You do need to account for the compute, storage and networking indeed.
Cost wise, if you do it right, it can be comparable to the capital purchase equivalent (over 36 months).
It does feel expensive from a glance though.
RDP sessions server
- Thread starter Ish
- Start date
You do need to account for the compute, storage and networking indeed.
Cost wise, if you do it right, it can be comparable to the capital purchase equivalent (over 36 months).
It does feel expensive from a glance though.
Soldato
- Joined
- 30 Sep 2005
- Posts
- 16,736
the capital purchase can be used onprem though, making azure even more expensive in comparison
Azure is fine as you can scale up but also scale down. We moved most of our infra to it and it's costs less. We only have a few onsite servers left for AD and updates and thats about it.
The Windows virtual desktop has worked out well for us, but of course you need to cost the VM's. For example you can run a 4 core, 16 GB machine with a 128GB for 143 a month. Compute is 126 while the managed SSD is 17, but it the compute runs from 7am to 7pm so the cost is cut in half for the compute (Works for us as the users don't do much bar emails after 7pm and they can use the portal for that. At the moment I have 3 of them running 50 users for general office and some HR software and they run fine. I have other resource pools running smaller VM's and some running largeer ones, but it depends on the load. Overall 100+ users were able to log in over the course of a week by the time I fine tuned some of the images and built new resource pools correcting some of my mistakes.
Of course we do have a large number of users simply using their work laptops and VPN's but this was stood up to cover for people who normally sat at a desktop. Over time laptops will be issued and this will be stood down which is what I love, as the cost just ends.
WVD you can also publish just apps etc which allows for you to have the apps on a "local" box so it is quicker for some users over a vpn if their internet is crap.
The Windows virtual desktop has worked out well for us, but of course you need to cost the VM's. For example you can run a 4 core, 16 GB machine with a 128GB for 143 a month. Compute is 126 while the managed SSD is 17, but it the compute runs from 7am to 7pm so the cost is cut in half for the compute (Works for us as the users don't do much bar emails after 7pm and they can use the portal for that. At the moment I have 3 of them running 50 users for general office and some HR software and they run fine. I have other resource pools running smaller VM's and some running largeer ones, but it depends on the load. Overall 100+ users were able to log in over the course of a week by the time I fine tuned some of the images and built new resource pools correcting some of my mistakes.
Of course we do have a large number of users simply using their work laptops and VPN's but this was stood up to cover for people who normally sat at a desktop. Over time laptops will be issued and this will be stood down which is what I love, as the cost just ends.
WVD you can also publish just apps etc which allows for you to have the apps on a "local" box so it is quicker for some users over a vpn if their internet is crap.
We found with various project in the pipe line that if we invest in onsite infra and storage half of would end up unused in 2 years, at least with Azure we can down scale.
Soldato
- Joined
- 30 Sep 2005
- Posts
- 16,736
ah, I guess all depends on long term plans. If you are building something only to last a couple of years, then I can see why Azure makes sense.
Soldato
- Joined
- 30 Sep 2005
- Posts
- 16,736
I guess the only niggle would be if the company internet line fails, or it's not fast enough.
Same as any office or workspace though isn't it, not everyone has their infrastructure on premise or in Azure, it could be CoLo too and suffer from the same problems. This is where scoping and redundancy comes into play, how much does 1 hour of downtime cost is the sort of question I pose to management. Azure isn't the answer for everyone, but it's also a great option. I did a contract migrating from CoLo services to Azure for their main LOB Apps and Services and this saved the company £700k a year, we used Azure really well by having things automatically shut down and start up when they weren't required, and fortunately the company had their own licensing which could be reused for a lot more savings.
We're deploying more WVD at the moment, but also still doing a lot of on-premise or CoLo deployments, it's a great time to have so many fantastic options available.
For us it's being used by people at home so it depends on their internet only as they connect directly to MS. I did notice a few users on 1mbps and at times tell me it's slow and delayed typing etc but it's always the same 3 users all on 1mbps as they live in the middle of no where. The rest never complain.
We do have our azure talking to onsite as well via an express route and vpn for backups.
Associate
- Joined
- 14 Dec 2004
- Posts
- 1,205
I don’t get how people save money going to azure. On prem stuff must be badly managed somehow.
We have 7 esx servers @ 25k each running 130 server VMs ranging from 8gb RAM 4 cores up to 170gb RAM and 32 cores
Hardware life is 5 years so each VM is costing us £22 per month on average
A quick look at the azure calculator and it would literally cost 10 times the cost of on prem. I don’t get it
If you have a bunch of PCs on site, maybe splashtop or the likes would be cheaper and as handy.
I manage several companies infrastructure and depending on budgets and how quickly WFH was demanded they got different solutions.
From RDP or remote apps to splashtop and teamviewer, as long as the guys can get their work done, they've alm been happy with what they have. Internet connection to site is by far the biggest bottleneck.
I manage several companies infrastructure and depending on budgets and how quickly WFH was demanded they got different solutions.
From RDP or remote apps to splashtop and teamviewer, as long as the guys can get their work done, they've alm been happy with what they have. Internet connection to site is by far the biggest bottleneck.
It's really easy to save money on Azure with caveats not all workloads are necessarily suitable for Azure in terms of cost. You need to look at the following:
- Is your workload static, or does it need to scale up and down
- Are you scoping Azure workloads based on direct replication of your compute into Azure
- Are you costing your comparison correctly - you mention 7 ESXi Hosts at 25k each, that's just hosts though, are you calculating your switches, your SAN(s) and your SAN Switches into that, and then your backup targets, as well as the power costs, the licensing costs (this one is key) and the engineering costs for maintaining these and updating them usually it's out of hours for the likes of SANs and SAN switches which is likely to be a more expensive engineering time
I much prefer the traditional infrastructure market, I like being in control of my Private Cloud and my DataCentres, but Azure and any other Public Cloud offering is fantastic way to save money, dependent on workloads. I saved 700k a year moving LOB, SQL, ADFS->AAD, SharePoint etc over to Azure and SaaS all whilst keeping a vSAN Stretch Cluster across two physical sites for our VDI deployment, we costed up all this and it made zero sense for anything but the VDI solution, some domain controllers, file servers and a couple of print servers to stay on premise.
A well optimised Azure environment can absolutely be comparable to a traditional deployment; go with a decent reseller, get 3 year reserved instances, automatically de-alocate VMs when not in use, right-size your VMs, utilise Hybrid-benefits, etc etc.
The problem with Azure is most dive head first into it without realising how to properly configure the environment, and get bill shock as a result. Key to any Azure deployment is to realise some workloads shouldn't touch it, never go full cloud, and do your research!
I know that, the savings come from using the advantages that the platform gives you - whether that's scaling down when business is quiet, moving databases to their cloud instances etc. If your planning goes as far as doing a lift-and-shift though then you'll find it looks expensive. So in this virtual desktop example if you do your price comparison based on zero seasonal changes in headcount and a static VM size and quantity for three years with hosts running 24x7 then it's understandable that the value proposition doesn't make much sense, but that's because you're approaching it wrong.
We have customers running in Azure now that scaled back at the start of Covid19 and are paying about a quarter of their regular rate to keep a few services ticking over for a reduced workforce to access - anybody on-prem is still depreciating their hardware at the same rate they were before, their support contracts will still need renewing at the same time they did before etc.
We have customers running in Azure now that scaled back at the start of Covid19 and are paying about a quarter of their regular rate to keep a few services ticking over for a reduced workforce to access - anybody on-prem is still depreciating their hardware at the same rate they were before, their support contracts will still need renewing at the same time they did before etc.