Using metric math with Amazon EFS
Using metric math, you can query multiple CloudWatch metrics and use math expressions
to
create new time series based on these metrics. You can visualize the resulting time
series
in the CloudWatch console and add them to dashboards. For example, you can use Amazon
EFS metrics to
take the sample count of DataRead
operations divided by 60. The result is the
average number of reads per second on your file system for a given 1minute period.
For more
information on metric math, see Use Metric
Math in the Amazon CloudWatch User Guide.
Following, find some useful metric math expressions for Amazon EFS.
Topics
Metric math: Throughput in MiB/second
To calculate the average throughput (in MiB/second) for a time period, first choose
a
sum statistic (DataReadIOBytes
, DataWriteIOBytes
,
MetadataIOBytes
, or TotalIOBytes
). Then convert the value to
MiB, and divide that by the number of seconds in the period.
Suppose that your example logic is this: (sum of TotalIOBytes
÷ 1,048,576
(to convert to MiB)) ÷ seconds in the period
Then your CloudWatch metric information is the following.
ID  Usable metrics  Statistic  Period 

m1 

sum  1 minute 
Your metric math ID and expression are the following.
ID  Expression 

e1  (m1/1048576)/PERIOD(m1) 
Metric math: Percent throughput
To calculate the percent throughput of the different I/O types
(DataReadIOBytes
, DataWriteIOBytes
,
or MetadataIOBytes
) for a time period, first multiply the respective
sum statistic by 100. Then divide the result by the sum statistic of
TotalIOBytes
for the same period.
Suppose that your example logic is this: (sum of DataReadIOBytes
x 100
(to convert to percentage)) ÷ sum of TotalIOBytes
Then your CloudWatch metric information is the following.
ID  Usable metric or metrics  Statistic  Period 

m1 

sum  1 minute 
m2 

sum  1 minute 
Your metric math ID and expression are the following.
ID  Expression 

e1  (m2*100)/m1 
Metric Math: Percentage of permitted throughput utilization
To calculate the percentage of permitted throughput utilization
(TotalIOBytes
) for a time period, first multiply the throughput in
MiB/second by 100. Then divide the result by the sum statistic of
PermittedThroughput
converted to MiB for the same period.
Suppose that your example logic is this: (metric math expression for throughput in
MiB/second x 100 (to convert to percentage)) ÷ (sum of PermittedThroughput
÷
1,048,576 (to convert bytes to MiB))
Then your CloudWatch metric information is the following.
ID  Usable metric or metrics  Statistic  Period 

e1 


m1 

sum  1 minute 
Your metric math ID and expression are the following.
ID  Expression 

e2  (e1*100)/(m1/1048576) 
Metric math: Throughput IOPS
To calculate the average operations per second (IOPS) for a time period, divide the
sample count statistic (DataReadIOBytes
, DataWriteIOBytes
,
MetadataIOBytes
, or TotalIOBytes
) by the number of seconds in
the period.
Suppose that your example logic is this: sample count of DataWriteIOBytes
÷ seconds in the period
Then your CloudWatch metric information is the following.
ID  Usable metrics  Statistic  Period 

m1 

sample count  1 minute 
Your metric math ID and expression are the following.
ID  Expression 

e1  m1/PERIOD(m1) 
Metric math: Percentage of IOPS
To calculate the percentage of IOPS per second of the different I/O types
(DataReadIOBytes
, DataWriteIOBytes
, or
MetadataIOBytes
) for a time period, first multiply the respective sample
count statistic by 100. Then divide that value by the sample count statistic of
TotalIOBytes
for the same period.
Suppose that your example logic is this: (sample count of MetadataIOBytes
x 100 (to convert to percentage)) ÷ sample count of TotalIOBytes
Then your CloudWatch metric information is the following.
ID  Usable metrics  Statistic  Period 

m1 

sample count  1 minute 
m2 

sample count  1 minute 
Your metric math ID and expression are the following.
ID  Expression 

e1  (m2*100)/m1 
Metric math: Average I/O size in KiB
To calculate the average I/O size (in KiB) for a period, divide the respective sum
statistic for the DataReadIOBytes
, DataWriteIOBytes
, or
MetadataIOBytes
metric by the same sample count statistic of that
metric.
Suppose that your example logic is this: (sum of DataReadIOBytes
÷ 1,024
(to convert to KiB)) ÷ sample count of DataReadIOBytes
Then your CloudWatch metric information is the following.
ID  Usable metrics  Statistic  Period 

m1 

sum  1 minute 
m2 

sample count  1 minute 
Your metric math ID and expression are the following.
ID  Expression 

e1  (m1/1024)/m2 
Using metric math through an AWS CloudFormation template for Amazon EFS
You can also create metric math expressions through AWS CloudFormation templates.
One such template
is available for you to download and customize for use from the Amazon EFS tutorials