1

We are currently preparing for a move of a reasonably high traffic web site to the cloud.

We are thinking of using scalr to help us manage the whole setup especially since we dont have experience with amazon.

We are unsure about whether we should use Scalr's MySQL functionality that relies on EBS backed EC2 instances or whether we should be making use of RDS or even xeround and enjoy much easier maintenance and management.

Our dataset is about 40GB and we consume a bandwidth of 4000 GB per month between the application server and the database server.

any experiences on similar setups?

thanks in advance

Michel Galibert
  • 57
  • 5
webgr
  • 203
  • 1
  • 2
  • 7

2 Answers2

2

I can tell you from my experiences with a large db.. but much larger then yours around 90gigs.

We went with RDS and at least 3-4 times a day we would get horrible latency on queries. Like queries that would take a second would go on for 10-20 seconds. We moved to their largest instance backed with raided EBS and performance was about the same but we didn't hit those really bad latency spikes.

Mike
  • 21,910
  • 7
  • 55
  • 79
-2

Going to the Cloud is indeed a very good option - The hardest part in scaling your cloud application is scaling the database. Don't get confused, MySQL has failover solutions (how long it takes…), that can support multiple replicas to handle reads. Scalr and RDS are very relevant options if you know what the limitations are… With Scalr – they will scale your DB by provisioning database slaves, masters, and configuring data replication. While auto-provisioning does ease and replication does provide for some remedy, keep in mind that adding read replications won't fix the OLTP writes…, nor will it handle true on-demand linear elasticity. Each time you add a read replica it will likely be a service event.
For HA, Scalr uses EBS. If the last AWS EBS very long down time is of any indication.., make sure your data/storage is also HA…

Scalable solution needs to be linear and elastic (scale out, in, up, down) on demand without any downtime. Cloud applications need native true HA – multiple replicas performing read and write all the time, not failover. RDS will prove the same "pre-configured MySQL" setup and thus have the same limitations. Last but not least, make sure there's no single point of failure and what the impact on your developers is… every time you make a change to the application. At Xeround, our design goal was to address these issues. Our Telco grade genes and solution developed from day one as virtual for a cloud operations enable us to offer an on-demand, elastic, super scalable and highly available "worry free" cloud DB.
We have run a yearlong beta with thousands of users many of which exactly like you.
We're now GA - please sign in for a 30 free trial (no credit card required) and check this out for yourself.

Razi Sharir (Xeround).

Razi Sharir
  • 1