Overview of cloud disk volumes
ACK Serverless clusters support Alibaba Cloud disk volumes as persistent block storage. This topic covers disk types, performance specifications, key constraints, and billing to help you choose the right disk before provisioning.
Considerations
Before mounting a disk volume, note the following constraints:
One pod per disk — disks cannot be shared. If multiple replicated pods are scheduled to different nodes, they attempt to mount the same disk simultaneously, causing mount failures.
Use a StatefulSet, not a Deployment — if you must use a Deployment, set the replica count to 1. With multiple replicas, upgrade policies can interrupt disk mounting during pod restarts.
Disk type must match the ECS instance type — verify compatibility before provisioning. See Instance families.
16 disks per node maximum — each disk can be up to 32 TiB.
Pay-as-you-go billing only — do not change the disk billing method to subscription after changing an ECS instance from pay-as-you-go to subscription. Doing so prevents the disk from being mounted.
`securityContext.fsGroup` increases mount time — when set, kubelet runs
chmodandchownafter mounting, adding latency. For more information about how to reduce the mounting time, see Why does it require a long time to mount a NAS volume? and Why does it require a long period of time to mount an OSS volume?.
Disk types
Tip: For most production workloads requiring low latency and high throughput, use ESSDs (PL1 or higher). For development and testing, Ultra disks are the most cost-effective option.
Alibaba Cloud disks are block-level storage resources backed by a distributed triplicate mechanism. If hardware errors disrupt a zone, data is automatically replicated to an unaffected disk in another zone.
Four disk families are available:
Enhanced SSD (ESSD) — built on next-generation distributed block storage architecture using 25 Gigabit Ethernet and remote direct memory access (RDMA). Delivers low-latency I/O, up to 1,000,000 random read/write IOPS. Best for online transactional processing (OLTP) databases, NoSQL databases, and ELK (Elasticsearch, Logstash, and Kibana) distributed logs.
Standard SSD — high-performance disks with consistent random IOPS and high data reliability. Best for I/O-intensive applications and small-to-medium relational and NoSQL databases.
Ultra disk — cost-effective with medium random IOPS and high data reliability. Best for development and testing as system disks.
Basic disk — previous-generation disks. No longer available for purchase.
Performance specifications
| Disk type | Capacity range (GiB) | Max IOPS | Max throughput (MB/s) | Write latency (ms) | API parameter |
|---|---|---|---|---|---|
| ESSD AutoPL | 40–65,536 | 1,000,000 | 4,096 | 0.2 | cloud_auto |
| ESSD PL3 | 1,261–65,536 | 1,000,000 | 4,000 | 0.2 | cloud_essd |
| ESSD PL2 | 461–65,536 | 100,000 | 750 | 0.2 | cloud_essd |
| ESSD PL1 | 20–65,536 | 50,000 | 350 | 0.2 | cloud_essd |
| ESSD PL0 | 20–65,536 | 10,000 | 180 | 0.3–0.5 | cloud_essd |
| Standard SSD | 20–32,768 | 25,000 | 300 | 0.5–2 | cloud_ssd |
| Ultra disk | 20–32,768 | 5,000 | 140 | 1–3 | cloud_efficiency |
| Basic disk | 5–2,000 | Several hundreds | 30–40 | 5–10 | cloud |
Latency values are one-way random write latency at 4 KB block size.
IOPS formulas:
| Disk type | Formula |
|---|---|
| ESSD AutoPL (baseline) | min{1,800 + 50 × Capacity, 50,000} |
| ESSD AutoPL (provisioned) | min{Provisioned IOPS, 50,000} |
| ESSD AutoPL (burst) | min{IOPS supported by the instance type, 1,000,000} |
| ESSD PL3 | min{1,800 + 50 × Capacity, 1,000,000} |
| ESSD PL2 | min{1,800 + 50 × Capacity, 100,000} |
| ESSD PL1 | min{1,800 + 50 × Capacity, 50,000} |
| ESSD PL0 | min{1,800 + 12 × Capacity, 10,000} |
| Standard SSD | min{1,800 + 30 × Capacity, 25,000} |
| Ultra disk | min{1,800 + 8 × Capacity, 5,000} |
| Basic disk | — |
Throughput formulas (MB/s):
| Disk type | Formula |
|---|---|
| ESSD AutoPL (baseline) | min{120 + 0.5 × Capacity, 350} |
| ESSD AutoPL (provisioned) | min{16 KB × Provisioned IOPS / 1,024, Maximum throughput per disk} |
| ESSD AutoPL (burst) | min{Throughput supported by the instance type, 4 GB/s} |
| ESSD PL3 | min{120 + 0.5 × Capacity, 4,000} |
| ESSD PL2 | min{120 + 0.5 × Capacity, 750} |
| ESSD PL1 | min{120 + 0.5 × Capacity, 350} |
| ESSD PL0 | min{100 + 0.25 × Capacity, 180} |
| Standard SSD | min{120 + 0.5 × Capacity, 300} |
| Ultra disk | min{100 + 0.15 × Capacity, 140} |
| Basic disk | — |
For full block storage performance details, see Block storage performance.
Billing
Only pay-as-you-go disks can be mounted to ACK Serverless clusters.
If the ECS instance billing method changes from pay-as-you-go to subscription, keep the disk billing method as pay-as-you-go. Changing the disk to subscription prevents it from being mounted.
To reduce costs on disks that do not support subscription billing, purchase storage capacity units (SCUs). See Introduction to SCUs.
For general billing information, see Billing.
What's next
Mount a disk volume to your ACK Serverless cluster: