0

I am planning on running Sharepoint Foundation on one VM size A3 and SQL Server on another of size A6. As far as I understand this is not enough to achieve SLA and I should use 2 more instances - one for Sharepoint and one for SQL Server configured in 2 seperate availability sets.

Can I use scaling (by CPU usage) to turn off one instance and leave only one running at a time in an availability set? This would reduce the costs but I wonder if this solution will be good enough to achieve Azure's SLA. The way I see it one instance is running at a time while other one is shut down so I am billed for one instance. When there is an update or failure going on, the instance that until then has been running is shut down and the other one comes online. Is this the way it works? Can I cut costs of availability sets like this?

Tomasz Tuczapski
  • 319
  • 4
  • 17
  • Hi I have same scenario where I would like to have one of the VM stoped and only when a failure happens then is turned on. Did you figure out how to do this in azure? Thanks – VAAA Feb 23 '20 at 14:56
  • For this kind of scenario I would use Azure Site Recovery. – Tomasz Tuczapski Feb 25 '20 at 05:29

1 Answers1

1

no, the SLA requires two running instances. However, if you want to control your costs, the approach you have in place will work. Just keep in mind that the duration/window for a disruption will be dependent on how quickly you detect that the primary VM has failed, and how fast you can start the secondary VM. And depending on the nature of the service disruption, it may not be possible for you to start the secondary. So its a risk.

BrentDaCodeMonkey
  • 5,493
  • 20
  • 18
  • Thanks for your answer BrentDaCodeMonkey. Do you have any experience regarding Azure vms? Is it a big difference in vm availability when it is a single machine comparing to have 2 of them in an availability set? I am wondering how often my single vm could be offline. – Tomasz Tuczapski Sep 04 '14 at 10:40
  • Yes, I have some experience. :) Random outages are (unfortunately), part of the cloud in this day and age. Hopefully over time this will be reduced. In the meantime, the question you really need to be asking yourself is how much downtime can you incur (RTO/RPO), and what is the cost to you and your customers. Then use those to determine how much should be invested to help you meat your requirements with regards to availability. For some, this is thousands of $$$, for others, its only a few. – BrentDaCodeMonkey Sep 04 '14 at 12:46
  • Thanks Brent, this is all understandable but I was also wondering if there are any statistics or estimation of monthly downtime of vms in Azure (and only vms as I am particularly interested in IaaS right now) so I could decide whether it makes a significant difference to use availability sets or not. Searched the internet but no success... Maybe this information is too sensitive to make it public ;) – Tomasz Tuczapski Sep 05 '14 at 10:44
  • There are some third parties that have monitored this. The few I've looked at fall within a couple percentage points of the SLA when you exclude the rare widescale service disruptions. But its anecdotal evidence at best. The reality is that any SLA is the insurance check you really don't want to have to cash. :) – BrentDaCodeMonkey Sep 05 '14 at 12:35
  • @tomasz-tuczapski According to CloudHarmony which tracks status of services of popular cloud providers, Azure VMs had 99.9738% availability across 13 regions with 66 outages - https://cloudharmony.com/status-1year-of-compute-group-by-regions – mvark Oct 11 '15 at 07:23