Cloud Spanner global availability and read latency

PCD By islever Editorial Last updated

Storing and accessing data. Designing high-performing applications and APIs.

Short answer

AC Use a three-continent multi-region Cloud Spanner instance with nam-asia-eur1 and provision at least three Spanner nodes to meet 99.999% availability and global read latency.

Question

Your company wants to expand their users outside the United States for their popular application. The company wants to ensure 99.999% availability of the
database for their application and also wants to minimize the read latency for their users across the globe.
Which two actions should they take? (Choose two.)
  1. Create a multi-regional Cloud Spanner instance with "nam-asia-eur1" configuration. Correct Answer
  2. Create a multi-regional Cloud Spanner instance with "nam3" configuration.
  3. Create a cluster with at least 3 Spanner nodes. Correct Answer
  4. Create a cluster with at least 1 Spanner node.
  5. Create a minimum of two Cloud Spanner instances in separate regions with at least one node.
  6. Create a Cloud Dataflow pipeline to replicate data across different databases.

Key Takeaways

The question tests that Spanner's 99.999% availability comes from a multi-region topology, while read latency depends on replica placement and sufficient node capacity; the trap is choosing a North America-only multi-region configuration.

Cloud Spanner multi-region configuration is the key to 99.999% database availability and low global read latency. This page confirms the two required actions are A and C.

Common mistake: The most tempting wrong choice is B: nam3 is multi-region, but it is a North America configuration and leaves European and Asian reads far from a replica, so it cannot meet the global latency requirement.

Expert Analysis

The governing rule: availability follows the instance configuration. Cloud Spanner's 99.999% SLA applies to multi-region instances whose replicas are placed in multiple regions; a regional instance has a lower SLA. Option A is the only choice that creates a multi-region instance spanning North America, Asia and Europe (nam-asia-eur1), so it both meets the availability target and puts read replicas near non-US users. Option B's nam3 configuration is a North American multi-region layout: it can improve availability inside North America, but it does not reduce read latency for European or Asian users, so it fails the global-latency constraint.

Read latency is a placement problem, not a replication-pipeline problem. Spanner serves reads from the closest replica in the instance configuration while keeping one logical database. Option E's two separate instances in separate regions are independent databases, and Option F's Dataflow pipeline is asynchronous batch replication; neither provides Spanner's global consistency or the documented multi-region SLA. They also add operational complexity and stale reads.

Node count supplies capacity. The remaining choice is between C and D. A multi-region Spanner deployment must be provisioned with at least three nodes (the minimum-sized multi-region instance) and needs enough compute to handle the global read load; a single-node cluster is under-provisioned and is not the design intended for this scale. More nodes do not move replicas closer to users, but they prevent CPU saturation from becoming the latency bottleneck once the global configuration is in place.

Thus A and C together satisfy both constraints: A provides the three-continent replica topology and the 99.999% availability posture, and C provides the minimum node capacity for that multi-region deployment.

Official Reference

These links open the official vendor documentation in a new tab.

Exam Strategy

Separate the SLA requirement from the latency requirement. First eliminate any option that is regional or limited to one continent, even if it says multi-region (nam3 is the trap). Then eliminate DIY replication such as separate instances or Dataflow, because Spanner's guarantee comes from one multi-region instance. Finally, for capacity choices, pick the option that satisfies the minimum multi-region node count rather than the smallest possible cluster.

Frequently Asked Questions

Why does nam3 not satisfy the global read-latency requirement?

nam3 is a North America multi-region configuration. It can provide multi-region availability within North America, but it does not place replicas in Europe or Asia, so users there still read across an ocean.

Does adding Spanner nodes reduce read latency for global users?

More nodes add CPU and memory capacity and help avoid load-induced latency, but geographic latency is mainly reduced by choosing a multi-region configuration with replicas near users.

Can two separate Cloud Spanner instances replace a multi-region instance?

No. Two instances are separate databases without Spanner's global consistency or multi-region 99.999% SLA; Dataflow replication would be asynchronous and not equivalent.

More PCD FAQ →

Community Votes

A
50%
C
45%
E
5%

50% of anonymous learners picked answer A. Votes are pick records left by other test-takers — they are not the verified answer.

Discussion 22 contributions

Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.

