This week in Google Cloud —“Cloud VPC’s seriously unique features and another decision tree”
You’ve heard it before, Google’s network is fast, high performance, and global. Check out how Cloud VPC (Virtual Private Cloud) can also be sharable, expandable and how one instance span multiple regions — “Reimagining virtual private clouds” for more details.
If you liked the Storage/DB decision tree from a few weeks ago, you’ll enjoy “Choosing the right compute option in GCP: a decision tree” just as well.
“Counting uniques faster in BigQuery with HyperLogLog++” discussed how BigQuery uses an implementation of the HyperLogLog algorithm for cardinality estimation when speed is more important than exact count.
How exactly-once processing in Cloud Dataflow is supported by sources and sinks. This is the third and final part of the “Exactly-once processing in Cloud Dataflow” series.
From the “technical coverage for recent GCP announcements” department :
- Google App Engine Gets Full Java 8 Support (programmableweb.com)
- Spanner: becoming a SQL system (blog.acolyer.org)
- Google Cloud Load Balancer Setup Tweaking and Observations (medium.com)
- Google Updates Video Learning and Vision Machine Learning APIs (programmableweb.com)
- Google Touts Value of Cloud IoT Core for Analyzing Connected Car Data (eweek.com)
Completing the list is a more generic piece on how committed and involved Google is wrt Open Source for GCP: “How Google Turned Open Source Into A Key Differentiator For Its Cloud Platform” (forbes.com)
Romin Irani was lucky to attend Google I/O this year and was granted $700 worth of GCP credit. Here’s what he suggests you should do (this also applies to the $300 free trial from cloud.google.com/free) : “You have $700 GCP Credit … Now What?” (rominirani.com)
From the “ICYMI” department :
- Add, remove, or update a payment method (also covers accepted credit cards around the world)
- Dataproc Regional endpoints
From the “Bigtable and BigQuery partners and customers say it best” department :
- You probably shouldn’t use DynamoDB (syslog.ravelin.com)
You’ll never guess what happened when Ravelin threw 100x their peak throughput at Google Cloud Bigtable… - How we moved our Historical Stats from MySQL to Bigtable with zero downtime (fastly.com)
- How to speed up your BigQuery query 31x by replacing a self-join with two UNNEST() operations (medium.com) “instead of a query that takes 380 seconds to complete in billing tier 8, my modified query runs in 12 seconds in billing tier 1”
Decision trees make for great pics of the week :
That’s it for this week!
-Alexis

