Skip to main content
Version: User Guides (BYOC)

FAQ: Resource Planning

This topic lists the possible issues that you may encounter while planning your resources on Zilliz Cloud and the corresponding solution.

Contents​

FAQs​

What is a Compute Unit (CU)?​

A compute unit (CU) is a group of hardware resources for serving your indexes and search requests. You can simply consider a CU as a fully-managed physical node for deploying search service.

For more details, see Select the Right CU.

What is a vCU? How does it get calculated?​

A vCU is a virtual compute unit used to measure the resources consumed by read operations (such as search and query) and write operations (such as insert, upsert, bulk insert, and delete). The data volume written or read will be converted from GB to vCUs.

How can I avoid expenses on unused clusters?​

We recommend suspending unused clusters to save computing costs. You can resume them later when necessary.

How many CUs do I need for a given collection?​

A Performance-optimized CU can serve 7.5 million 128-dimensional vectors or 1.5 million 768-dimensional vectors.

A Capacity-optimized CU can serve 25 million 128-dimensional vectors or 5 million 768-dimensional vectors.

Since your collection's schema may differ from the ones in the simple guide above, we highly recommend you test the actual requirements against different CU types.

Which type of CU should I pick?​

Select the Performance-optimized CU if you require high throughput and low latency for demanding use cases. Alternatively, choose the Capacity-optimized CU if your priority is to host large volumes of data with less concern for throughput and latency, as it offers a better balance of performance and cost.

For more information, please refer to Select the Right CU.

What's the difference between Performance-optimized CU and Capacity-optimized CU?​

The "Performance-optimized Compute Unit" suits low latency or high throughput similarity searches. This option works best for high-search performance scenarios.

The "Capacity-optimized Compute Unit" suits data volumes that are five times larger than the performance-optimized CU option. This option works best for increased storage capacity scenarios.

For more details, see Select the Right CU.