It was a bit quiet here due to the current COVID 19 pandemic. But now I’m back with a pretty interesting story on how my colleagues and I moved most of our on-prem server stuff to Microsoft Azure and Office 365.
It all started with the COVID19 lockdown in Germany in March 2020. We moved into our home offices after setting up a small VMware Horizon View deployment to access our PCs using physical View Agents and manual desktop pools. Most projects were stopped, and we did most of our work remote. No lay-offs or short-time work.
We were running a small VMware vSphere cluster for a couple of years. Nothing fancy: Two HPE ProLiants, vCenter, two DCs, File-/ Printserver, WSUS, Exchange, Linux maschines for web services, Sophos UTM, a pfSense, View Connection Server, UAG, ADFS/ ADFS Proxy, PKI etc. In sum 18 VMs on two hosts, some VLANs with firewalls in between etc. We were running Exchange 2016, AzureAD Sync, Exchange Hybrid, but we only used Microsoft Teams from our Office 365 deployment. Veeam Backup & Replication was used for backups, a backup copy to a NAS and some Robocopy jobs that moved Veeam Backups to USB drives for DR. Everything was pretty simple and designed to work without much operations. Our focus is on our customers, not on our internal IT. It was stable, secure and pretty slick.
In March 2020 we asked “What if?”. What if we lose our offices due to a fire (we are located in a bigger office building and we had a couple of fire alarms this year due to remodeling work). How can we work if your DSL line is cut? How can we get our backups offsite? How can we modernize our IT withtout big invests? Money, that we don’t spend on our internal IT can given to our employees. ;) (By the way, that’s the same reason why we try to drive smaller and more efficient cars…).
We developed a couple of ideas, including new servers, storage etc. and put that stuff into a datacenter. But in the end, we decided to move most of our stuff to Microsoft Azure and Office 365.
I want to share some of the things we have learned on the road to Azure.
We used the Azure Migrate Server Assessment tool to assess our vSphere environment. We wanted to get a ball park on how we had to size the VMs. We knew, that we wont need to migrate all VMs. For example our virtual pfSense firewall, the vCenter, the Sophos UTM or our ADFS setup were not planned to migrate.
After the first assessment, we started to play around the the Azure pricing calculator. Just to get an idea on how different VM sizes affect the costs.
As a Microsoft partner, we were able to use our internal user rights (IUR) for Microsoft Office 365 and Azure. Microsoft offers us 25 Office 365 E3 plans and a 6000 US-$ budget for Azure (Azure Sponsorship Subscription). Our plan was to stretch the Azure budget over 12 months, so that we don’t have additional costs until we re-apply for our Microsoft partnership. Starting with 6000 US-$ Azure budget, it makes ~16 US-$ per day for our complete Azure deployment.
Now, as we knew that we have 16 US-$ per day, we planned our Azure deployment. First of all, we planned the number of VMs. We had 18 VMs on-prem, and we managed to get down to 9 VMs.
- two Domain Controller
- Management Server
- Remote Desktop Host (all-in-one Deployment)
- Ticket System
- SQL/ App server
- Exchange 2016 for Hybrid Deployment
The View Connection Servers and UAG are still running on-prem. Our virtual pfSense will be moved to a WatchGuard Firebox soon. Sophos UTM and ADFS are gone. A dedicated WSUS server is not necessary any more, we moved back to simple Windows Update and Delivery Optimization.
To connect to our Azure deployment, we had to setup Site-2-Site VPN. We deployed a simple Gen1 Basic SKU VPN Gateway. We had no need for more than 100 Mbit (we’re using a 50/10 VDSL at our office location), BGP or zone redundancy.
Instead of Standard SSD or Premium SSD storage, most of our VMs are using simple Standard HDD storage.
Backups are kept in a Recovery Services Vault with pretty simple polices. Either a VM needs to be current, in this case we keep 7 restore points, or we might need to keep more restore points. In this case we keep 7 daily, 5 weekly, 12 monthly and 3 yearly restore points. And this is only the case for our fileserver.
Additional cost savings
But with this setup we would not get under 16 US-$ a day. :( So we took another approach to break the mark: We shut down VMs at night and at the weekends! It took a bit until my colleagues and I get used to this. Nobody wants to shut down servers without a good reason.
But: We are currently at 18 US-$ per workday, and 10 US-$ for saturday and sunday. Everything, except domain controllers and ticket system, is shutdown at night and on the weekends.
We are using an Automation Account with some simple scripts and schedules to shutdown VMs and start them again.
The next blog post will be around how we planned the usage of Office 365, and how we started with Azure.
- Configure VMware Horizon View client device certificate authentication - March 11, 2021
- VMware vCenter 7.0 U2 deployment fails at stage 2 - March 10, 2021
- Veeam B&R backup failes with “No scale-out repository extents are available” - February 24, 2021