Engineering / advanced
Object storage from metadata to repair
Examine object placement, multipart protocols, erasure coding, integrity checks, and repair under explicit consistency and durability contracts.
Your route through engineering
Follow the path or begin with a question. Every lesson is open.
Your place, at your pace. Progress is optional and saved only in this browser when available.
Read / explore / reason
Inside this module
Module 01 advanced
Object storage from metadata to repair
Examine object placement, multipart protocols, erasure coding, integrity checks, and repair under explicit consistency and durability contracts.
- 01
An object name selects data within a defined namespace
Separate bucket/key identity, object metadata, backend placement, and single-key consistency from filesystem and multi-object transaction assumptions.
24 min - 02
Multipart upload separates staging from publication
Track upload IDs, numbered parts, completion results, ambiguous responses, and cleanup races without mistaking staged bytes for a completed object.
25 min - 03
Put the precondition inside the object write
Use supported atomic If-Match and If-None-Match writes while treating ETags as opaque validators rather than universal hashes or version histories.
25 min - 04
Redundant fragments recover known erasures under a code model
Calculate k+m storage overhead and reconstruction thresholds while separating known missing fragments from undetected corruption and operational write policy.
27 min - 05
Fragment placement determines which failures overlap
Connect Ceph placement groups and CRUSH topology to host, rack, and shared-dependency failures without assuming distinct disks imply independent risk.
26 min - 06
Detecting corrupt bytes is different from repairing them
Separate transfer checksums, stored-data scrubs, authoritative versions, and repair evidence while budgeting the cost of reading cold data.
25 min - 07
Repair must restore the intended committed generation
Compare access-triggered and background repair, establish authoritative version evidence, and budget reconstruction traffic without assuming repair is free.
26 min - 08
Deleting the current name need not erase retained versions
Track S3 version-qualified data and delete markers, distinguish visibility from physical retention, and recover deliberately without assuming versioning is a backup by itself.
25 min - 09
Eligibility, execution, and protected retention are separate states
Reason about asynchronous lifecycle actions and version-scoped Object Lock while distinguishing technical storage controls from a complete retention policy.
26 min - 10
Capacity must leave room to restore redundancy
Budget logical payload, encoding, retained state, failure-domain reserve, and repair bandwidth without deriving product durability from raw redundancy counts.
27 min