Today, we’ll cover the top questions about the first, full Office 365 backup using BackupAssist 365.
- How long will the first full backup take?
- I have large files, lots of files, many users, email hoarders: what kinds of data will really slow things down?
- Will my tenant get throttled during a backup, slowing down everything else?
- Should I only run backups overnight? How?
- Will my first full backup capture all my data?
- What if the full backup gets interrupted?
- Does my Internet connection’s speed make a difference?
- Office 365 Permissions & security – best practices, do I need an extra user license, multi-factor authentication
- After my full backup is complete, will subsequent backups be incrementals?
1. How long will the first full backup take?
This is highly dependent on the amount of data in Office 365. But we have solid figures to help you estimate how things will go for you.
Let’s assume you have:
- a regular business-grade Internet connection (50 to 100 Mbps), and
- a “typical” mix of data sets (files & mailboxes)
Here are some estimates based on our experiments, and you can see where your tenant fits.
|Typical duration||Amount of data|
|1 hour||1.3 GB, 14,000 items|
|Overnight – 14 hours||17.6 GB, 200,000 items|
|One weekend – 60 hours||75 GB, 850,000 items|
For files (SharePoint Online or OneDrive for Business):
|Typical duration||Amount of data|
|1 hour||25 GB, 500 files|
|Overnight – 14 hours||360 GB, 7000 files|
|One weekend – 60 hours||1.5 TB, 30,000 files|
In Q7, we’ll discuss how different Internet speeds can influence the download time.
If you’re interested in our raw results, here they are:
|80Mbps fiber||12.44||268,550||14h 24m 01s||48.6||14.7||310.8|
|4G||12.15||275,004||14h 9m 11s||46.3||14.7||323.9|
|80Mbps fiber||35.67||405,471||1d 5h 32m 07s||92.2||20.6||228.8|
|80Mbps fiber||35.14||402,216||1d 5h 26m 17s||91.6||20.4||227.8|
|4G||41.73||471,903||1d 11h 25m 37s||92.7||21.3||235.4|
|100Mbps fiber||76.28||890,079||2d 12h 6m 13s||89.9||21.7||246.8|
|90Mbps fiber||19.15||485||49m 42s||40.4||400.2||9.9|
|80Mbps fiber||123.95||70,391||1d 15h 20m 36s||1.8||53.8||29.8|
|80Mbps fiber||154.65||70,398||1d 15h 43m 15s||2.2||66.5||29.5|
|100Mbps fiber||184.27||4,337||7h 7m 9s||43.5||441.9||10.2|
|80Mbps fiber||241.98||4,219||17h 32m 0s||58.7||235.5||4.0|
|90Mbps fiber||250.75||4,274||10h 18m 26s||60.1||415.5||6.9|
|80Mbps fiber||254.33||131,989||3d 1h 29m 36s||2.0||59.1||29.9|
|100Mbps fiber||369.57||4,951||13h 34m 55s||76.4||464.9||6.1|
2. Large files, lots of files, many users, email hoarders: what kinds of data will really slow things down?
Obviously, the more data you have, the longer it will take to back up. However, there are a number of things that will make a full backup take substantially longer.
- Lots of calendar items. Extracting calendar items is very slow, and on top of that, Exchange only allows us to download around 500-1000 calendar items per backup run. So while your first full backup will contain all your emails, it’s likely you won’t get all your calendar items on your first run if you have a lot of them! We once witnessed a salesman’s inbox with over 10,000 appointments – so the first full backup got about 500 items, and the remainder were downloaded over the next two weeks of incremental backups.
- Lots of folders. Each folder takes time to scan, so many folders increases this scanning time.
In OneDrive for Business or SharePoint:
- A lot of small files – each file takes on average between 0.5s and 1.0s to download, even if they contain just 1 byte of data. So if you have one million such files, that will take 5 to 11 days to download a million files, even though the data size is just 1 MB.
Apart from these anomalies, the duration of your first full backup will vary quite linearly with the volume of data.
What matters less is the number of users. For example, if you take one big mailbox and spread it over 10 smaller mailboxes, it won’t take much longer to back up.
3. Will my tenant get throttled during a backup, slowing down everything else?
Microsoft’s throttling algorithm isn’t publicly documented, but we’ve been running experiments for years, so we’re familiar with how it works.
Short answer: if you follow our best practice guide for creating a backup user, then the answer is “no”.
We’ve noticed that throttling is based on user, and to a lesser extent, the server. When you create a separate backup user login into Office 365, it means that usage is recorded against that backup user, and not an actual user. Therefore, slow-downs (discussed in Q4) are limited to the backup user.
Sometimes, a server is slow because it’s highly loaded across many tenants. In those situations, things may run slower for everyone… but that is a quality of service issue from the cloud provider, not something caused by BackupAssist 365.
