Clearly RDS prints money - doesn’t take a genius to realize AWS has been trying to duplicate that success with every other OSS db-aaS they’ve launched since.
Given its success, lack of pricing pressure makes sense - until users go away, there’s little reason to decrease margins.
RDS is quite expensive, even for the peace of mind it gives.
I used to work on a big enterprise which had many talented engineers and it was clear that RDS for the many dozens of teams wouldn’t be the best idea (money-wise) and so a custom HA solution was built and saved a lot of money.
At a certain scale it may be a good idea to find alternatives but as usual YMMV.
RDS is expensive, but for a small team of 5 devs (my current setup) it is an amazing solution.
Scaling up and down the database is as simple clicking a button on the UI, which we used quite effectively when our services saw a sudden spike in the load. We were able to scale down the same way after the spike dropped.
We still had to tune our queries and indices so that we weren't paying the cost over the long term, but the peace of mind that comes with not having to figure out why your database went down when there is no one in the office is a huge win-win.
We currently also manage an OLD version of Cassandra and Elasticsearch and we are also looking to migrate to the new managed version of Cassandra. At the end of the day, the extra cost for RDS even for small teams, is much lower than the cost in terms of developer-hours * cost per hour for developer.
Clickbait title. RDS pricing has doubled relative to the instance pricing. RDS absolute pricing has gone down, just not as much as the instance pricing. RDS may have some fixed cost (or slower moving variable cost) component separate from the raw instance cost.
Your article does not support the idea of the cost having doubled - hence (I assume) the suggestion of clickbait.
In order to show that the pricing has doubled, you'd need to show an increase in actual monetary cost to a customer, not that the margins have changed. The table of prices has no historical data, making change impossible to determine.
I'm sympathetic to the idea that RDS is overpriced, but can't say I'm convinced by this article alone.
RDS is a value-add on top of EC2. Therefore its actual price is the extra premium.
If a food delivery company doubles their fees from 10% to 20% and the restaurant cuts its price proportionally, do you really say the delivery price is the same?
The total cost remains the same but the value the customer receives from the delivery service per dollar spent has been cut in half.
Back when RDS did not have Postgresql supported, our dev team build something similar on top of EC2, EBS, Wal-e (S3) and XFS.
It was a lot of work to get all the moving parts working nicely; lot of testing to perfect all procedures and I can tell you, it was no fun at all when we got the instance shutdown email notification from AWS (or if you wanted to spin a new server from snapshot / upgrade instance type, ...)
Here's a list of things that you get with RDS:
- Somebody else will make sure that failover, snapshotting, backup recovery work. If you use EC2 this is going to be something you need to test periodically (unless you like bad surprises)
- Automated failover election mechanism
- Logging integrates with S3 out of the box
- Proxy (no need to run your own pgbouncer anymore)
- Engine upgrades work with replicase and multi-az failovers
The list is much longer than this but my point is: the amount of overhead that comes with running PG on EC2 is high and does not improve that much over time (probably the opposite is true if you consider that developers/DBAs come and go).
Not saying it’s a bad deal. RDS is a great service. I’m saying that suddenly doubling the cost of a service is not a good look unless the value delivered is also doubled. As a customer, in what situation would you find this acceptable?
I seem to miss something here? If RDS went from $0,465 to $0,342 per hour as shown in the table, that’s roughly a decrease by one third in cost for me. Where does the „suddenly doubling the cost“ part come from? Isn’t it that the total cost for me decrease, just not at the rate of other AWS offerings?
> I’m saying that suddenly doubling the cost of a service is not a good look unless the value delivered is also doubled.
That's not how markets and pricing work. Prices can go up for any number of reasons, demand is high, demand is too low, past pricing was too aggressive. Amazon has historically been operated with very low profits, maybe Bezos wants to change that?
I'm not sure why you feel offended or deserve a certain price point. Go with some other cloud provider, and if it's too much work, well then there you have it: Amazon has just justified their RDS pricing.
AWS has always had high margins. And I don’t care if they are in a market position to dictate the short-term price, I believe it’s bad business over the long term. Charging more for the same thing is certainly not customer obsession.
Oh, and thanks for the middle school economics lesson.
We did the same, self hosted to Rds was so much easier. I didn’t think it had a proxy though, and the new proxy is still just MySQL, so still needs pgbouncer?
This was a while ago, maybe even before i families, but our boss refused to use ebs at the time because of the ebs failure a couple years before that. What changed his mind was a instance termination that took way to long to recover from, and there been a ton of ebs improvement since then
1) db.m3.Large, db.m4.large, and db.m5.large are all around $130/month, no where near a 20% difference between any one of them.. so to not enough information to find this supposed 78% premium
2) AWS is totally for the cost concerned.... But you can't just lift and shift... That's exactly how NOT to save money. I run many projects out of AWS for less than $5/month ... Which is the usual expected cost for a web server. If you can leverage something like Dynamo DB for your DB you can get cheaper databases
However, everything in AWS is a pricing game. API gateway is cheap until you get to 5 million requests pet month, then you should look at ELB proxy... Unless you really need API gateway's additional features... Though HTTP proxy just came out and I think that is a good replacement for many people for API gateway which reduce costs 70% and makes it a good low cost front end listener up to 15 million requests or so before you may want to consider a load balancer instead (of course, assuming the features line up with needs)
The article said m3 to m5 went up 78%, I don't see that.
Another reader pointed out the author is comparing on demand ec2 to RDS... That wasn't clear to me when I made my statement. Party my fault for not reading the table, it's rather small on my phone.
