- LPU MCA Semester 1
- CAP470
- Unit 2
CAP470 · Unit 2: Cloud computing architecture
Cloud infrastructure performance, networking and resource management
10 practice questions · in the MTE and ETE syllabus
Key points
- Latency is delay, bandwidth is capacity
- 99.9 percent availability allows about 43 minutes a month
- Load balancers and auto-scaling manage demand
- CDNs reduce latency by caching near users
Key performance measures are latency (delay before a response begins, in milliseconds), bandwidth (maximum data rate of a link), throughput (data actually delivered per second) and availability (percentage of time a service is up).
An SLA (service level agreement) states promised targets and remedies, such as service credits. Downtime allowed per 30-day month (43,200 minutes): 99 percent about 7.2 hours, 99.9 percent about 43 minutes, 99.99 percent about 4.3 minutes.
Resource management techniques:
- Load balancing spreads requests across servers.
- Auto-scaling adds and removes capacity with demand.
- Provisioning can be static (fixed) or dynamic (on demand).
- CDNs cache content at edge locations near users to cut latency.
- Over-provisioning wastes money, while under-provisioning hurts performance.
Networking in clouds relies on virtual networks, subnets and software-defined networking (SDN).
Exam trap: high bandwidth does not guarantee low latency, because distance and routing add delay.
Try two questions
What is an SLA in cloud computing?
- A software license activation key tied to one single machine only
- A contract stating promised performance and uptime
- A storage layer allocation table for the disk blocks of a volume
- A secure login assertion used for signing on to a remote account
Show the answer
B. An SLA states measurable service targets such as uptime and the remedies if they are missed.
Why not A: An SLA is a contract about service, not a license key.
Why not C: That is a storage concept, not an SLA.
Why not D: That is an authentication mechanism, not an SLA.
Why is resource over-provisioning a problem in the cloud?
- It makes the provider remove all of the customer's data
- It wastes money on capacity that is not used
- It stops virtual machines from ever starting
- It increases network latency to the nearest router
Show the answer
B. Provisioning more than needed means paying for idle resources.
Why not A: Over-provisioning does not cause data removal.
Why not C: Over-provisioned machines still start.
Why not D: Idle capacity does not change latency.