SAS in the cloud - Amazon Redshift & EC, SAS Cloud Analytics

[sashannover]

Hi, I have been using SAS for over a year with an on-premise Server-based solution with roughly 10 users. From my experience so far, the I/O has been the bottleneck when it comes to multiple queries. For a new project I am playing with the idea to set it up as a cloud solution. The options I see so far are:

• Amazon Redshift (Data Warehouse) + Amazon EC2 for the SAS virtual server

• SAS Cloud Analytics + Amazon Redshift (Data Warehouse)

• maybe others?

My questions so far:

• For a set-up of 10 SAS users what are the options for building a cloud SAS solution for analytics and information marts?

• What should be considered when calculating the costs? Is there a calculator that does rough estimates based on users, data size, expected queries?

• What are the differences in performance for on premise vs in the cloud?

• What are the differences in terms of pricing?

At what point does it make sense to have an on-premise solution?

Thank you in advance!

[Reeza]

How big is your data? That’s usually the most significant issue.  You may be able to get a significant performance gain by connecting SAS to a Hadoop installation as an alternative.  SAS sales is super helpful in this regard. 

[JuanS_OCS]

Hello @sashannover,

@Reeza already gave you great insights. Please let me provide mine as well:

- In one hand, I wonder what I/O troubles did you found in premises and I wonder as well why they could not be solved.

- Moving to the Cloud/Amazon won't solve your I/O troubles necessary. Because of the ephemeral storage. Only one kind of offering by AWS/Amazon would solve this ( as pointed out by @MargaretC in https://communities.sas.com/t5/Administration-and-Deployment/Important-Performance-Considerations-Wh... , a great post for SAS/AWS!)

In the end, all goes to fall into a proper sizing questionnaire and sizing proposal document/design, either on premises or the Cloud.

the information of 10 users won't be enough, the load, kind of queries, data outputs and your own company security policies are very important for this sizing design.

The best you can do is to contact your SAS Sales representative or SAS Account manager you they put you in contact with SAS EEC (SAS Enterprise Excellence Center) which is the team responsible of providing the best reliable sizing designs. Margaret Crevar probably would be involved if your question is regarding Cloud/AWS.

Please read:

https://blogs.sas.com/content/sgf/2013/11/27/sas-deployment-whats-the-right-size/

https://blogs.sas.com/content/sascom/2014/03/13/something-old-something-new-planning-for-a-happy-mar...

https://communities.sas.com/t5/Administration-and-Deployment/SAS-Sizing-Questionnaire/td-p/140480

http://support.sas.com/resources/papers/proceedings12/363-2012.pdf

https://communities.sas.com/t5/Administration-and-Deployment/Two-new-storage-array-tuning-papers/m-p...

[sashannover]

Thank you both for the quick response!

@JuanS_OCS thanks a lot for the vauable ressources! I will go thriugh them and hopefully will get a better picture. You are right that data and type if querries are the more important parameters.