cmaven

작업노트

:bulb: A summary of Object Storage.

[01] Definition

A storage system that assigns a unique ID (identifiable by the user or server) to data regardless of its physical location, stores it in a container or bucket, and retrieves it on demand using that ID.

  • Modern data is split into structured data (classifiable by some criteria: student ID, name, age, etc.) and unstructured data (typically large images, videos, etc.) that can’t be neatly classified.
  • Existing storage methods — file systems (file-based, hierarchical) and block storage (block-based with sectors/tracks) — suit structured data. Object storage suits unstructured data.
  • Analogy: a valet parking service. If the data object is your car, the unique ID is your receipt. You hand over the receipt and you get your car back — you don’t have to know where it was parked.

[02] Advantages

2-1. Scalability

  • No need to partition
  • Holds data regardless of capacity

2-2. Efficiency

  • No hierarchical directory system → no inter-layer bottleneck

2-3. Availability

  • Auto-replication and rolling updates supported
  • No downtime

[03] How It Works (IDA)

3-1. IDA (Information Dispersal Algorithm)

  • Splits data into pieces
  • Distributes pieces across local or globally distributed storage nodes over the network
  • The distributed nodes form a single namespace storage

:small_blue_diamond: Reference: IBM Tech Forum — Object Storage

참고