Performance
Improve the performance of your service by checking your service quotas (formerly referred to as limits), so that you can take advantage of provisioned throughput, monitor for overutilized instances, and detect any unused resources.
You can use the following checks for the performance category.
Check names
- Amazon Aurora DB cluster under-provisioned for read workload
- Amazon EBS Provisioned IOPS (SSD) Volume Attachment Configuration
- Amazon RDS instance under-provisioned for system capacity
- High Utilization Amazon EC2 Instances
- Large Number of EC2 Security Group Rules Applied to an Instance
- Large Number of Rules in an EC2 Security Group
- Overutilized Amazon EBS Magnetic Volumes
Amazon Aurora DB cluster under-provisioned for read workload
- Description
-
Checks whether Amazon Aurora DB cluster has the resources to support a read workload.
- Check ID
-
c1qf5bt038
- Alert Criteria
-
Yellow:
Increased database reads: The database load was high and the database was reading more rows than writing or updating the rows.
- Recommended Action
-
We recommend that you tune your queries to decrease the database load or add a reader DB instance to your DB cluster with the same instance class and size as the writer DB instance in the cluster. The current configuration has at least one DB instance with a continuously high database load caused mostly by read operations. Distribute these operations by adding another DB instance to the cluster and directing the read workload to the DB cluster read-only endpoint.
- Additional Resources
-
An Aurora DB cluster has one reader endpoint for read-only connections. This endpoint uses load balancing to manage the queries contributing the most to database load in your DB cluster. The reader endpoint directs these statements to the Aurora Read Replicas and reduces the load on the primary instance. The reader endpoint also scales the capacity to handle concurrent SELECT queries with the number of Aurora Read Replicas in the cluster.
For more information, see Adding Aurora Replicas to a DB Cluster and Managing performance and scaling for Aurora DB clusters.
- Report columns
-
-
Status
-
Region
-
Resource
-
Increased database read (count)
-
Last detection period
-
Last Updated Time
-
Amazon EBS Provisioned IOPS (SSD) Volume Attachment Configuration
- Description
-
Checks for Provisioned IOPS (SSD) volumes that are attached to an Amazon EBS optimizable Amazon Elastic Compute Cloud (Amazon EC2) instance that is not EBS-optimized.
Provisioned IOPS (SSD) volumes in the Amazon Elastic Block Store (Amazon EBS) are designed to deliver the expected performance only when they are attached to an EBS-optimized instance.
- Check ID
-
PPkZrjsH2q
- Alert Criteria
-
Yellow: An Amazon EC2 instance that can be EBS-optimized has an attached Provisioned IOPS (SSD) volume but the instance is not EBS-optimized.
- Recommended Action
-
Create a new instance that is EBS-optimized, detach the volume, and reattach the volume to your new instance. For more information, see Amazon EBS-Optimized Instances and Attaching an Amazon EBS Volume to an Instance.
- Additional Resources
- Report columns
-
-
Status
-
Region/AZ
-
Volume ID
-
Volume Name
-
Volume Attachment
-
Instance ID
-
Instance Type
-
EBS Optimized
-
Amazon RDS instance under-provisioned for system capacity
- Description
-
Checks whether Amazon RDS instance or Amazon Aurora DB instance has the required system capacity to operate.
- Check ID
-
c1qf5bt039
- Alert Criteria
-
Yellow:
Out-of-memory kills: When a process on the database host is stopped because of memory reduction at the OS level, the Out Of Memory (OOM) kills counter increases.
Excessive swapping: os.memory.swap.in and os.memory.swap.out metric values were high.
- Recommended Action
-
We recommend that you tune your queries to use less memory or use a DB instance type with higher allocated memory. When the instance is running low on memory, this impacts the database performance.
- Additional Resources
-
Out-of-memory kills were detected: Linux kernel invokes the Out of Memory (OOM) Killer when the processes running on the host require more than the memory physically available from the operating system. In this case, the OOM Killer reviews all the running processes, and stops one or more processes, in order to free up system memory and keep the system running.
Swapping is detected: When the memory isn't sufficient on the database host, the operating system sends a few minimum used pages to the disk in the swap space. This offloading process impacts the database performance.
For more information, see Amazon RDS Instance Types
and Scaling yourAmazon RDS instance . - Report columns
-
-
Status
-
Region
-
Resource
-
Out-of-memory kills (count)
-
Excessive swapping (count)
-
Last detection period
-
Last Updated Time
-
High Utilization Amazon EC2 Instances
- Description
-
Checks the Amazon Elastic Compute Cloud (Amazon EC2) instances that were running at any time during the last 14 days. An alert is sent if daily CPU utilization was greater than 90% on four or more days.
