Database Reference
In-Depth Information
cloud is the extraordinarily fast network. It also is clear that the network
is much less sensitive to physical locality than you might expect. These
super-fast internal connections enable BigQuery to execute multiterabyte
queries in seconds.
The connections between datacenters are as impressive as the connections
within the datacenter. Google bought up a lot of unused fiber-optic cable
around the world and uses it to move data between datacenters. For users
of BigQuery, this allows your data to be replicated to different geographical
regions extremely quickly so that your data will be available even in the
event of a natural disaster or large-scale power outage.
Query Computation
We mentioned that BigQuery's query computation engine does not use
MapReduce. And it clearly isn't a relational database…. So how does it work?
The computation engine is Dremel, a parallel, distributed, column-oriented
query engine that Google has used internally since 2006. The principal
innovation is the tree-structured architecture that allows queries to be
processed in parallel in the leaves but aggregates results at the higher levels
of the tree. The tree architecture also enables multiple queries to run at once
within the tree, which lets different users share the same hardware. Chapter
9 discusses Dremel and how it works in considerable detail.
Google Cloud Platform
BigQuery is one component of Google's Cloud Platform. The other principal
products are Compute Engine, AppEngine, Cloud Storage, Cloud SQL, and
Cloud Datastore. It can be confusing, at first, to figure out what all of these
products are, how they fit together, and when to use one versus another.
This section walks through the Google Cloud Platform offerings and
describes how they work with BigQuery or when you might want to use them
instead of BigQuery. Figure 2.3 shows the relationship between BigQuery
and various Google Cloud Platform offerings. The ones that are not listed are
services that do not have any direct connection to BigQuery (such as Google
Prediction API).
 
Search WWH ::




Custom Search