Compute Availability

The GCP offers several compute services. We'll consider availability in four of these services.

• Compute Engine
• Kubernetes Engine
• App Engine
• Cloud Functions

Each of these services can provide high availability compute resources, but they vary in the amount of effort required to achieve high availability.

Storage Availability

Highly available storage is storage that is available and functional at nearly all times. The storage services can be grouped into the following categories:

• Object storage
• File and block storage
• Database services
• Caching

Let's look at availability in each type of storage service.

Network Availability

When network connectivity is down, applications are unavailable. There are two primary ways to improve network availability:

• Use redundant network connections
• Use Premium Tier networking

Redundant network connections can be used to increase the availability of the network between an on-premises data center and Google's data center. One type of connection is a Dedicated Interconnect, which can be used with a minimum of 10 Gbps throughput and does not traverse the public internet. A Dedicated Interconnect is possible when both your network and the Google Cloud network have a point of presence in a common location, such as a data center. When your network does not share a common point of presence with the Google Cloud network, you have the option of using a Partner Interconnect. When using a Partner Interconnect, you provision a network link between your data center and a Google network point of presence. Traffic flows through a telecommunication provider's network from your data center to Google Cloud's network. Traffic does not travel over the internet.

VPNs can also be used when sending data over the internet is not a problem. You should choose among these options based on cost, security, throughput, latency, and availability considerations. Google Cloud offers a high availability VPN, known as HA VPN, which uses redundant connections and offers a 99.99 percent SLA.

Data within the GCP can be transmitted among regions using the public internet or Google's internal network. The latter is available as the Premium Network Tier, which costs more than the Standard Network Tier, which uses the public internet. The internal Google network is designed for high availability and low latency, so the Premium Tier should be considered if global network availability is a concern. Note, if you plan to use global load balancing, you will need to use Premium Tier networking.

Application Availability

Application availability builds on compute, storage, and networking availability. It also depends on the application itself. Designing software for high availability is beyond the scope of this book, and it is not a subject you will likely be tested on when taking the Professional Cloud Architect exam.

Architects should understand that they can use Cloud Monitoring and Cloud Logging to observe the state of applications so that they can detect problems as early as possible. Applications that are instrumented with custom metrics can provide application-specific details that could be helpful in diagnosing problems with an application.

Scaling Compute Resources

Compute Engine and Kubernetes Engine support automatic scaling of compute resources. App Engine and Cloud Functions autoscale as well, but they are managed by the Google Compute Platform.

Scaling Storage Resources

Storage resources are virtualized in GCP, and some are fully managed services, so there are parallels between scaling storage and compute resources.

The least scalable storage system is locally attached SSDs on VMs. Locally attached storage is not considered a persistent storage option. Data will be retained during reboots and live migrations, but it is lost when the VM is terminated or stopped. Local data is lost from preemptible VMs when they are preempted.

Zonal and regional persistent disks and persistent SSDs can currently scale up to 64 TB per VM instance. You should also consider read and write performance when scaling persistent storage. Standard disks have a maximum sustained read IO operations per second (IOPS) of 0.75 per gigabyte and write IOPS of 1.5 per gigabyte. Persistent SSDs have a maximum sustained read and write IOPS of 30 per gigabyte. As a general rule, persistent disks are well suited for large-volume batch processing when low-cost and high-storage volume are important. When performance is a consideration, such as when running a database on a VM, persistent SSDs are the better option.

Adding storage to a VM is a two-step process. You will need to allocate persistent storage and then issue operating system commands to make the storage available to the filesystem. The commands are operating system specific.

Managed services, such as Cloud Storage and BigQuery, ensure that storage is available as needed. In the case of BigQuery, even if you do not scale storage directly, you may want to consider partitioning data to improve query performance. Partitioning organizes data in a way that allows the query processor to scan smaller amounts of data to answer a query. For example, assume that Mountkirk Games is storing summary data about user sessions in BigQuery. The data includes a date indicating the day that the session data was collected. Analysts typically analyze data at the week and month levels. If the data is partitioned by week or month, the query processor would scan only the partitions needed to answer the query. Data that is outside the date range of the query would not have to be scanned. Since BigQuery charges by the amount of data scanned, this can help reduce costs.

Network Design for Scalability

Connectivity between on-premises data centers and Google data centers doesn't scale the way storage and compute scales. You need to plan ahead for what is the upper limit of what will be needed. You should plan for peak capacity, although you may only pay for bandwidth used depending on your provider.

Measuring Reliability

There are different ways to measure reliability, but some are more informative than others.