Consistent high utilization can indicate optimized, steady performance. However, it can also indicate that an application does not have enough resources. To get daily CPU utilization data, download the report for this check.
- Check ID
-
ZRxQlPsb6c
- Alert Criteria
-
Yellow: An instance had more than 90% daily average CPU utilization on at least 4 of the previous 14 days.
- Recommended Action
-
Consider adding more instances. For information about scaling the number of instances based on demand, see What is Auto Scaling?
- Additional Resources
- Report columns
-
-
Region/AZ
-
Instance ID
-
Instance Type
-
Instance Name
-
14-Day Average CPU Utilization
-
Number of Days over 90% CPU Utilization
-
Large Number of EC2 Security Group Rules Applied to an Instance
- Description
-
Checks for Amazon Elastic Compute Cloud (Amazon EC2) instances that have a large number of security group rules. Performance can be degraded if an instance has a large number of rules.
- Check ID
-
j3DFqYTe29
- Alert Criteria
-
-
Yellow: An Amazon EC2-VPC instance has more than 50 security group rules.
-
Yellow: An Amazon EC2-Classic instance has more than 100 security group rules.
-
- Recommended Action
-
Reduce the number of rules associated with an instance by deleting unnecessary or overlapping rules. For more information, see Deleting Rules from a Security Group.
- Additional Resources
- Report columns
-
-
Region
-
Instance ID
-
Instance Name
-
VPC ID
-
Total Inbound Rules
-
Total Outbound Rules
-
Large Number of Rules in an EC2 Security Group
- Description
-
Checks each Amazon Elastic Compute Cloud (Amazon EC2) security group for an excessive number of rules.
If a security group has a large number of rules, performance can be degraded.
- Check ID
-
tfg86AVHAZ
- Alert Criteria
-
-
Yellow: An Amazon EC2-VPC security group has more than 50 rules.
-
Yellow: An Amazon EC2-Classic security group has more than 100 rules.
-
- Recommended Action
-
Reduce the number of rules in a security group by deleting unnecessary or overlapping rules. For more information, see Deleting Rules from a Security Group.
- Additional Resources
- Report columns
-
-
Region
-
Security Group Name
-
Group ID
-
Description
-
Instance Count
-
VPC ID
-
Total Inbound Rules
-
Total Outbound Rules
-
Overutilized Amazon EBS Magnetic Volumes
- Description
-
Checks for Amazon Elastic Block Store (Amazon EBS) magnetic volumes that are potentially overutilized and might benefit from a more efficient configuration.
A magnetic volume is designed for applications with moderate or bursty input/output (I/O) requirements, and the IOPS rate is not guaranteed. It delivers approximately 100 IOPS on average, with a best-effort ability to burst to hundreds of IOPS. For consistently higher IOPS, you can use a Provisioned IOPS (SSD) volume. For bursty IOPS, you can use a General Purpose (SSD) volume. For more information, see Amazon EBS Volume Types.
For a list of instance types that support EBS-optimized behavior, see Amazon EBS-Optimized Instances.
To get daily utilization metrics, download the report for this check. The detailed report shows a column for each of the last 14 days. If there is no active EBS volume, the cell is empty. If there is insufficient data to make a reliable measurement, the cell contains
N/A
. If there is sufficient data, the cell contains the daily median and the percentage of the variance in relation to the median (for example,256 / 20%
). - Check ID
-
k3J2hns32g
- Alert Criteria
-
Yellow: An Amazon EBS Magnetic volume is attached to an instance that can be EBS-optimized or is part of a cluster compute network with a daily median of more than 95 IOPS, and varies by less than 10% of the median value for at least 7 of the past 14 days.
- Recommended Action
-
For consistently higher IOPS, you can use a Provisioned IOPS (SSD) volume. For bursty IOPS, you can use a General Purpose (SSD) volume. For more information, see Amazon EBS Volume Types.
- Additional Resources
- Report columns
-
-
Status
-
Region
-
Volume ID
-
Volume Name
-
Number of Days Over
-
Max Daily Median
-
Note
If you opted in your account for Amazon Compute Optimizer, we recommend that you use the Amazon EBS under-provisioned volumes check instead. For more information, see Opt in Amazon Compute Optimizer for Trusted Advisor checks.