Pre-requisite: EC2
Placement groups are a mechanism used in Amazon Elastic Compute Cloud (EC2) for controlling the placement of instances in a cluster. They have some rules and limitations that you need to be aware of when using them.
Here are some of the rules and limitations of Placement Groups:
- General Rules and Limitations
- Cluster Placement Group Rules and Limitations
- Partition Placement Group Rules and Limitations
- Spread Placement Group Rules and Limitations
General Rules and Limitations
The following guidelines should be considered when using Placement Groups
Each AWS service has its own set of limits, such as the number of Amazon EC2 instances that can be launched or the number of Amazon S3 buckets that can be created. These limits are designed to ensure that AWS resources are used in a manner that is fair to all customers and that the overall performance and stability of the AWS infrastructure are not impacted.
- One Placement Group at a Time: An instance cannot be launched in more than one placement group.
- AWS is designed to help customers meet their compliance requirements, but it is the customer’s responsibility to ensure that they are in compliance with all applicable laws and regulations.
- For EC2 instances in a particular Availability Zone, capacity reservations are made via on-demand capacity reservations and zonal reserved instances.
- The Instances in a Placement Group are permitted to use the capacity reservation. It is advised that you reserve space within a cluster placement group when utilizing one that allows for capacity reservations. See Capacity Reservations in Cluster Placement Groups for further details.
- For instances located in a certain Availability Zone, Zonal Reserved Instances offer a capacity reserve. The instances in a placement group are permitted to use the capacity reservation.
- AWS is committed to the Security of Customer Data and provides a number of security features to help customers secure their resources. However, customers are ultimately responsible for securing their own data and must follow best practices to ensure the security of their resources.
Cluster Placement Group Rules and Limitations
The following rules apply to Cluster Placement Groups
- Multiple Availability Zones cannot be covered by a cluster placement group. Cluster Placement Groups are limited to a single Availability Zone within a Region, so all instances within a Cluster Placement Group must reside in the same Availability Zone.
- The slower of the two instances in a cluster placement group has a limit on the maximum network throughput speed of traffic between them. Select an instance type with network connectivity that satisfies your needs if your applications have high throughput requirements.
- A cluster placement group can launch different instance kinds. This lessens the possibility that the necessary capacity will be accessible for your launch to be successful, though. All instances in a cluster placement group should be of the same instance type, per our recommendation.
- Not all EC2 instance types are supported in Cluster Placement Groups. For example, T2 and T3 instances are not supported.
Guidelines when improved Networking is enabled
- A Cluster Placement Group’s instances can use up to 10 Gbps of single-flow bandwidth. For single-flow traffic, instances outside of a cluster placement group can use up to 5 Gbps.
- All available instance aggregate bandwidth can be used for traffic to and from Amazon S3 buckets located in the same Region when it is routed across the public IP address space or through a VPC endpoint. Instances within a Cluster Placement Group are connected to multiple high-bandwidth, low-latency networks, which provides improved network performance compared to other placement strategies.
- The Maximum Amount of network traffic that can be sent to the internet or over an AWS Direct Connect connection to on-premises resources is 5 Gbps.
- Cluster Placement Group names must be unique within a region and cannot be changed once created.
Partition Placement Group Rules and Limitations
The following rules (guidelines) apply to Partition placement groups:
- The most partitions that can be supported by a partition placement group per Availability Zone is seven. Only your account limits can determine how many instances you can run in a partition placement group, Partition placement groups can only be created within a single availability zone.
- Amazon EC2 makes an effort to evenly divide the instances among all partitions when they are deployed into a group for partition placement. An equal distribution of instances among all partitions is not a promise provided by Amazon EC2. Some instance types are not supported, and others are supported but with limitations. For a list of supported instance types, see the Amazon EC2 documentation.
- The maximum number of partitions in a partition placement group with Dedicated Instances is two.
- A partition placement group’s capacity cannot be reserved using capacity reservations. A partition placement group’s capacity is not reserved by capacity reservations.
- Partition placement groups do not support termination protection, so be careful when terminating instances in a placement group.
- Partition placement group names must be unique within a region and must follow a specific naming convention. The name must be a combination of upper and lower case letters, numbers, and hyphens, and must start with a letter and be no longer than 255 characters.
Spread Placement Group Rules and Limitations
The following rules (guidelines) apply to the Spread Placement Group
- The most running instances that can be supported by a rack spread placement group are seven. A Region with three Availability Zones, for instance, allows you to run a total of 21 instances in the group, with seven instances in each Availability Zone. Eighth instances cannot launch if they are started in the same Availability Zone and spread placement group. We suggest using multiple spread deployment groups if you want more than seven instances in a given Availability Zone. Using several spread placement groups helps ensure the spread for each group but not guaranteeing the spread of instances between groups, hence reducing the impact of specific failure classes.
- Dedicated Instances do not allow spread placement groups. The Spread placement strategy is only available for EC2 instances that are homogeneous in terms of their instance type and network performance.
- Only AWS Outposts allow host-level spread placement groups for placement groups. With host-level spread placement groups, there are no limitations on the number of running instances.
- Capacity reservations cannot be used to reserve space in a spread placement group.
- Space in a spread placement group is not reserved by reservations for capacity.
- The Spread placement strategy does not support integration with Elastic Load Balancers (ELBs).
- The Spread placement strategy assumes that all instances in the Auto Scaling group have the same network performance characteristics, so it may not be suitable for instances with varying network performance needs.