Unanswered: What to consider when moving MSSQL 2000 from Data Center to Amazon EC2?
We are in the process of migrating our system which is currently using a Data Center to Amazon EC2. Our current system architecture is as such:
- STAGING [Data originates here]
- PRIMARY [Needed data gets pulled into here from STAGING, using our custom pull jobs]
- SUBSCRIBERS [Data gets replicated using Transactional replication from PRIMARY(Publisher)]
All machines are dedicated servers and we have Failover Clustering set up to ensure minimal downtime.
We are using MSSQL 2000.
We have used Amazon EC2 instances before (although not for production) and performance was not as good as our current one. I did a bit of reading about amazon instances, and they claim that when we purchase an instance we get a dedicated CPU(s) and memory, but sub-system disk and network are shared.
I know that for good performance it is often suggested to separate transaction logs from data and data from indexes and so on, but looks like it will not be possible with amazon, as we would not know who we are sharing the disk with.
Additionally, I don't think there is any support for failover clustering with Amazon.
This question goes to those who are familiar with Amazon EC2 and ideally those who were in a similar situation of having to migrate from a stable system in a data center to Amazon.
What things did you have to consider/change to ensure no performance degradation?