10 years ago developers and ops people were a lot more concerned about being locked into proprietary products especially to the sort of level of integration AWS brings.
Eventually I suspect Amazon will be all but required to chase the money and at that point a lot of companies will really struggle to get off it. OS/360 is still going, still charging a fortune for a mainframe.
Except it's not, not really -- it's just not kept pace with the decrease in underlying EC2 cost. It still costs less now for an RDS instance than it did before.
The premium over raw EC2 has gone up -- from $0.115/h to $0.15/hour -- but that's only a 30% jump. The total cost has gone down by $0.123/hour, which is ~25%.
29 comments
[ 3.2 ms ] story [ 75.7 ms ] threadGiven its success, lack of pricing pressure makes sense - until users go away, there’s little reason to decrease margins.
I used to work on a big enterprise which had many talented engineers and it was clear that RDS for the many dozens of teams wouldn’t be the best idea (money-wise) and so a custom HA solution was built and saved a lot of money.
At a certain scale it may be a good idea to find alternatives but as usual YMMV.
Scaling up and down the database is as simple clicking a button on the UI, which we used quite effectively when our services saw a sudden spike in the load. We were able to scale down the same way after the spike dropped.
We still had to tune our queries and indices so that we weren't paying the cost over the long term, but the peace of mind that comes with not having to figure out why your database went down when there is no one in the office is a huge win-win.
We currently also manage an OLD version of Cassandra and Elasticsearch and we are also looking to migrate to the new managed version of Cassandra. At the end of the day, the extra cost for RDS even for small teams, is much lower than the cost in terms of developer-hours * cost per hour for developer.
In order to show that the pricing has doubled, you'd need to show an increase in actual monetary cost to a customer, not that the margins have changed. The table of prices has no historical data, making change impossible to determine.
I'm sympathetic to the idea that RDS is overpriced, but can't say I'm convinced by this article alone.
If a food delivery company doubles their fees from 10% to 20% and the restaurant cuts its price proportionally, do you really say the delivery price is the same?
The total cost remains the same but the value the customer receives from the delivery service per dollar spent has been cut in half.
Back when RDS did not have Postgresql supported, our dev team build something similar on top of EC2, EBS, Wal-e (S3) and XFS.
It was a lot of work to get all the moving parts working nicely; lot of testing to perfect all procedures and I can tell you, it was no fun at all when we got the instance shutdown email notification from AWS (or if you wanted to spin a new server from snapshot / upgrade instance type, ...)
Here's a list of things that you get with RDS:
- Somebody else will make sure that failover, snapshotting, backup recovery work. If you use EC2 this is going to be something you need to test periodically (unless you like bad surprises)
- Automated failover election mechanism
- Logging integrates with S3 out of the box
- Proxy (no need to run your own pgbouncer anymore)
- Engine upgrades work with replicase and multi-az failovers
The list is much longer than this but my point is: the amount of overhead that comes with running PG on EC2 is high and does not improve that much over time (probably the opposite is true if you consider that developers/DBAs come and go).
That's not how markets and pricing work. Prices can go up for any number of reasons, demand is high, demand is too low, past pricing was too aggressive. Amazon has historically been operated with very low profits, maybe Bezos wants to change that?
I'm not sure why you feel offended or deserve a certain price point. Go with some other cloud provider, and if it's too much work, well then there you have it: Amazon has just justified their RDS pricing.
Oh, and thanks for the middle school economics lesson.
Back when we did this instance storage was much faster than EBS but the difference was not as dramatic as it can be these days. For example:
i3.2xlarge -> 180,000 write IOPS (4KB) 456$/month EBS 16TiB volume -> 16,000 IOPS (16KB) (more than 1,500$/month)
The difference in price and performance (NVMe can go much higher than this) could justify the EC2 self-hosted approach.
1) db.m3.Large, db.m4.large, and db.m5.large are all around $130/month, no where near a 20% difference between any one of them.. so to not enough information to find this supposed 78% premium
2) AWS is totally for the cost concerned.... But you can't just lift and shift... That's exactly how NOT to save money. I run many projects out of AWS for less than $5/month ... Which is the usual expected cost for a web server. If you can leverage something like Dynamo DB for your DB you can get cheaper databases
However, everything in AWS is a pricing game. API gateway is cheap until you get to 5 million requests pet month, then you should look at ELB proxy... Unless you really need API gateway's additional features... Though HTTP proxy just came out and I think that is a good replacement for many people for API gateway which reduce costs 70% and makes it a good low cost front end listener up to 15 million requests or so before you may want to consider a load balancer instead (of course, assuming the features line up with needs)
Another reader pointed out the author is comparing on demand ec2 to RDS... That wasn't clear to me when I made my statement. Party my fault for not reading the table, it's rather small on my phone.
r5.large: Aurora got 20% premium over MySQL
1. Aurora: $0.29 / hour
2. MySQL: $0.24 / hour
Though Aurora you don't need to manually provision disk space and can have higher performance.
https://aws.amazon.com/rds/aurora/pricing/
https://aws.amazon.com/rds/mysql/pricing/
Eventually I suspect Amazon will be all but required to chase the money and at that point a lot of companies will really struggle to get off it. OS/360 is still going, still charging a fortune for a mainframe.
The premium over raw EC2 has gone up -- from $0.115/h to $0.15/hour -- but that's only a 30% jump. The total cost has gone down by $0.123/hour, which is ~25%.