2

We are currently in the process of moving our dev and production work on AWS. I was assigned to create the necessary VMs and databases, so I am reading here and there about these, since this is my first aws attempt.

Firstly, I noticed that I can not create serverless Aurora instance under some regions. Paris for example does not support it, where Ireland does.
We are based in Greece, so I was thinking to use the closest region (by distance) from us, and this would be the Paris region. So I created a normal Mysql database there, not realising that the serverless option was not around to choose. Then I have read that the serverless instances can not be accessed by an outside host, only from VM inside the same cluster, or an ssh tunnel to be used. Just because we have been using a private server that is behind an ssh tunnel and it was a pain to connect to at the beginning, I am not a great fan of it. Let alone that I have spend some good time to set up the jdbc connection for it on my java applications. Is that correct?

So, the real questions,

  • Is the serverless approach worth the trouble? Is it better compared to the rest of the options there?
  • In any case, serverless or not, the rest of the VMs should be on the same region/cluster? I am guessing that it would be faster if everything is under the same region, but is there any reason to scatter the processes to different regions?
  • By using serverless I am charged by the power I am using. So I can choose the max available ram and if not used, I would be charged by a smaller fee. Does the same apply on the other options?
  • On the region where the serverless option is available i am given the choice of Provisioned and Provisioned with Aurora parallel query enabled, where on the Paris region I got no such option. I am guessing that there are no options because the Provisioned is selected as default. Am I right?

Thanks in advance, Ilias

Improve this question

1 Answer 1

Reset to default
2

Aurora Serverless is good for infrequently used databases. Something that sits around in dev/test and is only accessed from time to time. I wouldn't recommend it e.g. for a public website, because there is some time overhead when the DB is accessed for the first time after it goes idle. That may lead to quite a poor user experience.

If I were you I would use standard non-serverless Aurora in whatever region you prefer. Once you're comfortable with that you can look at switching to Serverless Aurora for your non-production environments, but start with the normal one for ease of use.

Regarding your other questions:

Ireland vs Paris region

I doubt you will notice a difference between running your workload from Paris, Frankfurt or Ireland. Europe is very well interconnected and the latency difference will be minimal. Furthermore if you're deploying a website you can reduce the end user latency by using Cloud Front content delivery network.

the rest of the VMs should be on the same region/cluster?

Keep them all in the same region because that will make your networking much easier. Resources in the same region can communicate with each other (if they are in the same VPC - Virtual Private Cloud). Resources in different regions need VPNs or VPC peerings set up in order to talk to communicate with each other - that's probably not worth the trouble unless you've got some advanced needs.

choice of Provisioned and Provisioned with Aurora parallel query enabled

Unless you're doing some Big Data analytics you most likely don't need parallel queries. Go ahead with standard Provisioned database.

Hope that helps :)

Improve this answer
3
  • Thanks MLu. Very helpful info. Actually, we are into big data business, so I might read about Aurora parallel query if it is worth using on our main database. We are only going to have one database connected to web server, the rest might as well go into serverless, the one connected to the web page inot a normal Provisioned DB and our linked data DB to a parallel DB. Thanks again
    – Skaros Ilias
    Oct 17, 2018 at 6:32
  • Aurora serverless doesn't support being a replication slave, this isn't documented anywhere but has been confirmed by AWS support. So this will make migration more difficult.
    – Rwky
    Nov 14, 2018 at 14:27
  • One thing I noticed when I was working with Serverless Aurora, trying to do a load with a large mysqldump (one with 80000 rows in a single table with LOBs) seem to make it hang. I am trying to see if using a provisioned one works better.
    – Archimedes Trajano
    Dec 3, 2018 at 22:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .