# ScyllaDB maximum node size, RAM:Disk ratio, CPUs

**URL:** <https://forum.scylladb.com/t/scylladb-maximum-node-size-ram-disk-ratio-cpus/1432>\
**Category:** ScyllaDB\
**Tags:** sizing\
**Created:** [March 28, 2024, 5:48am UTC](https://forum.scylladb.com/t/scylladb-maximum-node-size-ram-disk-ratio-cpus/1432 "2024-03-28T05:48:27Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Guy](https://sea2.discourse-cdn.com/flex016/user_avatar/forum.scylladb.com/guy/32/9_2.png) [@Guy](https://forum.scylladb.com/u/Guy)\
**Post date:** [March 28, 2024, 5:48am UTC](https://forum.scylladb.com/t/scylladb-maximum-node-size-ram-disk-ratio-cpus/1432/1 "2024-03-28T05:48:27Z")

</div>

> [@](#):
>
> [**Originally from the User Slack**](https://scylladb-users.slack.com/)
> 
> **@Nguyen\_Huu\_Trung\_Ha:** Dear experts. How big of data a ScyllaDB node can handle?
> 
> ![dor](https://us1.discourse-cdn.com/flex016/uploads/scylladb/original/1X/27b698a3e588f84cb82a722e458d31c997a48bed.jpeg) **@dor:** There is no limit, it depends on the rest of the node resources. In a machine like i3en.metal, we can utilize all of the 60TB  
> It’s good not to exceed a ratio of 1;100 ram:disk
> 
> **@Nguyen\_Huu\_Trung\_Ha:** Thank you for your reply @dor.  
> I have another question: what is the recommended data size a node can handle well, including: decommissioning, adding nodes, compaction, cleanup, etc in acceptable time.
> 
> ![dor](https://us1.discourse-cdn.com/flex016/uploads/scylladb/original/1X/27b698a3e588f84cb82a722e458d31c997a48bed.jpeg) **@dor:** It’s primarily a function of your compaction strategy. We recommend having about 30% free space for all of those
> 
> **@Nguyen\_Huu\_Trung\_Ha:** For archiving purposes with seldom/rarely access, can we use machines with less CPU, RAM and big storage, for example: 12 CPU cores, 128GB RAM and 20-60TB storage?
> 
> ![dor](https://us1.discourse-cdn.com/flex016/uploads/scylladb/original/1X/27b698a3e588f84cb82a722e458d31c997a48bed.jpeg) **@dor:** The CPUs are less of a problem, it’s the ram:disk ratio, at the end it depends on the access pattern, right now the recommendation is 1:100 and we will work to enlarge it this year to even more than you need