Extremely slow restore speeds?

[t7MevELx0]

Has anyone else experienced extremely slow restore speeds with version 8.1?

We've noticed that even when restoring locally, the transfer speed spikes to 30–150 Mbps, then drops to 0, repeating this cycle. Restoring 150 GB of data takes about 3 hours — that’s an average of 14.22 MB/s (or 113.76 Mbps).

To troubleshoot, we’ve tested restores using the following methods:

1. From a local system using Veeam Explorers

2. Through our custom restore portal (API-initiated)

3. Using the built-in restore portal provided with the Veeam software

All methods show the same inconsistent and slow performance.

[Polina]

Hi Cody,

What restore speeds would you expect based on your usage experience before 8.1?

[t7MevELx0]

Hi Polina,

To properly test what we were seeing before version 8.1, I’d need to reinstall 7.1. That said, given the changes Microsoft has made to their service, many older builds of your product no longer function reliably, so there’s a good chance it wouldn't work at all.

However, based on what we now know about the hardware and current config, we can confidently estimate expected performance.

Hitachi Vantara Object Storage; multiple storage nodes.

Per Node
Sequential Read: ~3.6–4.5 GB/s
Sequential Write: ~3.0–3.6 GB/s

Across All 3 Nodes:
Operation Aggregate Estimate
Sequential Read ~10.8 – 13.5 GB/s
Sequential Write ~9.0 – 10.8 GB/s

Given this setup, we’d expect ~11–13.5 GB/s read throughput and ~9–11 GB/s write throughput under optimal conditions. So seeing only ~113 Mbps during restores is far below what the hardware is capable of delivering.

The storage vendor has already reviewed the health of the system and they've noted no issues.

[Polina]

There's also a networking impact and M365 itself...
If you are convinced that speeds have decreased compared to your previous experience, please open a case with our support to troubleshoot it. Without inspecting logs, it's not possible to identify the root cause.

[t7MevELx0]

I have opened a case with support. I've attached logs. My copy jobs haven't been working since Jan 28th and I'm still waiting for those to get fixed. Could the extremely poor performance be related?

[Polina]

I doubt it, but I'm not a support engineer, either.

Support cases should be opened per a single specific issue, so if you would like to troubleshoot restore performance, please open a new support case for it and provide corresponding logs.

Thanks!

[t7MevELx0]

I updated our support case. We were able to test directly downloading from the storage without Veeam and that download was 1.2 Gbps sustained. I've given support the logs.

[edh]

We got the same sense, a full NVME cluster just for Veeam Office 365 as repo and download speed are toooo slow... we spent hours downloading like 20 psts...

[Polina]

Manuel,

Download (from M365) or upload (to M365)?

[edh]

Creating a PST from our S3 NVME storage... Download from Microsoft we're able to reach 10-15Gbps from all tenants.

[pat_ren]

Hi Polina,

To properly test what we were seeing before version 8.1, I’d need to reinstall 7.1. That said, given the changes Microsoft has made to their service, many older builds of your product no longer function reliably, so there’s a good chance it wouldn't work at all.

However, based on what we now know about the hardware and current config, we can confidently estimate expected performance.

Hitachi Vantara Object Storage; multiple storage nodes.

Per Node
Sequential Read: ~3.6–4.5 GB/s
Sequential Write: ~3.0–3.6 GB/s

Across All 3 Nodes:
Operation Aggregate Estimate
Sequential Read ~10.8 – 13.5 GB/s
Sequential Write ~9.0 – 10.8 GB/s

Given this setup, we’d expect ~11–13.5 GB/s read throughput and ~9–11 GB/s write throughput under optimal conditions. So seeing only ~113 Mbps during restores is far below what the hardware is capable of delivering.

The storage vendor has already reviewed the health of the system and they've noted no issues.

are you testing save as file eg. pst or restore direct to a mailbox? save as file performance for me seems normal, but restore to 365 takes an extremely long time, often it's throttled my microsoft

[t7MevELx0]

This is testing by saving to a PST, of data that we already have from Microsoft via the restore portal, or the explorers. We've proven the array is substantially faster than this. I can copy data directly off of the array much faster. I'll just hope this gets fixed soon.

[t7MevELx0]

This issue is still present in version 8.1.1.

I can reproduce the problem regardless of the storage backend. I've tested with both Wasabi and our local Hitachi Vantara system. During Veeam restores, data transfer starts, then pauses intermittently—starting and stopping repeatedly. However, downloading the same data directly from the storage system is consistently fast.