引導式測驗 · 中等 等級
預配置吞吐量與即用即付測驗
使用工作負載形狀、利用率、條款和測量的服務結果來比較特定於提供者的預留容量和按需容量。
問題 1 的 8
How do pay-as-you-go and provisioned-throughput offers generally differ?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- How do pay-as-you-go and provisioned-throughput offers generally differ?
- They always cost exactly the same
- Provisioned throughput has no contract or commitment
- Pay-as-you-go bills usage; provisioned throughput commits to reserved capacity under provider terms
- Pay-as-you-go guarantees dedicated capacity
- Why might a reserved capacity plan fit a steady workload?
- It guarantees every request is instant
- It is always cheaper at any usage level
- It removes the need to forecast
- A predictable baseline can use committed capacity consistently
- What can happen with shared pay-as-you-go capacity during demand spikes?
- All requests receive reserved hardware
- Temporary contention or quota errors may occur, depending on provider behavior
- Latency becomes fixed
- Every provider returns the same behavior
- Does “provisioned throughput” universally guarantee per-request latency?
- No; read the provider’s specific service terms and performance guarantees
- Yes, for all models and regions
- Yes, because capacity is reserved
- No provider offers throughput reservations
- What should a cost comparison include?
- Only the token price
- Only the reservation monthly amount
- Utilization, peak demand, commitment terms, overages, and quota behavior
- Only average traffic
- What does Google Vertex AI’s Dynamic Shared Quota describe?
- A fixed dedicated reservation for every user
- Shared resources allocated based on real-time availability and demand
- A guaranteed response-time SLA
- An output token pricing formula
- What can help decide whether to purchase a reservation?
- A single request test
- A comparison of product names
- An assumption that peak equals average
- A production-like pilot with measured utilization and latency
- Why should reserved and pay-as-you-go product details be checked for each provider?
- Units, commitments, supported models, overages, and capacity behavior differ
- All cloud providers use identical plans
- The term provisioned has a legal definition shared by vendors
- Only the model name matters