Create long-running applications

Question 1

Designing reliable applications

Answer 1

1) Reliability is the probability that a system functions correctly during any given period of time.
2) Azure helps you to handle these challenges.

Question 2

Fault domains and update domains

Answer 2

1) hardware failures are unavoidable

2) To help you to cope with such failures, Azure introduces two concepts: fault domain and update domain.

Question 3

Fault domain

Answer 3

1) This is a group of resources that can fail at the same time
2) you can distribute service instances evenly to multiple fault domains so that all service instances won’t fail at the same time due to hardware failures.

Question 4

Update domain

Answer 4

1) An update domain is a logical group of resources that can be simultaneously updated during system upgrades. 2) When Azure updates a service, it doesn’t bring down all instances at the same time.
3) Instead, it performs a rolling update via an update domain walk.
4) Service instances in different update domains are brought down group by group for updates.

Question 5

Transient errors

Answer 5

Transient errors are caused by some temporal conditions such as network fluctuation, service overload, and request throttling. Transient errors are quite elusive; they happen randomly and can’t be reliably re-created. A typical way to handle transient error is to retry a couple of times

Question 6

Loose coupling

Answer 6

1) dynamic scaling and load leveling.
2) don’t have direct dependencies on one another
3) failing component won’t produce the ripple effect on the entire system
4) integrated by Azure Service Bus queue

Question 7

Health monitoring

Answer 7

1) Azure Diagnostics and Application Insights

Question 8

workload of an application

Answer 8

two typical workload change patterns: gradual changes (scaling up, out) and spikes.

Question 9

Dynamic scaling

Answer 9

1) Predicable (weekend demand)

2) UnPredicable news website might experience unexpected spikes when breaking news occurs

Question 10

Two major autoscaling methods

Answer 10

1) Scheduled scaling This is suitable for expected workload changes.
2) Reactive scaling Reactive scaling is suitable for unexpected workload changes. It monitors certain system metrics and adjusts system capacity when those metrics attain certain thresholds.

Question 11

A practical challenge of reactive scaling

Answer 11

is the latency in provisioning new resources. Provisioning a new VM and deploying a new service instance need time. So, when you design your reactive-scaling solution, you need to leave enough space for new resources to be brought online

Question 12

Containers to avoid latency

Answer 12

1) Container technologies such as Docker make it possible for you to package workloads in light-weight images, which you can deploy and activate very quickly.
2) Such agility affords new possibilities to reactive scaling by eliminating the needs to account for latencies.

Question 13

Workload partitioning

Answer 13

1) total workload is sliced into small portions, and each portion is assigned to a number of designated instances. 2) There are several advantages to using workload partitioning compared to homogeneous instances, among them is tenant isolation.

Question 14

Tenant isolation

Answer 14

you can route workloads for a certain tenant to a designated group of instances instead of being randomly routed to any instances in a bigger instance pool

Question 15

Tenant isolation new scenarios

Answer 15

1) per-tenant monitoring
2) tiered service offerings,
3) independent updates to different tenants.

Question 16

dynamic partioning

Answer 16

1) dynamic scaling.
2) When a new node is joined to a cluster, the new node takes on a fair share of the workload without impacting more running nodes than is necessary.
3) Techniques such as consistent hashing are commonly used to facilitate such workload relocation

Question 17

Cloud Services

Answer 17

1) (PaaS) offering for building and hosting cloud-based applications
2) It’s designed for building n-Tiered applications in the cloud
3) When the application is deployed to Azure, it might become a SaaS application that is accessible to all authenticated users across the globe.

Question 18

Role endpoints

Answer 18

1) Input Endpoints,
2) Internal Endpoints, and
3) Instance Input Endpoints

Question 19

Input Endpoints

Answer 19

1) Accessed openly over the Internet.
2) The Input Endpoints of a role point to an Azure-provided load balancer.
3) User requests are distributed to role instances by the load balancer.
4) The public port is what the load balancer exposes to service consumers;
5) the private port is what the load balancer uses to communicate with role instances.
6) Input Endpoints support HTTP, HTTPS, TCP, and UDP.

Question 20

Internal Endpoints

Answer 20

1) are private to the cloud service.
2) They are used for role instances to communicate with one another.
3) Internal Endpoints are not load-balanced.
4) support HTTP, HTTPS, TCP, and ANY.

Question 21

Instance Input Endpoints

Answer 21

1) These are publicly accessible endpoints with port ranges.
2) A service consumer can directly access different instances by choosing different ports in the given range. For instance, an Instance Input Endpoint with port range from 8008 to 8012 corresponds to 5 instances, with 8008 mapped to the first one, 8009 mapped to the second one, and so on.
3) Instance Input Endpoints support TCP and UDP.

Question 22

Access cloud service

Answer 22

1) By default, all cloud services are deployed to the cloudapp.net domain.
2) service1.cloudapp.net.
3) http://service1.cloudapp.net will be mapped to the role that has an HTTP-based Input Endpoint at port 80.

Question 23

Cloud Services availability features

Answer 23

1) Built-in load balancer
2) Rolling updates
3) Swap deployment

Question 24

Cloud Services reliability

Answer 24

1) Automatic instance recovery
2) Built-in diagnostics
3) Integration with Application Insight
4) Multiregion deployment with Traffic Manager Use this to fail-over from a primary site to a secondary site when needed.

Question 25

scalability

Answer 25

1) Planned or reactive autoscale.

2) Independent adjustments of instance numbers of different roles.