How to collect your workload for Savio
The collector measures how busy your Amazon RDS SQL Server instances really are (processor, memory, disk) and writes the results to files you upload to Savio. Savio uses them to show whether a smaller, cheaper setup would still fit. The collector does not read or change the data in your application databases.
How it works
The collector walks you through five steps, shown at the top of its window:
- Servers. Make the list of servers to measure. The collector can build it for you from AWS.
- Collector. Choose how Savio reads them: by signing in to SQL Server, or from AWS CloudWatch only.
- Length. Choose how long to measure (7 days is usual), and where to save the files.
- Check. The collector checks each server's sign-in, database and permissions before anything starts.
- Collect. Measuring runs on your servers. When it finishes, you export the files and upload them to Savio.
With CloudWatch, steps 3 and 4 are skipped: you pick 14 or 30 days and it collects straight away.
Which collector should I use?
Signs in to each server and reads SQL Server's own counters, including activity for each database. Gives the most detail. You need each instance's master user login and password, and a small database on each server for the collector's own tables (explained below).
Reads from AWS only, with no SQL Server login and nothing created on your servers. Use it when you cannot get a SQL Server login. You need AWS CLI v2 and a read-only AWS identity.
| SQL Server collector | CloudWatch collector | |
|---|---|---|
| Reads | SQL Server directly | AWS only: RDS settings, CloudWatch, Enhanced Monitoring, Database Insights |
| You need | Each server's master user login and password, and a small collection database on each server | AWS CLI v2 with a read-only AWS identity (the policy is in SavioCW_README.txt) |
| Creates on your server | One SQL Agent job and its own tables in the collection database | Nothing |
| Days covered | The days ahead. It collects for the length you choose: 7 days; for many databases or longer periods, use CloudWatch | The days already past: 14 by default, 30 at most. You can upload the same day |
| CPU and disk activity | Yes | Yes |
| Memory | SQL Server's own memory readings and page life expectancy | Page life expectancy from Database Insights, only where it is turned on and only for the days it keeps (7 by default). Without it, Savio does not lower memory |
| tempdb | Checked against the smaller size | Shown for information only. CloudWatch does not report tempdb's size |
| Busiest databases | Yes | No |
| Enterprise-to-Standard check | Yes | No |
| RDS size and storage settings | From your server list (SavioServers.csv) | Read from AWS |
| Your sign-in details | Never exported | AWS credentials stay on your computer |
Before you start
- Unzip the download into one folder and keep every file together, including
SavioPackageToken.json. The token ties your results to your Savio account; without it Savio refuses them. If it is missing, download the collector again from app.usesavio.com/cost/assessment. - Start the collector by double-clicking
SavioCollector.bat. - Run it on a Windows computer that can reach your RDS servers over the network, the same way SQL Server Management Studio (SSMS) would.
- The SQL Server collector uses Microsoft's SqlServer PowerShell module. If the collector says it is not ready, open PowerShell and run
Install-Module SqlServer -Scope CurrentUser(orImport-Module SqlServerif it is already installed), then open the collector again.
Step 1: Your server list
Both collectors use the same list, SavioServers.csv, with one row per server. A CSV file is a simple table that Excel opens and saves. The first screen shows the list and what each server still needs, and reads the file again every time you save it.
- Build from AWS fills the list with every RDS SQL Server instance your AWS account can see, in every Region, with their size and storage settings. It only reads from AWS, asks before it starts, and needs the AWS CLI set up (see CloudWatch collector).
- Or Open a list to use a list you already have (a CSV, or a text file with one server endpoint per line).
- Edit in Excel opens the list. Delete the rows of servers you do not want measured, fill in Auth and Database (and Login and Password if you like), then save it, keeping the CSV format.
| Column | What to put in it | |
|---|---|---|
ServerName | Required | The server's endpoint: its address, for example mydb.abc123.us-east-1.rds.amazonaws.com. Where to find it. |
Auth | SQL Server | s to sign in with a SQL Server login and password (use this with the master user), or w for Windows sign-in (only in some setups) |
Database | SQL Server | The name of the collection database on that server, for example SavioCollect. Letters, numbers and underscores only. What it is and how to create it. |
Login, Password | SQL Server | The master user's name and password, when Auth is s. You can leave them empty and type one login for every server in the collector instead. |
RDSSize | Optional | The instance class, for example db.r5.4xlarge. Without it, Savio uses the processor count the collector measures. |
StorageType, AllocatedStorageGb, ProvisionedIops, ProvisionedThroughputMbps | Optional | From the RDS console (Configuration, Storage). Without them, Savio cannot check whether your storage could be lowered. Build from AWS fills them in. |
Where to find a server's endpoint
- Sign in to the AWS console and open RDS. Choose the Region your database is in.
- Choose Databases, then the instance name.