Total system uptime is one measure. This sounds simple and straightforward, but it is not—at least when dealing with distributed systems. Specifically, what measure do you use to determine whether a system is up? If at least one server is available, is a system up? If there was a problem with an instance group in Compute Engine or the pods in a Kubernetes deployment, you may be able to respond to some requests but not others. If your definition of uptime is based on just having one or some percentage of desired VMs or pods running, then this may not accurately reflect user experience regarding reliability.

Rather than focus on the implementation metrics, such as the number of instances available, reliability is better measured as a function of the work performed by the service. The number of requests that are successfully responded to is a good basis for measuring reliability. Successful request rate is the percentage of all application requests that are successfully responded to. This measure has the advantage of being easy to calculate and of providing a good indicator for the user experience.

Reliability Engineering

As an architect, you should consider ways to support reliability early in the design stage. This should include the following:

• Identifying how to monitor services. Will they require custom metrics?
• Considering alerting conditions. How do you balance the need for early indication that a problem may be emerging with the need to avoid overloading DevOps teams with unactionable alerts?
• Using existing incident response procedures with the new system. Does this system require any specialized procedures during an incident? For example, if this is the first application to store confidential, personally identifying information, you may need to add procedures to notify the information security team if an incident involves a failure in access controls.
• Implementing a system for tracking outages and performing post-mortems to understand why a disruption occurred.

Designing for reliability engineering requires an emphasis on organizational and management issues. This is different than designing for high availability and scalability, which is dominated by technical considerations. As an architect, it is important to remember that your responsibilities include both technical and management aspects of system design.

