01')
TABLESPACE cold_archive;
-- A scheduled job moves aging partitions from fast_ssd to cold_archiveFollow-up Questions
- How do you decide the threshold between hot and cold data?
- What happens to query latency when data lives on the cold tier?
- How does tiered storage interact with partitioning strategy?
- Can data be re-promoted from cold back to hot storage if access patterns change?
MCQ Practice
1. In tiered storage, "hot" data refers to:
Hot data is accessed frequently and kept on fast, more expensive storage to maintain low latency.
2. What is the main goal of moving cold data to a cheaper storage tier?
Tiering cuts cost by storing rarely accessed data cheaply, while still allowing it to be queried, just more slowly.
3. A typical trigger for demoting data from hot to cold storage is:
Tiering policies commonly key off data age or recency/frequency of access to decide when to migrate data to a cheaper tier.
Flash Cards
What is hot data? — Frequently accessed data kept on fast, more expensive storage for low latency.
What is cold data? — Rarely accessed data moved to cheaper, slower storage to save cost.
Is cold data still queryable? — Yes, just with higher latency since it is fetched from slower storage.
What triggers tier migration? — Policies based on data age or access recency/frequency automatically move data between tiers.