S3 (Simple Storage Service) Storage Classes and Performance#
Two things get tested: which storage class for which data (mostly a cost question - see s3-cost - but with performance implications) and how to move big data in and out fast.
Storage classes at a glance#
| Class | Durability / AZs (Availability Zone) | Access pattern | Min storage |
|---|---|---|---|
| S3 Standard | 99.99% / 3+ AZs | Frequent, low latency | - |
| Standard-IA | 99.99% / 3+ AZs | Infrequent but instant | 30 days |
| One Zone-IA | 99.5% / 1 AZ | Infrequent, re-creatable | 30 days |
| Intelligent-Tiering | 99.99% / 3+ AZs | Unknown/changing patterns | None (small monitoring fee/object) |
| Glacier Instant Retrieval | 99.99% / 3+ AZs | Archival, ms retrieval | 90 days |
| Glacier Flexible Retrieval | 99.99% / 3+ AZs | Archive, minutes-hours | 90 days |
| Glacier Deep Archive | 99.99% / 3+ AZs | Archive, 12-48h retrieval | 180 days |
Rules of thumb:
- “Unknown or changing access” → Intelligent-Tiering (no retrieval fees!)
- “Must be retrieved in milliseconds but rarely accessed” → Glacier Instant Retrieval
- “Regulatory 7-year retention” → Deep Archive + Object Lock
- “Reproducible data, cost critical, 1 AZ acceptable” → One Zone-IA
Performance fundamentals#
- Strong consistency for all operations (since Dec 2020) - read-after-write is guaranteed
- Request rates: 3,500 PUT/COPY/POST/DELETE and 5,500 GET/HEAD per prefix - scale by spreading prefixes
- Multipart upload: required above 5 GB, recommended above 100 MB; up to 10,000 parts; faster and resumable
- S3 Transfer Acceleration: upload via CloudFront edge locations over AWS backbone - “uploads from far-away regions are slow”
- Byte-range fetches: download a slice of an object - parallelize big reads
- S3 Batch Operations: run jobs (copy, tag, restore) on billions of objects
- Max object size 5 TB
flowchart LR
client[Client in Asia] -->|accelerated upload| edge[Nearest CloudFront edge]
edge -->|AWS backbone| s3[(S3 us-east-1)]Hands-on: aws-s3-objects, aws-s3-bucket.
S3 Replication (CRR & SRR) and Event Notifications#
Cross-Region (CRR) & Same-Region Replication (SRR) rules the exam tests:#
- S3 Versioning MUST be enabled on both the source and destination buckets.
- Existing objects are NOT replicated automatically! CRR/SRR only replicates new objects uploaded after the rule is enabled. To replicate objects that already existed, run an S3 Batch Replication job.
- Delete markers are NOT replicated by default (can be enabled optionally); permanent deletions (deleting a specific
versionId) are never replicated (protects against malicious deletion). - S3 Replication Time Control (S3 RTC): guarantees 99.99% of new objects replicate within 15 minutes (backed by an SLA) — pick this when a compliance question specifies a strict 15-minute replication SLA.
S3 Event Notifications vs Amazon EventBridge:#
- Standard S3 Event Notifications: send events (
s3:ObjectCreated:*) directly to SNS (Simple Notification Service), SQS (Simple Queue Service), or Lambda (filters only on prefix/suffix; delivery takes seconds to minutes). - S3 + Amazon EventBridge: enable EventBridge on the bucket when you need advanced JSON filtering (object size, metadata, tag), routing to 18+ AWS services (Step Functions, Kinesis, ECS…), or event archiving and replay.
Sınav Tuzağı (Exam Trap): “CRR açtık ama eski dosyalar karşı bölgede görünmüyor” sorusunun cevabı S3 Batch Replication‘dır. Ayrıca S3 performans soruları iki kalıba iner: “büyük dosya yükleme yavaş” = Multipart Upload + S3 Transfer Acceleration, “yüksek istek hızı (
503 Slow Down)” = prefix’leri dağıt.
Exam check#
Q1: Large video files (50 GB each) are uploaded from offices worldwide and uploads frequently fail on slow links. What improves reliability AND speed?
- A) Enable versioning
- B) Multipart upload with S3 Transfer Acceleration
- C) Switch to One Zone-IA
- D) Increase the bucket’s storage class
Cevabı göster
Answer: B. Multipart = resumable parallel parts; Transfer Acceleration = edge-based fast path.
Q2: Medical images must be retrievable in milliseconds but are rarely accessed, and kept for 10 years. Which class?
- A) Glacier Deep Archive
- B) Glacier Instant Retrieval
- C) S3 Standard
- D) One Zone-IA
Cevabı göster
Answer: B. “Milliseconds but rare” = Instant Retrieval. Deep Archive takes hours.
Ozet#
- Standard = sik erisim, IA = nadir ama aninda, One Zone-IA = tek AZ, yeniden uretilebilir
- Intelligent-Tiering = bilinmeyen erisim deseni, geri cekme ucreti yok
- Glacier Instant = milisaniye, Flexible = dakika/saat, Deep Archive = 12-48 saat
- Tutarlilik artik strong consistency - her islemde
- 3.500 yazar / 5.500 okuyucu per prefix = istek orani siniri
- Multipart upload buyuk dosyalarin hem hizi hem guvenilirligi
- Transfer Acceleration = dunyanin diger ucundan hizli yukleme
- 5 TB = tek obje limiti