Many, many (all?) companies come to GCP solely to use its database products. This is intentional and obvious from the way GCP prices its network egress. Egressing directly from a GCP db costs 10x more than if you put a VM in between, which basically forces companies to shift their entire infra into GCP GCP Spanner had the monopoly on the "perfect database" - arbitrary distributed transactions, unlimited horizontal scaling, zero-touch zero-downtime zero-loss failover. Multi-region, quorum replication, relational, blah blah. Just imagine the perfect database with no tradeoffs (except cost) and Spanner was that. Nobody else had anything like Spanner... until a month ago. AWS released a new product Aurora Limitless in preview and it's basically AWS Spanner. It also uses the old AWS Aurora's compute-storage disaggregation goodness at the shard level, which makes it a combo of Spanner and AlloyDB. AWS delivered what is basically the "final" database. This is the biggest bombshell in cloud computing history. The following companies no longer have a bright future: - GCP - Oracle (too far behind now) - Cockroach Labs (clone of Spanner by ex-Googlers) - Yugabyte (clone of Spanner by some Indian dude) - PingCAP/TiDB (clone of Spanner by some Chinese dudes) - IBM (explained below) What the future holds: Like IBM Cloud, Google will give up on its public cloud ambitions and will instead pivot to selling data software that runs on other public clouds and hybrid with on-prem. Of course it will do a better job at this than IBM (decimating IBM), but this is still a very shitty place to operate because you're basically just testing product-market fit for AWS and Azure to copy, and the platform has better insights about your user than you. (Edit: title is clickbait hyperbole, obviously)
Ok
A [SERIOUS] question for people in the cloud space: how fast do you adapt to new offerings in the cloud space? For example, like OP stated, AWS releases an alternative for GCPโs cloud spanner. Do you have the power to just pivot and try new products? Do you need permission to try new offerings? If you have a sandbox environment, how fast does management approve new tools for you (if they need to)? If AWSโs product is approved, do you have to jump through hoops to officially have it supported by your company?
Takes some time for us... Some teams might be able to make a POC pretty quickly..
from my experience, clients with glaring infra issues will switch over almost immediately if the product will solve it. you would be surprised how fast some businesses can move. also in the sales / presales departments, this stuff will get picked up super quickly.
Holy moly did I just wake up in a system design interview
Haha! I actually enjoyed this post. Good amount of technical info and was engaging. I wonder which other forums are out there which has these kinds of discussions routinely?
Hacker News
Snowflake?
Snowflake isnโt OLTP, different workload
Ok
๐ฅฑ yawn (AI, AI, AI)
Rest and vest spotted lol
Competition is a good thing. Let AWS and GCP compete so we get better products.
Agreed. Spanner used to cost a minimum of $600/month. Then they started selling 0.1x instances for $60/month (probably when they found out AWS is coming for them). Now let's see.
Cool tiktok. Thanks for sharing
Hmmm....gcp has more offerings than just spanner. AI and ML offerings are great. AWS AI services are complete crap.
Yes AI/ML will be the next big draw to GCP. But Microsoft exists. And it seems to have a crystal ball. Azure on one end. AWS on other end. GCP finna boutta get spit roasted iykwim ๐
Rightly said. AWS AI ML is a joke. Look at SageMaker and the bloated mess it offers. So many components that make no sense. AWS is so behind when it comes to GenAI. GCP is eons ahead. Telling AWS is a premium cloud is a butt joke. People patronizing AWS are delusional. Do a favor for yourselves and start with a comparative analysis of the ML offerings between SageMaker and Vertex AI - you will know the truth. Vertex is an industry leader and the true innovator, while AWS is the copy cat badly implemented. I worked for AWS by the way before starting my own company
Ok