- On the Connectivity & security tab, copy the Endpoint.
Passwords stay on your computer. See Your passwords below.
Your passwords stay with you
Savio does not collect, save or use your database passwords.
- The collector uses them only on this computer, to connect to your servers.
- They never go into the files you upload. Before the collector makes the ZIP file, it checks it and stops if a login or password is inside.
- The support log hides them: they show as
[REDACTED]. - Savio never asks you for a database password. You sign in to Savio with a link sent to your email.
When you have finished, delete the passwords from your server list.
The collection database: what it is and why it is needed
The SQL Server collector does not stay connected for the whole week. Instead it sets up a small scheduled task on each server, a SQL Server Agent job named SQL_IOCollection, that wakes up every minute and writes a reading (processor, memory, disk activity per database) into a few tables. Those tables need a home: that is the collection database.
- Why a separate database? So nothing is ever added to your application databases. Everything the collector creates lives in this one database, and is easy to see and remove.
- What goes in it: the tables
SQL_CollectionStatus,SQL_CPUCollection,SQL_MemCollection,SQL_DBIOandSQL_DBIOTotal, plusCO_WorkloadSamplesandCO_WaitStatsSampleswhen full workload evidence is on. They hold only measurements, never your data. - At the end: Export reads these tables into your files. Clean up then deletes the tables and the job. You can delete the database yourself afterwards.
- One per server: each server in your list needs its own collection database. It can use the same name on every server.
How to create it
The collector does not create the database for you. Ask your database administrator, or do it yourself in SSMS:
- Connect to the server in SSMS with the master user (see step 2).
- Select New Query and run the line below. Repeat on each server.
CREATE DATABASE SavioCollect;
Then put SavioCollect in the Database column of your list. Whoever creates a database becomes its owner, so the master user that created it already has every permission the collector needs there.
What the SQL Server collector does on your server
What it creates
- One SQL Server Agent job named
SQL_IOCollection, which takes a reading every minute. - Its own tables, only in the collection database (listed above). Nothing is added to your application databases.
What it reads
Only SQL Server's own system views, the same ones monitoring tools use:
- Activity and memory counters, including page life expectancy (
sys.dm_os_performance_counters). - Processor use over time, and memory use (
sys.dm_os_ring_buffers,sys.dm_os_sys_memory). - Disk activity for each database file (
sys.dm_io_virtual_file_stats). - What SQL Server waits on (
sys.dm_os_wait_stats). - Database names, file sizes and server settings (
sys.databases,sys.master_files,sys.configurations). - Enterprise features in use, for the Enterprise-to-Standard check (
sys.dm_db_persisted_sku_features).
What it never reads
Your application tables or rows, SQL text, query plans, Query Store or traces.
At the end
Export copies the readings into your files. Clean up deletes the collector's tables and the job.
CloudWatch collector: creates nothing on your servers and reads only from AWS.
Effect on your server
In our September 2026 test on Amazon RDS for SQL Server, the collector caused no measurable CPU or memory increase and only a small, expected increase in disk reads. The test application completed about 6.6% fewer transactions in most runs while the collector ran; we are still investigating why.
The full results, and how to run a careful first collection, are on the Security and data page.
Step 2: Sign-in and permissions
Sign in with the master user
Use each instance's master user login and password. When your RDS instance was created, AWS asked for a master user name and password: the instance's own administrator login. You can see the name in the RDS console under the instance's Configuration tab. On RDS this login can create databases, and it already belongs to the SQL Agent role the collector needs, so it passes every check. Nobody has the classic sa or sysadmin login on RDS.
In your list, put s in Auth and the master user's name and password in Login and Password. If every instance has the same master user, you can type it once instead (see One login for every server).
Before anything starts, the collector checks each server and says which check failed.
One login for every server
Each server uses the Auth, Login and Password in your list. To use the same login for every server instead, tick Use one login for all servers instead on the sign-in screen and type it there. It is used only on this computer and never saved in the files.
Windows sign-in (Auth w)
Works only if your RDS instance is joined to an AWS Managed Microsoft AD domain and a SQL Server login was created for your Windows account. The collector then signs in as the Windows user running it. If you are unsure, use s with the master user.
Step 3: How long, and where to save
| Length | What Savio can do with it |
|---|---|
| 7 days (10,080 minutes) or more | A size you can rely on, because it covers a full week of your busy and quiet days. For a server with more than about 40 databases, or to cover more than 7 days, use the CloudWatch collector. |
| 48 hours to under 7 days | A size, shown only with a warning. |
| Under 48 hours | No size. |
- Customer name: your company or project name. The collector makes a folder with this name inside the folder you choose, so different collections do not mix.
- Save the files to: the folder on this computer where the result files go.
