L16 - Cloud Computing 2

Question 1

What are the 10 obstacles of using Cloud Computing?

Answer 1

• Business Continuity and Service Availability
• Data lock in
• Data Confidentiality / Auditability
• Data Transfer Bottlenecks
• Performance Unpredictability
• Scalable Storage
• Bugs in Large-Scale Distributed Systems
• Scaling Quickly
• Reputation Fate Sharing
• Software Licensing

Question 2

Explain ‘Business Continuity and Service Availability’ in Cloud Computing

Answer 2

• Cloud computing means handing data and operational capacity over to service providers
• Could cause big performance and data confidentiality issues if cloud provider goes down.

Question 3

If Cloud Provider goes down, why is the knock on effect on the user usually not as bad as opposed to if using a local server?

Answer 3

Because cloud provider will handle restoration processes for the users

Question 4

Explain ‘Data Lock-in’ in Cloud Computing…

Answer 4

• The fact that when a company uses one cloud provider to store their data, it’s extremely hard to change cloud provider.
• Companies worry that their data is only available in one providers cloud.

Question 5

Why does data lock-in occur?

Answer 5

Because there are no standard API’s that enables interaction between discrete cloud providers.

Question 6

Explain the obstacle of ‘Data Confidentiality / Auditability’ in Cloud Computing. How can it be overcome?

Answer 6

• How can we be sure that our data is being kept confidential? How can we tell industry regulators that our data is safe?
• Confidentiality: Cloud providers use encryption and firewalls to guard against malicious agents.
• Auditability: Ensure data can be audited correctly through region locking into a cloud provider

Question 7

Explain the obstacle of ‘Data Transfer Bottlenecks’ in Cloud Computing. How can we overcome it?

Answer 7

• Huge data transfer is expensive and slow
• Edge Caching → Cloud providers cache data locally to improve download response time. For example, Netflix uses this for efficient streaming.

Question 8

Explain the obstacle of ‘Performance Unpredictability’ in Cloud Computing. How can we overcome it?

Answer 8

• Companies are worried about cloud performance unpredictability due to sharing storage and network
• This is most often an issue in network sharing and disk storage sharing.
• Disks are bottleneck devices (serialised) which can cause poor performance with heavy load.
• Sharing CPU’s → Works well
• Sharing GPU’s and Networks → Not so well
• Therefore, sharing VM’s (CPU’s) reduces unpredictability

Question 9

Explain the obstacle of ‘Scalable Storage’ in Cloud Computing. How can we overcome it?

Answer 9

• Will cloud storage scale as per the data requirements?
• Scaling up → works well on both local and cloud data storages.
• Scaling down → Due to fragmentation, scaling down can be awkward and difficult. Fragmentation means we would need to move clusters of data around when scaling down.

Question 10

In Scalable Storage, what is the fragmentation issue?

Answer 10

• Occurs when scaling down
• Fragmentation means we would need to move clusters of data around when scaling down

Question 11

Explain the obstacle of ‘Bugs in Large-Scale Distributed Systems’ in Cloud Computing. How can we overcome it?

Answer 11

• Some companies felt using cloud computing would lead to Heisenberg Bugs…
• 1 thread → Simple to track down bugs.
• Multi-threaded → More challenging to locate bugs. E.g Thousands of processes, each running hundreds of threads (Heisenberg Bugs).
• Cloud services use logs to communicate which offers traceback stacks to identify the time and trigger of a bug.

Question 12

What is a Heisenberg Bug?

Answer 12

A challenging bug associated with multi-threaded systems that consist of huge numbers of processes. Typically occurs in cloud computing.

Question 13

Explain the obstacle of ‘Scaling Quickly’ in Cloud Computing. How can we overcome it?

Answer 13

• Can the cloud keep up with rapid, huge growth such as seen in social networking start ups?
• Cloud providers are almost effectively infinite → Can continuously scale provided you pay for the service.
• Cloud scales horizontally

Question 14

What does it mean that Cloud Providers avoid cold starts? Why is this important?

Answer 14

• Cloud providers create containers ahead of time to ensure users don’t encounter cold starts on VM’s.
• Cold Starts are slow, cumbersome start-up processes. Thus, cloud providers ensure this is done beforehand, improving efficiency and experience for end user.

Question 15

Explain the obstacle of ‘Reputation Sharing’ in Cloud Computing. How can we overcome it?

Answer 15

• Cloud users were worried that their reputation relies on, and could be ruined by, other cloud users. For example, if a cloud provider user sends spam email, other users who send emails from the cloud provider may be blocked by customers who have put a blanket block on all emails from the cloud provider.
• Try to help users use cloud services in a customer friendly way

Question 16

Explain the obstacle of ‘Software Licensing’ in Cloud Computing. How can we overcome it?

Answer 16

• Ambiguity around software licensing. E.g If you have license for software, can I run it on only one VM or multiple VM’s? etc.
• Azure → Microsoft is looking very powerful in the cloud due to comprehensive licensed solutions.