Metal

PlanetScale Metal is the fastest way to run databases in AWS or GCP. With blazing fast NVMe drives, you can unlock unlimited IOPS, ultra low latencies, and the highest throughput for your workloads.

Read about the results of PlanetScale Metal from our customers:

Getting these results for yourself is as simple as selecting a Metal node in the PlanetScale cluster configuration dropdown.

Benefits of Metal

Unlimited I/O — Metal's local NVMe drives offer higher I/O bandwidth than network-attached storage. You will run out of CPU long before you use all your I/O bandwidth.
Query latency — Local NVMe means lower latency I/O, ensuring queries execute faster. Real customer workloads saw an average of 70% decrease in p99 and p95 latency.
Consistency — This high-bandwidth, low-latency performance is also more consistent. Even under heavy load, Metal offers predictable performance without sudden stalls.
Cost — Metal almost always comes in less expensive than similar Amazon Aurora, RDS for MySQL, and Google Cloud SQL.

The following graphs show some of the latency percentile changes we saw after switching a few of our workloads from traditional network-attached storage to Metal.

Query Insights p50, p95, p99, p99.5

Customer testimonials

PlanetScale Metal has been a game changer for Intercom. All of our biggest, most critical PlanetScale databases now run on Metal. It has allowed us to significantly improve performance, reduce costs, and simplify our infrastructure—all with minimal operational or engineering overhead.
Miles McGuire, Staff Engineer @ Intercom [Source]

We are very happy with our decision to migrate to PlanetScale Metal which enabled us to achieve the rare outcome of improvements in performance, cost, and reliability – a win for our customers and our business.
Aaron Young, Engineering Manager @ Block [Source]

By switching to PlanetScale Metal, our database queries are up to 8x faster, and we’ve been able to remove data retention limits for build analytics data!
Jacob Gillespie, Co-founder and CTO @ Depot [Source]

omg
Rafer, PlanetScale engineer, after migrating PlanetScale Insights to Metal

FAQ

What makes PlanetScale Metal performance so much better?
With your storage and compute on the same server, you avoid the I/O network hops that traditional cloud databases require. We can also use better hardware: super fast NVMe drives. Pair this with explicit sharding for large databases, and you can't beat the performance.
How does PlanetScale ensure durability?
Vitess was built to run on Google's Borg, so it has always been architected to handle possible failure.
Every PlanetScale database requires at least 2 replicas in addition to the primary. Semi-synchronous replication is always enabled. This ensures every write has reached stable storage in two availability zones before it’s acknowledged to the client.
PlanetScale Metal was forged in the fires of production at Cash App, Intercom, and several other companies who have been running their core production workloads on Metal for months prior to our public launch. PlanetScale itself has been running on Metal for months.
Metal databases have already excecuted trillions of queries across petabytes of data. For more information about Metal durability, see Why Metal databases are durable.
What do you mean by unlimited IOPS?
Though every NVMe SSD technically has a maximum IOPS, it is so high for Metal instances that it is impossible to exhaust it using MySQL. You will be maxing out your CPU to 100% long before you are able to reach the IOPS limit of the underlying drives, even for IO-intensive database workloads.
What workloads work best on Metal?
The majority of workloads will get faster on Metal, but performance-sensitive workloads are an especially good fit. High-I/O databases, the most demanding and critical workloads, massive write throughput, databases where microseconds matter — Metal was made for these workloads.
One way to measure this is if you get close to the limits of AWS EBS IOPS or have a large gp3 or io2 EBS volume bill. Metal provides unlimited IOPS and will likely yield performance improvement, cost savings, or both. There is no need to pay extra for access to the I/O throughput of the local drive.
For more information, see What is Metal for.
Why isn't every other database company offering this?
Much of the industry has recently embraced the separation of storage and compute for database architecture. This is an excellent choice for OLAP workloads, but we have found it's slow for OLTP workloads. Keeping storage and compute together has enabled us to run Vitess clusters on these NVMe drives.
The durability aspect of these drives also makes this harder for companies to offer, as you must have an extremely well-designed and well-tested solution to ensure it's safe.
Vitess was already well-equipped to solve the durability problem. With automatic built-in failovers and the requirement to have at least 2 replicas for every Vitess cluster, we can confidently say we have solved the durability issue. In addition, Vitess was originally built to run on Borg at YouTube, which had similar durability concerns. Vitess was architected to solve this problem from the beginning and has been extensively tested in production for over a decade.
Since storage doesn't autoscale, what happens if I max out my storage?
With traditional network-attached drives, you can continue adding as much data as needed and we will autoscale your storage. This isn't true with Metal - you have to select a set amount of storage upfront and must choose a larger instance if more space is needed.
If you run out of storage or get very close to running out, we will switch your database into read-only mode to prevent a full outage. Your database will no longer accept writes. This is an emergency situation that we want to make sure you avoid.
We will send automatic emails to your organization admins when your keyspaces hit certain thresholds of storage. We will also inform you about the predicted date that you'll run out of storage based on your current rate of growth. To add more storage, you simply go to the cluster configuration page in the dashboard, and select a larger drive.
Because we have to migrate your data to the new drive, so this is not an immediate operation. Fortunately, there is no downtime or impact to service while your data is migrated. Because it takes some time to migrate, we recommend upgrading storage size far before running out of storage.
I'm interested in Metal. What are the next steps?
You can sign up for a PlanetScale account today to import an existing database into a Metal instance or start from scratch.
If you would like an evaluation and hands-on migration services, we can help. Please reach out and we will get back to you with options.
Are there limitations to regions and cloud providers I can use with Metal?
You can run Metal instances on both AWS and GCP. For the full list of supported regions, see the pricing calculator.