- Full workload evidence: also records the workload and wait readings Savio needs for the full assessment. Leave it ticked unless Savio has told you otherwise.
Steps 4 and 5: Check, then collect
- Check servers and start checks every server first. Measuring starts only when every server passes. If one fails, fix what it says (see If something goes wrong) and start again.
- The measuring runs on SQL Server itself, so you can close the collector. To come back, open the collector with the same list and setup and start again: it reconnects to the running collection.
- Select Check now to see progress. Export opens when every server has finished.
- If a server still has an earlier collection, you choose: Export keeps its data, Clean up removes it (you are asked first).
CloudWatch collector
Reads RDS settings, CloudWatch metrics, Enhanced Monitoring and Database Insights from AWS. It changes nothing in your AWS account, creates nothing on your servers, and your AWS credentials never leave your computer. It collects 14 days by default; you can choose 30 days, the most.
Set up the AWS CLI
- Install AWS CLI version 2 from Amazon's AWS CLI page (search "install AWS CLI v2").
- Open PowerShell and run
aws configure. It asks for an access key, a secret key and a default Region. Your AWS administrator gives you the keys for an identity with the policy below.
Read-only permissions
These are the only AWS calls it makes. Ask your AWS administrator to attach this policy to the identity you use with aws configure.
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "SavioCloudWatchCollectorReadOnly",
"Effect": "Allow",
"Action": [
"ec2:DescribeRegions",
"rds:DescribeDBInstances",
"rds:DescribeDBParameters",
"cloudwatch:GetMetricData",
"logs:GetLogEvents",
"pi:GetResourceMetrics",
"pi:ListAvailableResourceMetrics"
],
"Resource": "*"
}
]
}
| Permission | What it is for |
|---|---|
ec2:DescribeRegions | Finds the AWS Regions to look in. |
rds:DescribeDBInstances | Finds your RDS SQL Server instances, their size, storage settings and license model. |
rds:DescribeDBParameters | Reads SQL Server's max memory setting. Without it, Savio uses the installed memory and says so. |
cloudwatch:GetMetricData | CPU, free memory, disk activity and latency, and the tempdb readings. |
logs:GetLogEvents | Enhanced Monitoring (memory and vCPU), where it is turned on. |
pi:GetResourceMetrics, pi:ListAvailableResourceMetrics | Database Insights counters (page life expectancy, memory grants), where it is turned on. |
For the most complete result, turn on Enhanced Monitoring and Database Insights in the RDS console. Database Insights keeps 7 days by default; the collection notes say how many days of page life expectancy came back.
Your files
- The collector writes
Savio_ZIP files to the folder you chose. The last screen lists them, with Show in folder. - Upload them at app.usesavio.com/cost/assessment.
- After exporting, select Clean up to remove the collector's tables and job from your servers. Your files are kept. You can then delete the collection database.
If something goes wrong
Select Show activity at the bottom of the collector to see each step and each server's result. These are the messages you are most likely to see:
| The collector says | What to do |
|---|---|
| server: login failed. Check the Login and Password for this server. | The login name or password is wrong for that server. Check them in your list, or in the one-login boxes. |
| Administrative database '…' does not exist. | The collection database is missing on that server, or the name in your list is misspelled. Create it. |
| Administrative database '…' exists, but this login cannot access it. | The login has no user in that database. Sign in with the master user that created the database. |
| … lacks CREATE TABLE or ALTER permission on schema dbo … | The login cannot create the collector's tables. Sign in with the master user. |
| … not a member of an approved SQL Agent role in msdb. | Sign in with the master user, which is already in this role. |
The collector cannot reach a server at all
- In the RDS console, check that the instance's status is Available.
- Check that the endpoint in your list is exactly the one on the Connectivity & security tab.
- The instance's VPC security group must allow connections from this computer. If SSMS on this computer cannot connect either, ask your AWS administrator to check the security group's inbound rules.
Support log
Each session writes a support log, SavioCollector_<date>_<time>_<id>.log, in the collector folder. It never contains passwords, and logins are hidden. Send it to Savio support if asked.
Words used here
| Endpoint | The address of an RDS instance, used to connect to it. |
| Instance | One RDS database server. Your list has one row per instance. |
| Master user | The administrator login set when the RDS instance was created. |
| Login | A user name and password SQL Server accepts. |
| SQL Server Agent job | A task SQL Server runs on a schedule by itself. The collector's job takes a reading every minute. |
| msdb | SQL Server's built-in database that stores scheduled jobs. |
| Collection database | The small database you create for the collector's tables. See above. |
| SSMS | SQL Server Management Studio, Microsoft's free tool for connecting to SQL Server. |
| CSV | A plain table file that Excel opens and saves. |
| AWS CLI | Amazon's command-line tool. The CloudWatch collector and Build from AWS use it to read from your AWS account. |