From public community threads 22

  1. saurabh1805 👍 9
    The more number of node less read latency hence i will go with option A and C
  2. emmet 👍 7
    I think the answer should be A) + something. They wants 99.999% availability - only multi-regional instance fits this. To minimize read latency nam-asia-eur1 instance works best as it has replicas in Noth America, Europe and Asia regions. As for second answer - I do not have strong opinion.. As per documentation "Adding nodes gives each replica more CPU and RAM, which increases the replica's throughput" and they recommend to choose number of nodes to "keep high priority total CPU utilization under 65%". So nodes are not about SLA and read latency. From another hand spanner "Cloud Spanner automatically replicates your data between regions with strong consistency guarantees" so no DataFlow pipeline needed to replicate data, unless the app has other DBs and ETL between Spanner and that DBs.
  3. syu31svc 👍 6
    https://cloud.google.com/spanner/docs/instances Above link supports A for "users across the globe" as per qn https://cloud.google.com/spanner/docs/latency Above link supports C for "minimize the read latency "
  4. maxdanny 👍 4
    C because the number of nodes increases the computational capabilities ( queries for seconds and so minor latency), F to cover worldwide availability; A it's wrong because the configuration "nam-asia-eur1" not exist, Google suggests "nam-eur-asia1" o "nam-eur-asia3", D it's wrong because the number of nodes is too few; E is too expensive a solution
  5. jeeet_ 👍 4
    There is no region called ""nam-asia-eur1"" A is wrong, B&C is correct
Show 17 more comments
  1. fraloca 👍 3
    i think that the answer is A & F. A becouse minimize the latency compared to B. F becouse is the best way to import data: https://cloud.google.com/spanner/docs/import-non-spanner The answers B & C is not possible because the number of node depends to amount of the data, that we don't know.
  2. wilwong 👍 3
    Agree with A and C
  3. tomato123 👍 3
    AC are correct
  4. closer89 👍 2
    99.999% availability and reduce latency Option A gives us 99.999% availability (think its typo in region name) Option C is about compute capacity, more nodes -> less latency https://cloud.google.com/spanner/docs/instances#compute-capacity B - there is no such multi-region configuration nam3 D - its better to create cluster with 3 nodes, not 1 E,F - overengineering
  5. brunoguzzo18 👍 2
    We need multi-regin db (spanner) to satisfy 99,999% SLA and have multi-node to ensure resource needed.
  6. DonWang 👍 2
    there is nam3 https://cloud.google.com/spanner/docs/instance-configurations
  7. santoshchauhan 👍 1
    To achieve 99.999% availability and minimize read latency for a globally distributed user base, the company should: - Deploy a multi-regional Cloud Spanner instance with the "nam-asia-eur1" configuration (Option A). This setup will provide the geographical distribution of data across three continents — North America, Asia, and Europe. The multi-regional nature of this option is designed to maintain high availability and ensure users across these regions experience low latency when accessing the database. - Ensure that the Cloud Spanner instance has a minimum of three nodes (Option C). This configuration will contribute to the high availability and fault tolerance of the database. Spanner's built-in replication across these nodes in different regions will further support the five nines (99.999%) availability target, while also providing scalability for read operations.
  8. theseawillclaim 👍 1
    Let's think about all possible answers: - B does not make sense, as "nam3" is not the global configuration we want; - D is not ok, 'cause 1 node is not enough - E is not enough, as we want a global configuration; - F is totally unrelated.
  9. jcataluna 👍 1
    3 regions at least, and for those that says there is no region "man-asia-eur1", take a look at the console, its multiregion nomenclature!
  10. Srini_A 👍 1
    I see A and E as option A covers all the regions (NortAmerica, Asia, Europe) option E which ensures atleast one instance of Spanner
  11. zevexWM 👍 1
    There is according to their documentation: https://cloud.google.com/spanner/docs/instance-configurations#three_continents
  12. closer89 👍 1
    A - global and provides 99.999% availability C - more nodes - less latency
  13. __rajan__ 👍 1
    AC are best suited here.
  14. ParagSanyashiv 👍 1
    This should be A and F.
  15. jdx000 👍 1
    A and C make sense
  16. Lunaisxiaoxin 👍 1
    why not C and F
  17. tuanbo91 👍 1
    it's obvious

Reader comments & corrections

No reader corrections yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

Uploading Files to Cloud Storage for Dataproc

S-grade

Alerting when a Unix process stops running on Compute Engine

A-grade

Quickly Comparing SSD and Standard Persistent Disks

A-grade

Practice All PCD Questions

Access 84 questions with complete answers and detailed explanations.

View Full PCD Practice Test →

← Back to PCD Study Guide