Performance Tuning

Aspera has worked hard to create a transfer cluster that works well out of the box. The default configuration of the system is designed to maximize the price-to-performance ratio. However, if you require transfer speeds significantly faster than 1 Gbps, or if you are transferring hundreds of thousands of files then some tuning can improve cluster performance.

Your ATC cluster nodes can be optimized for maximum performance by changing certain configuration settings and selecting an instance type that supports your desired bandwidth usage.

Optimizing Instance Type

AWS offers several instance types (see https://aws.amazon.com/ec2/instance-types/) and you can select an instance type to support your transfer needs:

If most of your transfers are less than 1 GB, Aspera recommends using General Purpose M4 instances for your nodes.
If your transfers typically exceed 1 GB, Aspera recommends using Compute Optimized C4 instances for your nodes.

Increasing Bandwidth Cap (Vlink)

The default cluster node templates are configured with a vlink that sets a bandwidth cap on all transfers to 750 Mbps. To achieve transfers that exceed this limit, edit the vlink configuration either in the cluster node template before launching the cluster or in the template of a running cluster.

Open the ATC Manager UI.
If you are creating a new template, create a duplicate template and open it for editing.
Click Templates. Duplicate the template of your current template or a default template by clicking Action > Duplicate.
For the newly created template, click Action > Edit Template.
If you are modifying a running cluster, open the cluster transfer configuration for editing.
Click Action > Edit Transfer Configuration.
Edit the vlink configuration.
Locate the following section in the transfer configuration:
```
<trunks>
        <trunk>
            <id>70</id>
            <capacity>
                <schedule format="ranges">750000</schedule>
            </capacity>
            <ignore_direction>true</ignore_direction>
            <on>true</on>
        </trunk>
    </trunks>
```
Change the value for <schedule> from 750000 kbps to your desired transfer speed, in kilobits per second.
Click Save Changes.

Note: If you are updating the transfer configuration of a running cluster, saving your change updates the configuration file on all cluster nodes and restarts Aspera services. Any active transfers are stopped and might be resumed, depending on the client's configuration. Transfers that use the Connect Plugin retry every 30 seconds, up to three times, after a transfer fails.

Disabling Non-critical Features

The default configuration for cluster nodes has several features that can be disabled with little impact on their usability in exchange for a decreased workload on the Aspera Node daemon. You can modify a template before a cluster is launched, or modify the transfer configuration of a running cluster, as described in the previous section.

Disable activity logging

Cluster nodes report bandwidth usage in real time by default. Disable this reporting by locating the following line in the transfer configuration and changing the value from true to false:

<activity_bandwidth_logging>true</activity_bandwidth_logging>

Note: When bandwidth activity logging is disabled, you cannot monitor bandwidth usage.

Disable pre-calculate job

Cluster nodes attempt to estimate how long a transfer will take based on the size of the transfer and real-time available bandwidth. This is a parallel process to transfers that can decrease transfer performance in some scenarios. Disable pre-calculating transfer time by locating the following line in the transfer configuration and changing the value from yes to no:

<pre_calculate_job_size>yes</pre_calculate_job_size>

Reducing the Activity Retention Period

Cluster nodes track transfer activity and this information is collected by the ATC Manager and stored in a MySQL database. You can decrease the amount of transfer data that is managed by the ATC Manager by decreasing the length of time that activity information is stored on the nodes. Modify a template before a cluster is launched, or modify the transfer configuration of a running cluster, as described in the previous section.

Decrease the activity retention period by locating the following line in the transfer configuration and changing the value from six hours to no less than one hour (1h):

<activity_retention>6h</activity_retention>