Cost for backup to Azure or AWS

Badge +4
I'm looking into backing up our Nutanix Acropolis cluster to either Azure or AWS. What I'm curious about are the actual costs once you factor in not just their basic storage rate, but Puts and data writes and everything that happens during a backup and restore (rare as those would be).

We have around 5 TB total with all our VMs, with the daily snapshots averaging around 40GB in changed data. We would want to retain maybe a month's worth of backups with dailies and weeklies.

Is this enough data to do a decent estimate, or is it one of those 'Try and see' scenarios? I'd be interested in anybody else that's utilizaing it as a remote backup site and what kinds of costs they're seeing.


Best answer by AllBlack 2 August 2016, 00:55

View original

This topic has been closed for comments

22 replies

Badge +11
I trialled this but it was not the right solution for us, for several reasons.Initially it looked like we could make a cost saving on Commvault licensing but the benefits disappeared as our backup set grew.

Once you reach a certain amount of data, I believe 10 TB, you will need to upgrade your virtual controller instance which will double your compute cost. We also got stung by the exchange rate.Initially the Kiwi dollar was 1:1 to the US dollar but that changed to 1:1.5 which drove up the cost quite a bit. In the end it was costing us like $1,200 a month which was more expensive than our on prem solution. I cannot recall the exact retention details we used.
The performance was not ideal either as we do not have AWS datacentres in New Zealand and the closest one is over in Australia.

I think Cloud Connect is a great solution for the right use case and will definitely investigate again when it supports DR
Badge +4
Thanks, that helps. I only considered the storage costs, not compute...for some reason I thought that wasn't applicable if we're only using it for backup and restore. This is my first dive into these cloud services, guess I have some misconsceptions.

Right now we're using a combination of open source software and scripts to do backup and I'm trying to make it a little simpler. We're also considering Commvault, but we don't have any of their products in place at the moment.
Badge +11
You may want to have a look at Rubrik for your backups too. Trialling at the moment and it seems very simple to use. Another thing I am also considering is to use Nutanix storage nodes at a remote site and replicate my protection domains
Badge +4
Yeah, I saw Ribrik at the .Next conference. It looks like a very nice product, but it's way out of our price range. As is having an additional Nutanix cluster, though that would be my preference. Unfortunetely I've gotten our agency used to getting by on a shoestring budget. It was an accomplishment just getting Nutanix in the first place! I certainly sleep better though.

We're looking at about 10K in licensing for Commvault, backing up to an older NAS I'm taking out of production locally and probably some basic storage server for one of our other locations as our off-site replica. I was hoping the AWS/Azure option would be cheaper/simpler, but it sounds like that may not be the case!

I definately appreciate the feedback, thanks.
Userlevel 6
Badge +29
Yes, you do incure compute costs with cloud connect, as we run a CVM in the cloud to receive the replication data, and then pass that data on to s3/ebs (or blob in azure case).

That said, we're working on reducing that cost, by dynamically switching on an off that cloud CVM in a very future release.
Badge +4
That's good to hear, thanks Jon. Still not sure which way we'll go, hard to decide without knowing the ultimate cost! I wish they had a cost calculator or something, where you could put in your base data, change rate, restore frequency, retention period, etc and get a decent estimate.
Userlevel 6
Badge +29
The unfortunate part of cloud ... cloud can be a bit of a hotel california WRT to cost transparency.
Userlevel 1
Badge +10
AWS has a usefull cost calculator you can use:

Please note that storage is not the main costs in a small environment like yours. You need a M1.Xlarge EC2 instance, and that's more expensive than your S3 storage costs.

I just did a calculation and your S3 storage costs would be about 160 USD a month. You can save some money if you would use S3 IA (infrequent access) - this offers the same durability and availability at +/- 35% of the normal S3 pricing.

Transfers are more expensive however, so only recommended on older data. For example, you could set a policy to archive data to S3 IA after 30 days.

