Cloud Block Store Support Matrix
Cloud Block Store Models and Capacity Upgrades
| Model | Usable | Effective (4:1) | Controller Instances | VD instances |
EBS Volumes for NVRAM |
Provisioned IOPS for io1 (NVRAM) |
|
CBS //V10A-R1 |
6.9 TiB (7.6 TB) |
27.6 TiB (30.3 TB) |
2 x c5n.9xlarge **
|
7 x i3.2xlarge **
|
7 x 60 GB io1 volumes |
3000 per io1 volume Total: 21,000* |
|
CBS //V10A-R1 + First shelf upgrade |
13.8 TB (15.2 TB) |
55.2 TiB (60.7 TB) |
2 x c5n.9xlarge ** |
14 x i3.2xlarge ** | 7 x 60 GB io1 volumes |
3000 per io1 volume Total: 21,000* |
| Model | Usable | Effective (4:1) | Controller Instances | VD instances |
EBS Volumes for NVRAM |
Provisioned IOPS for io1 (NVRAM) |
|
CBS //V20A-R1 |
13.8 TiB (15.2 TB) |
55.2 TiB (60 TB) |
2 x c5n.18xlarge ** |
7 x i3.4xlarge **
|
7 x 120 GB io1 volumes |
6000 per io1 volume Total: 42,000* |
|
CBS //V20A-R1 + First shelf upgrade |
27.6 TiB (30.4 TB) |
110.4 TiB (121.4 TB) |
2 x c5n.18xlarge ** |
14 x i3.4xlarge ** |
7 x 120 GB io1 volumes |
6000 per io1 volume Total: 42,000* |
|
CBS //V20A-R1 + Second shelf upgrade |
55.2 (TiB) 60.8 (TB) |
220.8 TiB (242.8 TB) |
2 x c5n.18xlarge ** |
14 x i3.4xlarge ** + 7 x i3.8xlarge ** |
7 x 120 GB io1 volumes |
6000 per io1 volume Total: 42,000* |
* The provisioned IOPS refer to the number of provisioned IOPS for the io1 volumes which are used for Cloud Block Store NVRAM. This provisioned IOPS value does not reflect the total effective IOPS available from Cloud Block Store.
** Pure strongly recommends using Convertible Reserve Instances (rather than On-demand or Standard Reserve Instances) for the underlying controller and VD EC2 instances. This allows customers the flexibility to NDU to newer, cheaper, and more powerful EC2 types when made available by AWS.
Supported Regions
- us-east-1 (N. Virginia) *
- us-east-2 (Ohio)
- us-west-2 (Oregon) * **
- us-gov-west-1 (AWS GovCloud US)
- us-gov-east-1 (AWS GovCloud US-East)
- eu-central-1 (Frankfurt) *
- eu-west-1 (Ireland)
- ap-southeast-1 (Singapore)*
- ap-southeast-2 (Sydney)
- ap-northest-1 (Tokyo)
- ap-northeast-2 (Seoul)
- ca-central-1 (Canada Central)
* These regions are generally supported. However there are some Availability Zones within these regions that do not have the required c5n.9xlarge and c5n.18xlarge instances for Cloud Block Store. These Availability Zones are different for every customer. Customers can contact AWS Support to find out which Availability Zone does not include support for c5n.9xlarge and c5n.18xlarge instances, and avoid deploying Cloud Block Store in subnets tied to these Availability Zones.
**Cloud Block Store in an ActiveCluster configuration is not supported in Oregon if using with Pure1 Mediator. Customers who want to deploy Cloud Block Store with ActiveCluster in Oregon must use the On-Premises Mediator.
Note: Support for each region depends on the availability of EC2 resources for Cloud Block Store. For regions where there are low quantities c5n or i3 instances, customers have the option reserve the instances ahead of usage. See Capacity Reservations.
