I recently migrated my organizations SAS production environment to Amazon Web Services (AWS) and although I particularly enjoy the flexibility cloud-based virtualization offer for OS and application management, the storage is something less desired. To be frank, AWS storage, whether gp2, io1, hdd, softnas, etc, are not suitable for SAS workload, or any application that required fast iops. So my question is directed to those who have made the jump to the cloud (speaking in board terms and not just AWS). What recommendations do you have regarding storage? Is there something suitable to test that will help get better throughput? I've spoken with AWS EBS developers (with one of my favorite SAS engineers - Margaret Crevar) and it seemed as if they don't have a solution in sight. I asked them for dedicated storage channels like ephemeral, but persistent.
Thanks for any feedback!
Thanks Resa! Were you referring to this video?
https://communities.sas.com/t5/Administration-and-Deployment/Shared-File-Systems-webinar-recording-n...
The webinar is not applicable to Amazon. It is only applicable to on-premise hardware infrastructures.
Hi Margaret - thanks!
Have you heard anything back from Amazon? I also haven't heard if the ephemeral issue with the i3.* RHEL machines is fixed yet.
Yes, the ephermal issues with the AWS EC2 I3 have been resolved.
You need to use RHEL 7.3 and apply yum update -security to get the fix. Or you can use RHEL 7.4 which has the fix in it.
Just going through this process myself and was curious if you ever found a satisfactory solution? What kind of iops are you looking for?
Hello @RobP,
since one year ago, there has been a huge development and testing at SAS, due to the high amount of customers going to the Cloud, being Margaret Crevar ( @MargaretC ) probably the first point of contact for all of those topics.
First of all, the main point to consider, indeed, is that Cloud solutions are based on ephemeral resources, which makes sense for many applications, but if you look for any application that requires high performance, you need to align well with your cloud provider and SAS, to get the best possible out there, no matter if AWS, Google, or Azure. Otherwise, having ephemeral storage, it is impossible to know if you will get a storage volume that can provide 2000 MB/sec/core or 500 MB/sec/core throughput , as example. A huge difference.
To get the most basic and important concepts, you could start from those references:
https://www.sas.com/content/dam/SAS/en_us/doc/whitepaper1/sas-cloud-computing-105233.pdf
http://support.sas.com/resources/papers/Implementing-SAS9-4-Software-Cloud-Infrastructures.pdf
http://support.sas.com/rnd/scalability/grid/SGMonAWS.pdf
For me, a possible and very simplified summary would be:
So, in short, for normal deployments, you should not have huge problems if you stick to some basic concepts. But if you look for high performance, you better get your Cloud provider and SAS technical teams aligned, in order to provide you the best advise at the moment of your question/requirement.
Edit: do not look for IOPS, but Throughput.
Thank you@MargaretC, for the part that concerns to me.
The SAS Users Group for Administrators (SUGA) is open to all SAS administrators and architects who install, update, manage or maintain a SAS deployment.
SAS technical trainer Erin Winters shows you how to explore assets, create new data discovery agents, schedule data discovery agents, and much more.
Find more tutorials on the SAS Users YouTube channel.