CODEWITHSUNDEEP
google-cloud-bigtable
Magnus
Magnus

Reputation: 23

Availability of single-node Bigtables

I work for a company which, at this point in time, has a small IoT network. However, the network is expected to grow significantly over the coming years. In preparing for the future we want to use Bigtables to store our data, but our budget does not yet allow for the cost of running a Bigtable production instance.

In light of this my question is: Even though the documentation says there are no guarantees wrt. uptime or performance for development instances, is it at all possible to run a database on a single node until the need for a production instance arises? Is a single-node Bigtable at least relatively dependable, or is is better to set up a temporary SQL database?

Upvotes: 2

Views: 326

Answers (1)

Misha Brukman
Misha Brukman

Reputation: 13424

Disclosure: I'm a product manager for Google Cloud Bigtable.

In light of this my question is: Even though the documentation says there are no guarantees wrt. uptime or performance for development instances, is it at all possible to run a database on a single node until the need for a production instance arises? Is a single-node Bigtable at least relatively dependable, or is is better to set up a temporary SQL database?

Is it possible to run a Cloud Bigtable dev instance? Yes, and I've spoken to customers who are in the same position as you: they like the data model and the APIs, and know that they will be able to scale up when they need to, but their current situation does not necessitate the 3+ node production cluster, so they are using a dev cluster.

You may be at higher likelihood of unavailability if you do this, but it's no worse than running a single VM with another database (whether SQL or NoSQL): having a single one of anything means that the availability critically depends on that one thing being up and reachable.

In fact, the stateless nature of Cloud Bigtable actually makes this better — since the data isn't actually stored with the server node itself, but is stored on Colossus, our distributed, replicated file system, we can (and do) substitute another Bigtable node to serve your data if your node becomes unavailable, without any data loss.

That said, we are unable to provide an SLA or any guaranteed uptime or availability for a development cluster, so please keep that in mind.

Hope this helps.

Upvotes: 1

Related Questions