1. You are advising a customer on how to improve the availability of a data storage solution. Which of the following general strategies would you recommend?
   1. Keeping redundant copies of the data
   2. Lowering the network latency for disk writes
   3. Using a NoSQL database
   4. Using Cloud Spanner
2. A team of data scientists is analyzing archived data sets. Their statistical model building procedures run in batches. If the model building system is down for up to 30 minutes per day, it does not adversely impact the data scientists' work. What is the minimal percentage availability among the following options that would meet this requirement?
   1. 99.99 percent
   2. 99.90 percent
   3. 99.00 percent
   4. 99.999 percent
3. Your development team has recently triggered three incidents that resulted in service disruptions. In one case, an engineer mistyped a number in a configuration file and in the other cases specified an incorrect disk configuration. What practices would you recommend to reduce the risk of these types of errors?
   1. Continuous integration/continuous deployment
   2. Code reviews of configuration files
   3. Vulnerability scanning
   4. Improved access controls
4. Your company is running multiple VM instances that have not had any downtime in the past several weeks. Recently, several of the physical servers suffered disk failures. The applications running on the servers did not have any apparent service disruptions. What feature of Compute Engine enabled that?
   1. Preemptible VMs
   2. Live migration
   3. Canary deployments
   4. Redundant array of inexpensive disks
5. You have deployed an application on a managed instance group. Occasionally the application experiences an intermittent malfunction and then resumes normal operation. Which of these is a reasonable explanation for what is happening?
   1. The application shuts down when the instance group time-to-live (TTL) threshold is reached.
   2. The application shuts down when the health check fails.
   3. The VM shuts down when the instance group TTL threshold is reached and a new VM is started.
   4. The VM shuts down when the health check fails and a new VM is started.
6. An online gaming company is growing its user base in North America, Europe, and Asia. Executives are concerned that players in Europe and Asia will have a degraded experience if the game backend runs only in North America. What would you suggest to improve latency and game experience for users in Europe and Asia?
   1. Use Cloud Spanner to have a globally consistent, horizontally scalable relational database.
   2. Create instance groups running the game backend in multiple regions across North America, Europe, and Asia. Use global load balancing to distribute the workload.
   3. Use Standard Tier networking to ensure that data sent between regions is routed over the public internet.
   4. Use a Cloud Memorystore cache in front of the database to reduce database read latency.
7. What configuration changes are required to ensure high availability when using Cloud Storage or Cloud Filestore?
   1. A sufficiently long TTL must be set.
   2. A health check must be specified.
   3. Both a TTL and health check must be specified.
   4. Nothing. Both are managed services. GCP manages high availability.
8. The finance director at your company is frustrated with the poor availability of an on-premises finance data warehouse. The data warehouse uses a commercial relational database that only scales by buying larger and larger servers. The director asks for your advice about moving the data warehouse to the cloud and if the company can continue to use SQL to query the data warehouse. What GCP service would you recommend to replace the on-premises data warehouse?
   1. Bigtable
   2. BigQuery
   3. Cloud Datastore
   4. Cloud Storage
9. TerramEarth has determined that it wants to use Cloud Bigtable to store equipment telemetry received from vehicles in the field. It has also concluded that it wants two clusters in different regions. Both clusters should be able to respond to read and write requests. What kind of replication should be used?
   1. Primary–hot primary
   2. Primary–warm primary
   3. Primary–primary
   4. Primary read–primary write
10. Your company is implementing a hybrid cloud computing model. Line-of-business owners are concerned that data stored in the cloud may not be available to on-premises applications. The current network connection is using a maximum of 40 percent of bandwidth. What would you suggest to mitigate the risk of that kind of service failure?
    1. Configure firewall rules to improve availability.
    2. Use redundant network connections between the on-premises data center and Google Cloud.
    3. Increase the number of VMs allowed in Compute Engine instance groups.
    4. Increase the bandwidth of the network connection between the data center and Google Cloud.
11. A team of architects in your company is defining standards to improve availability. In addition to recommending redundancy and code reviews for configuration changes, what would you recommend including in the standards?
    1. Use of access controls
    2. Use of managed services for all compute requirements
    3. Use of Cloud Monitoring to alert on changes in application performance
    4. Use of Bigtable to collect performance monitoring data
12. Why would you want to run long-running, compute-intensive backend computation in a different managed instance group than on web servers supporting a minimal user interface?
    1. Managed instance groups can run only a single application.
    2. Managed instance groups are optimized for either compute or HTTP connectivity.
    3. Compute-intensive applications have different scaling characteristics from those of lightweight user interface applications.
    4. There is no reason to run the applications in different managed instance groups.
13. An instance group is adding more VMs than necessary and then shutting them down. This pattern is happening repeatedly. What would you do to try to stabilize the addition and removal of VMs?
    1. Increase the maximum number of VMs in the instance group.
    2. Decrease the minimum number of VMs in the instance group.
    3. Increase the time autoscalers consider when making decisions.
    4. Decrease the cooldown period.
14. A clothing retailer has just developed a new feature for its customer-facing web application. Customers can upload images of their clothes, create montages from those images, and share them on social networking sites. Images are temporarily saved to locally attached drives as the customer works on the montage. When the montage is complete, the final version is copied to a Cloud Storage bucket. The services implementing this feature run in a managed instance group. Several users have noted that their final montages are not available even though they saved them in the application. No other problems have been reported with the service. What might be causing this problem?
    1. The Cloud Storage bucket is out of storage.
    2. The locally attached drive does not have a filesystem.
    3. The users experiencing the problem were using a VM that was shut down by an autoscaler, and a cleanup script did not run to copy the latest version of the montage to Cloud Storage.
    4. The network connectivity between the VMs and Cloud Storage has failed.
15. Your development team has implemented a new application using a microservices architecture. You would like to minimize DevOps overhead by deploying the services in a way that will autoscale. You would also like to run each microservice in containers. What is a good option for implementing these requirements in Google Cloud Platform?
    1. Run the containers in Cloud Functions.
    2. Run the containers in Kubernetes Engine.
    3. Run the containers in Cloud Dataproc.
    4. Run the containers in Cloud Dataflow.
16. TerramEarth is considering building an analytics database and making it available to equipment designers. The designers require the ability to query the data with SQL. The analytics database manager wants to minimize the cost of the service. What would you recommend?
    1. Use BigQuery as the analytics database, and partition the data to minimize the amount of data scanned to answer queries.
    2. Use Bigtable as the analytics database, and partition the data to minimize the amount of data scanned to answer queries.
    3. Use BigQuery as the analytics database, and use data federation to minimize the amount of data scanned to answer queries.
    4. Use Bigtable as the analytics database, and use data federation to minimize the amount of data scanned to answer queries.
17. Line-of-business owners have decided to move several applications to the cloud. They believe the cloud will be more reliable, but they want to collect data to test their hypothesis. What is a common measure of reliability that they can use?
    1. Mean time to recovery
    2. Mean time between failures
    3. Mean time between deployments
    4. Mean time between errors
18. A group of business executives and software engineers are discussing the level of risk that is acceptable for a new application. Business executives want to minimize the risk that the service is not available. Software engineers note that the more developer time dedicated to reducing risk of disruption, the less time they have to implement new features. How can you formalize the group's tolerance for risk of disruption?
    1. Request success rate
    2. Uptime of service
    3. Latency
    4. Throughput
19. Your DevOps team recently determined that it needed to increase the size of persistent disks used by VMs running a business-critical application. When scaling up the size of available persistent storage for a VM, what other step may be required?
    1. Adjusting the filesystem size in the operating system
    2. Backing up the persistent disk before changing its size
    3. Changing the access controls on files on the disk
    4. Updating disk metadata, including labels