Supported Capabilities and Features
| Feature/Capability | Support | Notes |
| Nesting CloudFormation template | No | Cloud Block Store is deployed using CloudFormation. Pure provides customers with a CloudFormation (CF) template yaml file. Cloud Block Store must be deployed using its own standalone CloudFormation template. It is important that customers do not nest the Cloud Block Store CloudFormation template in other CloudFormation templates as this can lead to unexpected configuration issues over time. |
| Shut down or Stop | No | Stopping or Shutting down Cloud Block Store or its underlying resources are not supported |
| Cloud Block Store termination/deletion |
Support Driven: v5.3.0.aws0, 5.3.0.aws1, 5.3.0.aws2 Customer Driven: v5.3.3.aws0+ |
CBS with Purity version v5.3.0.aws0, 5.3.0.aws1, and 5.3.0.aws2 requires Pure Support in order to delete a CBS instance. CBS with Purity version 5.3.3.aws0 and higher can be deleted by customers. See CBS Deployment Guide for steps to terminate/delete a CBS instance. |
| Host ports | 2 (iSCSI) | One per controller |
| Replication ports | 2 | One per controller |
| Management ports | 2 | One per controller |
| Deduplication | Yes | |
| Compression | Yes | |
| Thin Provisioning | Yes | |
| Snapshots | Yes | |
| CloudSnap creation | Yes | |
| CloudSnap restore | Yes | |
| QoS (Fairness) | No | |
| QoS (Limits) | Yes | |
| Purity Run | No | |
| WFS | No | |
| Encryption | Yes | |
| REST APIs | Yes | |
| Controller NDU** | Yes |
Support Driven via CLI commands |
| Capacity NDU*** | Yes | Support Driven via CLI commands |
| Pure1 | Yes | |
| Pure1 Meta (Workload Planner) | No | |
| VM Analytics | N/A | No VMware Cloud support for CBS |
| VMware Cloud | No | |
| Changing Network IP or subnets | No | This is an AWS limitation |
| Replication Features | Support | Notes |
|
Async Replication
|
Yes
|
|
| Active Cluster (Synchronous Replication) |
Yes*
|
|
| Active/Active Async |
No |
* Cloud Block Store in an ActiveCluster configuration is not supported in Oregon if using with Pure1 Mediator. Customers who want to deploy Cloud Block Store with ActiveCluster in Oregon must use the On-Premises Mediator.
**Controller upgrade from //V10A-R1 to //V20A-R1 is supported but not recommended since backend performance will be limited because the virtual drives cannot be upgraded in-place from i3.2xlarge to i3.4xlarge. The i3.4xlarge virtual drives have more network bandwidth to support higher backend traffic.
*** Capacity upgrade is supported by adding additional virtual drives, seven at a time. However, capacity upgrade in-place from i3.2xlarge to i3.4xlarge is not yet supported. Support for in-place capacity upgrade will be post-GA.
//V10AR1 Limits
General Limits
|
Description |
5.3.x [AC Enabled] |
|---|---|
|
Max # of volumes |
500 |
| Max # of volume groups | 100 |
|
Max # of hosts |
200 [100]* |
|
Max # of IQNs |
200 [100]* |
| Max # of sessions | 2400 |
|
Max # of host groups |
50 [25]* |
EC2 Instance Bandwidth Limits
|
Description |
c5n.9xlarge |
|---|---|
|
Network Bandwidth (Aggregate)1 |
50 Gbps |
| Bandwidth per connection | 5 Gbps |
1 The 100 Gbps aggregate network bandwidth is shared across all virtual Elastic Network Interfaces (ENIs). These include the system, management, replication, and iSCSI interfaces. The virtual ENIs are backed by a physical Elastic Network Adapter (ENA).
Each connection or socket (source IP address:source port, target IP address:target port) is limited to 5 Gbps. If 4 iSCSI connections to the primary controller the resulting bandwidth would be 20 Gbps.
Snapshots and Asynchronous Replication Limits
|
Description |
5.3.x [AC Enabled] |
|---|---|
| Max # of volume snapshots per array | 1,000 |
| Max # of pgroups per array1 | 50 [50] |
| Max # of remote connected arrays | 1 |
| Minimum configurable replication frequency2 |
1 hour |
1 Example for pgroup and snapshot limits: If you create a pgroup containing 100 volumes and then create a single pgroup snapshot for that pgroup then the following would be counted against each maximum:
- 1 pgroup consumed from the array-wide maximum number of pgroups
- 1 pgroup snapshot consumed from the array-wide pgroup snapshot maximum
- 100 volume snapshots consumed from the array-wide volume snapshot maximum
Note: In cases where a pgroup snapshot request will result in more volume snapshots than are supported by the FlashArray, the pgroup snapshot request and/or scheduled pgroup snapshot will fail.
2 This is the minimum configurable replication frequency. Replication is a background process and priority is given to front-end workload. Maintaining the configured replication frequency depends on several factors such as the front-end workload on the array, the amount of data reduction, the amount of logical address space being replicated, and the available replication bandwidth.
