Think about it this way: OCZ has on paper SSDs that are simultaneously both faster and cheaper than, say, Intel. But I wouldn't put a OCZ drive into the lowest budged ricer gaming PC I could imagine (let's just say that have a bit of a reputation, to put it lightly), and I should be rightfully fired if I suggested putting one in a server.
Yes, your service might be massively faster, but... at 10-15% cheaper? No way! Not until you've been running for years with a great reputation. Now, if you enter the managed database space and pull a Digital Ocean ($5-$40 pricing vs comparable EC2 at $80-$400 AND better raw performance on top), then you can essentially create a new market niche. That's my suggestion to focus on, because businesses who operate 50TB databases aren't looking to save a few dollars. But people who use DO, need a database to go with it that is both faster and cheaper than AWS.
Honestly, performance ranks pretty low on the things I've found people look for when choosing a managed database service. You can serve a million users per day and average a whopping 12 IOPS, which might just about push a Raspberry Pi with sqlite slightly. Got more than a million users? Congrats, you now officially qualify to throw money at problems, the kind of problems RDS and Amazon loves to solve by injecting money.
The latest OCZ Trion series at least honor O_DSYNC correctly. You get shitty write performance, but it is the same for Samsung 950 Pro. The read performance is another story for OCZ though :(
Anyway enterprise SSD's have so much overall better performance that consumer grade SSD's is something I consider a joke. A joke even for consumer applications, web-surfing and email-clients. Remember that a lot of applications out there use SQLite, and poor DB performance means poor application performance.
You do not have an option to host anything larger then 6TB on RDS. I guess what we are trying to figure out is if there is a need for service for people who have needs beyond those RDS is addressing.
Not only disk space, you are limited to just two nodes (one primary and one standby). Of course you can create read replica, but that if you are write intensive, you can end up consuming all of your pools.
actually you would shard than, which of course is possible with rds aswell. since you mostly shard at the application level. there are some users that use extensions but I guess calling the support of aws will get these extension in your rds instances.
Think about it this way: OCZ has on paper SSDs that are simultaneously both faster and cheaper than, say, Intel. But I wouldn't put a OCZ drive into the lowest budged ricer gaming PC I could imagine (let's just say that have a bit of a reputation, to put it lightly), and I should be rightfully fired if I suggested putting one in a server.
Yes, your service might be massively faster, but... at 10-15% cheaper? No way! Not until you've been running for years with a great reputation. Now, if you enter the managed database space and pull a Digital Ocean ($5-$40 pricing vs comparable EC2 at $80-$400 AND better raw performance on top), then you can essentially create a new market niche. That's my suggestion to focus on, because businesses who operate 50TB databases aren't looking to save a few dollars. But people who use DO, need a database to go with it that is both faster and cheaper than AWS.
Honestly, performance ranks pretty low on the things I've found people look for when choosing a managed database service. You can serve a million users per day and average a whopping 12 IOPS, which might just about push a Raspberry Pi with sqlite slightly. Got more than a million users? Congrats, you now officially qualify to throw money at problems, the kind of problems RDS and Amazon loves to solve by injecting money.