Be aware that EC2 pricing highly depends on the pricing model you choose
(being "on demand", 1 or 3 years "no upfront reserved", 1 or 3 years "partial upfront reserved", 1 or 3 years "All upfront reserved")

Especially 3 years upfront saves you a lot of money (again about 35% of normal pricing),
see the screenshot below:

Check out this URL if you want to set S3 lifecycle management (tiering old data to S3 IA)
Badge +4
I had just found that calculator, but I didn't know what type of instance to use. Thanks for breaking it down for me, really helpful. At this point, I'm leaning towards getting Commvault and using some older storage hardware as repositories, should minimize the cost and maximize our restore capabilities!
Badge +7
Hey guys,

Commvault guy here - let me know if you had any questions.

Quick overview based on this thread of what we can help with.

We can talk natively to Amazon/Azure and a bunch of other cloud providers for backup storage and long term retention/tape replacement. We also support things like Glacier and Azure Cool storage - no need for infrastructure (VMs/Instances) running in the cloud to get fully compressed and deduplicated copies of data copied to the cloud. We track blocks locally on-prem and so can minimize the amount of data you need to shift over the wire only to unique blocks which can help with lower bandwidth connections.

For VMware and Hyper-V workloads we can recover those to Amazon and Azure as running instances agentlessly. Either by shifting data stored locally or by using the data you sent to the cloud already (latter requires a small VM instance to process the conversion). Towards the end of the year we will be adding replication as well to assist with DR scenarios.
Badge +1
"dynamically switching on an off that cloud CVM"

Is this feature out yet?
Userlevel 6
Badge +29
jsnyder - If it was a person coming come, it would be at the front door, with the door open, and one foot in the door.

Cheesy way of saying, no, but really, really close. Double checking with PM now cc MarkNijmeijer
Badge +3
Has that person come home yet? :)

Edit: Also, are there any issues if the CVM is manually shut down and scheduled to power on either side of a backup window?

Userlevel 6
Badge +29
We unfortunately had to push this feature, to make room for some other really neat cloud connect enhancements. The feature is done, code is checked in, but disabled in AOS 5.0. We'll be QA'ing it for a release shortly after
Badge +3
Hey thanks Jon, that's great to know.
Badge +3

Can I ask is the ability to replicate to the Azure cloud with Acropolis Hypervisor now suported?
Userlevel 1
Badge +10
Azure is supported as a replication target indeed - for backup purposes that is. It's not (yet) possible to use Azure as a DR solution.
Userlevel 6
Badge +29
Yes, bertvdl is correct. You can replicate to Azure for offsite backup purposes, so that if you have a failure you can pull that data back from Azure (meaning backup to cloud, not DR to cloud).

This is supported for all Hypervisors that we support, including AHV.
Badge +3
Thanks for the resonse.
Can the data that is backup up to Azure be accessed and used without having a Nutanix devices to pull back or reopen it?

For example if the source cluster was unavailable and backup data required how does the data look in Azure?
Userlevel 6
Badge +29
Give this section in the Nutanix bible a read for some background info:

Short answer is, No, you would need a Nutanix device to read the data locally and pull it back down.

The typical "DR" example is the datacenter burns down, so you stand up a new datacenter, put some equipment back in it (in this case, Nutanix), and point it to Azure, and you'd be able to pull that data back.

A quick/rough analogy here would be if you backed up your Data to tape, and sent it off to a tape storage warehouse, and your tape drive burned down. You'd need another tape drive of similar/same specifications to read that tape again.
Userlevel 3
Badge +17
Hi paulmathewson

While this is not a qualified method, it is possible to browse the snapshots within Azure and copy them out. We've had a customer stand up a VM within Azure and whitelist that VM to have access to the clould connect instance. They were able to pull out specific VMs as required. Again, this is not a fully supported method.

Userlevel 2
Badge +3
Jon wrote:We unfortunately had to push this feature, to make room for some other really neat cloud connect enhancements. The feature is done, code is checked in, but disabled in AOS 5.0. We'll be QA'ing it for a release shortly after

Quick question on this Jon - is there also going to be support for backup to other S3 compliant (private) object stores at some point? I realise there is the added complexity of needing the extra CVM to "front" this, but (let's create a quick what-if scenario) you had a single Nutanix "backup" node running a CVM and a large lump of S3 compliant e.g. Cloudian object storage behind that? Seems to me like a low cost backup/deep archive solution?