Snapshot Offload (NFS/Cloud) Replication Limits
|
Description |
5.3.x |
|---|---|
|
Max # of volume snapshots on an offload target |
10,000 |
| Max # of volume snapshots that one FlashArray can offload to an offload target | 10,000 |
| Max # of offload targets configurable per FlashArray | 1 |
| Max # of FlashArrays configurable per offload target | 4 |
Host Limits
|
Description |
5.3.x |
|---|---|
|
Max # of IQNs per host |
200 |
| Max # of LUNs (volumes) per host | 500 |
| Max # of private LUNs connected per host (LUNs not connected to host groups) | 500 |
|
LUN IDs assigned for LUN connections |
|
Host Group Limits
|
Description |
5.3.x |
|---|---|
| Max # of hosts per host group | No specific limit beyond max hosts |
|
Max # of LUNs (volumes) per host group |
500 Note: Private LUN connections count against host group max LUN limit. See above for LUN ID assignment. |
Volume Limits
| Description | 5.3.x |
|---|---|
| Max volume size | 4PB |
//V20AR1 Limits
General Limits
|
Description |
5.3.x [AC Enabled] |
|---|---|
|
Max # of volumes |
1000 |
| Max # of volume groups | 200 |
|
Max # of hosts |
400 [200]* |
|
Max # of IQNs |
400 [200]* |
| Max # of sessions | 4800 |
|
Max # of host groups |
100 [50] |
EC2 Instance Bandwidth Limits
|
Description |
c5n.18xlarge |
|---|---|
|
Network Bandwidth (Aggregate)1 |
100 Gbps |
| Bandwidth per connection | 5 Gbps |
1 The 100 Gbps aggregate network bandwidth is shared across all virtual Elastic Network Interfaces (ENIs). These include the system, management, replication, and iSCSI interfaces. The virtual ENIs are backed by a physical Elastic Network Adapter (ENA).
Each connection or socket (source IP address:source port, target IP address:target port) is limited to 5 Gbps. If 4 iSCSI connections to the primary controller the resulting bandwidth would be 20 Gbps.
Snapshot and Asynchronous Replication Limits
|
Description |
5.3.x [AC Enabled] |
|---|---|
|
Max # of volume snapshots per array2 |
2000 |
| Max # of pgroups per array2 | 100 [100] |
| Max # of remote connected arrays | 4 |
| Minimum configurable replication frequency3 |
30 minutes3 |
2 Example for pgroup and snapshot limits: If you create a pgroup containing 100 volumes and then create a single pgroup snapshot for that pgroup then the following would be counted against each maximum:
- 1 pgroup consumed from the array-wide maximum number of pgroups
- 1 pgroup snapshot consumed from the array-wide pgroup snapshot maximum
- 100 volume snapshots consumed from the array-wide volume snapshot maximum
Note: In cases where a pgroup snapshot request will result in more volume snapshots than are supported by the FlashArray, the pgroup snapshot request and/or scheduled pgroup snapshot will fail.
3 This is the minimum configurable replication frequency. Replication is a background process and priority is given to front-end workload. Maintaining the configured replication frequency depends on several factors such as the front-end workload on the array, the amount of data reduction, the amount of logical address space being replicated, and the available replication bandwidth.
Snapshot Offload (NFS/Cloud) Replication Limits
|
Description |
5.3.x |
|---|---|
|
Max # of volume snapshots on an offload target |
100,000 |
| Max # of volume snapshots that one FlashArray can offload to an offload target | 100,000 |
| Max # of offload targets configurable per FlashArray | 1 |
| Max # of FlashArrays configurable per offload target | 4 |
Host Limits
|
Description |
5.3.x |
|---|---|
|
Max # of IQNs per host |
400 |
| Max # of LUNs (volumes) per host | 500 |
| Max # of private LUNs connected per host (LUNs not connected to host groups) | 500 |
|
LUN IDs assigned for LUN connections |
|
Host Group Limits
|
Description |
5.3.x |
|---|---|
|
Max # of hosts per host group |
No specific limit beyond max hosts |
|
Max # of LUNs (volumes) per host group |
500 Note: Private LUN connections count against host group max LUN limit. See above for LUN ID assignment. |
Volume Limits
| Description | 5.3.x |
|---|---|
|
Max volume size |
4PB |