Aspera On Demand sizing guidelines

Summary

This article provides sizing guidelines to help you plan your On Demand deployments.

Details

This table contains the results of the most recent testing and validation on various platforms. We will continue to update the table as we gather more information.

Additional details about the testing parameters can be found in the pdf attached to this article. The performance numbers are provided as ranges, as the results are not consistent across all images/regions etc.

IaaS Provider Image type Image dimension info
Concurrency Max Bandwidth (mbps) Max File size (default)
Max Session size
Comments
AWS m3.large 7.5 GB memory, 2 vCPU 12 (estimated) 200 - 400  640 GB (1)    
AWS m3.xlarge 15 GB memory, 4 vCPU 25 300 - 500  640 GB (1)    
AWS c3.8xlarge 60 GB memory, 16 vCPU (not tested)

600-800

1.5 Gbps (6)

 640 GB (1)   High speeds on the 3.6 AOD release HVM images
Softlayer Virtual (CCI) 8 GB memory, 2 x 3 GHz vCPU (not tested) 600 Mbps TBD    
Softlayer Bare Metal 24 GB memory, 24 3 GHz vCPU (not tested) 1,800  TBD   Transfer to local disk
Google  n1-standard-2 2 vCPUs (Intel Haswell), 7.5 GB memory (not tested)  600 62 GB (2) 625 GB (2)  
Azure Aspera on Demand   6 (5) 350 (4) 200 GB (3) 1,800 (See note) Faster speeds coming soon.

 

(1) 640 GB is the default size.  You can increase this by modifying the chunk size setting in aspera.conf (details coming soon).

(2) This limit can be changed by modifying the chunk size setting in aspera.conf (details coming soon).

(3) This limit is due to the architecture of 'block blobs'. More info on this here.

(4) This limit is approximate per session.  With multiple sessions we should be able to scale out linearly.

(5) This concurrency limit is approx and only per transfer node. This is not an aggregate concurrency to the entire cluster. The maximum aggregate concurrent transfers will depend on the number of nodes running in the cluster.  

(6) As of the 3.6 release, using the c3.8xlarge images, you can achieve up to 1.5 Gbps (or more!).

Attachments

Have more questions? Submit a request

0 Comments

Please sign in to leave a comment.
Powered by Zendesk