July 6, 2015 at 2:30 pm
We are in the process of moving existing clustered SQL server databases to AWS. There is one major database that has intensive reads and writes transactions. I'm wondering what is the best design to optimize the performance for both R/W since we have constant issues historically with the current environment when massive updates are happening. Reads shall have higher priority over writes.
July 7, 2015 at 6:51 am
Michelle-138172 (7/6/2015)
We are in the process of moving existing clustered SQL server databases to AWS. There is one major database that has intensive reads and writes transactions. I'm wondering what is the best design to optimize the performance for both R/W since we have constant issues historically with the current environment when massive updates are happening. Reads shall have higher priority over writes.
My first question is whether or not your bandwidth to AWS can handle the I/O traffic as well as your existing infrastructure, because unless it's latency is at least as good as what you already have, you're going to be hurting big time. Of course, different people often have very different ideas about what constitutes "intensive reads and writes", so I have to ask just what kind of volume you're dealing with. Recognize that moving to AWS is not necessarily a whole lot different than putting in a SAN where one used to have directly attached disk drives... There's a sizable difference in the available I/O bandwidth and it has to be very carefully implemented. As you've provided no detail on just what kind of a move to AWS you're dealing with, it would be difficult to know whether to worry more about their server(s?) being the capacity equal or better to yours, or their bandwidth to connect to you or your customers (or both) is sufficient to handle the load, or if you're just outsourcing your disk storage there. Another rather important thing is the SLA's (aka Service Level Agreements) you agree to. And probably more important than the rest, your data is now in someone else's hands, so what will you do to secure it against loss or data breach?
Steve (aka sgmunson) 🙂 🙂 🙂
Rent Servers for Income (picks and shovels strategy)
July 7, 2015 at 8:22 am
Michelle-138172 (7/6/2015)
We are in the process of moving existing clustered SQL server databases to AWS. There is one major database that has intensive reads and writes transactions. I'm wondering what is the best design to optimize the performance for both R/W since we have constant issues historically with the current environment when massive updates are happening. Reads shall have higher priority over writes.
What are you planning on implementing in AWS? Are you going to be running SQL Server instances on EC2 Instances or on the RDS Platform?
If your planning on going down the RDS route you will have to provision IOPS to support your IO requirements. IOPS can be provisioned from 1000 to 30000. If you provision too little you can increase this value (along with the total storage of the instance), however if you over provision its harder to reduce the amount. The higher the provisioned IOPS the higher the cost.
You cannot dictate read and write IOPS seperatly as the value provisioned is accumulative of both so its not possible to prioritise read over write or vice versa. However you can eleviate read replicas to help with read requirements.
MCITP SQL 2005, MCSA SQL 2012
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply