Great video highlighting the key points of difference between aurora db and postgres. When you talk about "why not always use aurora db" you mention its mainly understanding if we are comfortable with the aurora db architecture of having separate storage and compute nodes. Could you please elaborate on what you mean by being comfortable? When my application supports gcp or azure too, would I need any application code changes because I connected to aurora db till then? If so what would they look like? Thanks.
Hi Amit, You shouldn't need any app changes if you are using something like a JDBC connection. Usually the issue, in regards to comfort, stems from companies that have strick security policies, or are heavily audited. Contracts, audit documentation or even external policies may state that storage must be on the same instance as compute. And issues then arise moving away from this will trodden path.
@@JohnnyChivers - Hi Johnny, I read more about aurora and found that the storage nodes are allocated only within the aws region so the data should never cross the region boundary. Even in a disaster recovery or multi-master setup where a secondary region is brought up, the data does not cross the boundary of that secondary region. I am not able catch the details of what security policies, audits or contracts could create a roadblock when choosing aurora over rds. It would be very helpful if you could elaborate more. Appreciate your time. Thanks
Today, using the same resources, aurora is 8$/month more expensive and if you go with equivalent serverless resources, it's more then 100$/month more expensive. Performance-wise, I only tested so far with serverless and it's disappointingly slow compared to the classic RDS - less then half transactions/s 112 (RDS) vs 47 (aurora) with more than double the latency of each transaction - 446 ms (RDS) vs 1050 ms (aurora serverless v2)
Serverless is 3x the cost of provisioned for the same specs. It’s not meant to be cheaper pound for pound. However, it can scale without downtime and very good if you have bursty traffic.
In your cost comparison, did you just compare an RDS multi-AZ instance over to a single instance cluster Aurora? Single-AZ RDS instance is more cheaper compared to a single instance cluster Aurora IMO.
Yip, it’s a fair comment. I picked this comparison because it’s the two default options. You actually get a lot higher throughput on the aurora side and the data is back up across three AZs. Admittedly the aurora instance takes longer to come back if you lose the single instance as you don’t have automatic fail over. So, what you lose in auto compute fail over for aurora you make in data redundancy. Your point is 100% valid. It just seems the AWS default is to compare them this way as a single, or multi node, comparison is not a 1 for 1 anyway. You have to love the nuances of AWS.
I think it's just a product naming nuance where we have AWS Aurora. Then within AWS Aurora we have both MySQL and PostgreSQL - so they decided to use the word 'compatible' to distinguish between the two Engines within Aurora.
Hi Johnny, What do you recommend for wordpress database (Amazon Aurora MySQL-Compatible Edition or Amazon Aurora PostgreSQL-Compatible Edition), considering ec2 horizontal scaling.
Great explaining
Thanks
Nice summary, great bit o' business
Glad you liked it!
Today, Aurora is costlier and Aurora serverless is even costlier!!
Very well explained!
Glad it was helpful and thanks for watching.
Great video highlighting the key points of difference between aurora db and postgres. When you talk about "why not always use aurora db" you mention its mainly understanding if we are comfortable with the aurora db architecture of having separate storage and compute nodes. Could you please elaborate on what you mean by being comfortable? When my application supports gcp or azure too, would I need any application code changes because I connected to aurora db till then? If so what would they look like?
Thanks.
Hi Amit,
You shouldn't need any app changes if you are using something like a JDBC connection. Usually the issue, in regards to comfort, stems from companies that have strick security policies, or are heavily audited. Contracts, audit documentation or even external policies may state that storage must be on the same instance as compute. And issues then arise moving away from this will trodden path.
@@JohnnyChivers - Hi Johnny,
I read more about aurora and found that the storage nodes are allocated only within the aws region so the data should never cross the region boundary. Even in a disaster recovery or multi-master setup where a secondary region is brought up, the data does not cross the boundary of that secondary region.
I am not able catch the details of what security policies, audits or contracts could create a roadblock when choosing aurora over rds. It would be very helpful if you could elaborate more. Appreciate your time.
Thanks
Today, using the same resources, aurora is 8$/month more expensive and if you go with equivalent serverless resources, it's more then 100$/month more expensive. Performance-wise, I only tested so far with serverless and it's disappointingly slow compared to the classic RDS - less then half transactions/s 112 (RDS) vs 47 (aurora) with more than double the latency of each transaction - 446 ms (RDS) vs 1050 ms (aurora serverless v2)
Serverless is 3x the cost of provisioned for the same specs. It’s not meant to be cheaper pound for pound. However, it can scale without downtime and very good if you have bursty traffic.
If you want to make a pound for pound comparison, it would be better to compare provisioned aurora vs postgres using the same specs
@@yangli8142 I did and the performance is not there - about 10-20% less on aurora then RDS on the same instance
Can you do local development with aurora without connecting to aws
In your cost comparison, did you just compare an RDS multi-AZ instance over to a single instance cluster Aurora?
Single-AZ RDS instance is more cheaper compared to a single instance cluster Aurora IMO.
Yip, it’s a fair comment. I picked this comparison because it’s the two default options. You actually get a lot higher throughput on the aurora side and the data is back up across three AZs.
Admittedly the aurora instance takes longer to come back if you lose the single instance as you don’t have automatic fail over. So, what you lose in auto compute fail over for aurora you make in data redundancy.
Your point is 100% valid. It just seems the AWS default is to compare them this way as a single, or multi node, comparison is not a 1 for 1 anyway. You have to love the nuances of AWS.
Please help understand why Aurora is categorically described as "postgreSQL compatible" and not AS PostgreSQL DB engine?
I think it's just a product naming nuance where we have AWS Aurora. Then within AWS Aurora we have both MySQL and PostgreSQL - so they decided to use the word 'compatible' to distinguish between the two Engines within Aurora.
Hi, thank for video. Does it make a price difference if I select "Production" or "Development/Test" when install Aurora serverless 2 ?
Hi Johnny,
What do you recommend for wordpress database (Amazon Aurora MySQL-Compatible Edition or Amazon Aurora PostgreSQL-Compatible Edition), considering ec2 horizontal scaling.
is this dude, Jack Black, or what?
Just a tribute.