We also provide the ability to define a backup window and run the backups overnight. See the next question for more details.
4. Should I only run full backups overnight? How?
According to Microsoft’s best practice and guidance, full backups are ideally run outside business hours.
Here are two ideas to do this:
- Start your first full backup on a Friday evening – for example, starting 6pm on Friday. Then let it run over the weekend – you’ll have over 60 hours before work starts on Monday. Typically, you can download:
- Mailboxes: 75GB of data, in 850,000 items
- SharePoint / OneDrive for Business: 1.5TB of data, in 30,000 files
- Run your backups overnight – say start the backups at 6:30pm, and stop them by 8:30am, giving you 14 hours. Typically, that’ll give you:
- Mailboxes: 17.5GB of data, in 200,000 items
- SharePoint / OneDrive for Business: 350GB of data in 7,000 files
We recommend for most businesses that you can create a backup task and run it straight away to make sure all the settings are right. But then once you know everything connects up just fine, you can stop the backup and schedule it to run later.
If you run a full backup in the middle of the work day, there is a chance you’ll run into throttling. You may see messages like “The Office 365 servers are overloaded” in the backup progress window – that happens when BackupAssist 365 gets a “Server busy” response when downloading data. We explore that more in the next question…
5. Will my first full backup capture all my data?
As noted in Q2, mailboxes with many calendar items (more than 1,000) are likely to require several backups to download them all.
If you have a lot of data, it may also require several runs to complete.
Aside from these cases, if you run your backups outside business hours, more often than not, your full backup will be complete.
The exceptions where Microsoft’s cloud service behaves unpredictably:
- Throttling – sometimes Microsoft says “you’ve downloaded enough, I’m not giving you any more for now”.
- Microsoft internal maintenance – sometimes Microsoft denies access to its service – for example, it might be migrating data internally from one server to another.
In these cases, you might see messages like these if you go digging into the the backup logs:
- The specified object was not found in the store., The process failed to get the correct properties.
- The mailbox database is temporarily unavailable., Cannot query rows in a table.
- The mailbox database is temporarily unavailable., Cannot open mailbox.
- Access is denied. Check credentials and try again., The process failed to get the correct properties.
- Mailbox does not exist
BackupAssist 365 translates these messages into friendly messages in the backup report:
One or more mailboxes are throttled and skipped on this backup. The backup will continue on the next run.
John.Citizen@domain.com: Unable to connect to mailbox. There may be a permissions problem, or the mailbox is throttled.
As the message suggests, it means that it was not possible to back up that data on this backup run, but the next run will try again.
6. What if the full backup gets interrupted?
BackupAssist 365 has very sophisticated retry logic that will automatically deal with temporary service outages or network issues. For example, if there’s a brief Internet outage, or if an Exchange server is busy, we will wait a few minutes and try again.
Most of the time, you wouldn’t even know that there were brief outages (which are very common).
However, sometimes the full backup can be interrupted in a non-recoverable way. As Murphy’s Law would dictate:
- Windows Update reboots your machine while a backup is running
- The Internet connection drops out for hours
- Microsoft might deny access while migrating data or reprovisioning their service internally.
If this happens, it’s no problem at all.
The next time the backup runs, the remaining data will be downloaded and backed up.
7. Does my Internet connection speed make a difference?
Yes, but not as much as you might think. Here’s a summary:
Mailboxes – generally there’s little difference in performance:
- Mailboxes generally have many small items of 50kB to 100kB in size.
- The speed of the Exchange server is the bottleneck here.
- Surprisingly, no difference between a 4G connection at 20Mbps vs. a fiber connection at 100Mbps. 4G was less than 2% slower.
For files – there can be differences in performance, depending on the average file size:
- Small files – very little difference in performance, as latency is the bottleneck.
- Large files – a faster connection will mean large files are downloaded faster as bandwidth is the bottleneck.
8. Office 365 Permissions & security – best practices, do I need an extra user license, multi-factor authentication
When you’re setting up a backup, we recommend that you create an unlicensed admin user login to act as your “backup user”.
This enables you to keep your automated backup login separate from regular user logins. This gives you particular benefits:
- Multi-factor authentication can be used to secure the automated backup user login
- If the tenant administrator changes his or her password, it won’t affect the backups
- No extra licenses are required
We recommend following our best practices guide.
9. After my full backup is complete, will subsequent backups be incrementals?
Conceptually, yes. The differences, or “deltas”, will be backed up. So many users think of the backups as incrementals.
Technically, the nightly backups are implemented not as incrementals, but as differentials.
BackupAssist 365 will always back up the differences between the cloud and what’s in your backup. This is a very reliable method that caters for:
- Unexpected interruptions in any previous backup
- Backups that have not been run for days, weeks or months
- Rolling back your backup data – like restoring your machine from an earlier backup