Business Class Server Hardware

[UNSC Trek]

Good day to you all.

I work for a growing IT Company and am experiencing a few things I'd love to have opinions on as well as possible links to modern standardized SOPs that I can show my higher-ups, so we can implement proper processes when it comes to hardware.

The majority of the customers we service are currently on either Server 2016, or 2019, so software isn't the issue. They also have dual xenon CPUs and 32+GB of ram for only a few VMs. However, there are some customers that we have who experience server reboots of nearly 120 minutes.

I've been told that this is "typical" for business class servers.

I have a personal server at home, dual xenons, 120GB of DDR3 ram, and mostly SSD storage, especially for boot media. I never experience more than a 5 min boot time, and that's with running updates.

* Could anyone suggest links to current Hardware recommendations for servers (Domain Controllers, Terminal Servers, file hosting servers, etc.)? I want to bring my company into the 21st century, but it's almost as if we're locked in the 20-year-old past.

[Needfuldoer]

What applications are you hosting?

 

What hardware are they running on?

[Electronics Wizardy]

Do you have vms here? 

 

If you have vms and multiple hosts, you can move a vm to another host(without stopping the vm, normally in the 10s of ms of downtime). And then reboot the hypervisor, then move the vm back.

 

The modern way to do this is not host your own servers and do it all on the cloud.

 

What is the reboot time doing? Is the os in control? Or is the os updating/saving something? Windows 2016 seems to love taking a while to update, I think the small buiness with 2016 I manage takes >1 hour to reboot and install updates normally.

[UNSC Trek]

@Needfuldoermost apps are things like QuickBooks and other accounting software, or databases of various kinds. Some are fileservers/print servers. Most are Terminal or Application servers.

@Electronics WizardyThey are all mostly VMs run on VM Host systems, but the Host systems are all Spinning HDDs, no SSDs. And I have no idea why anyone in their right mind would run a VM C drive off spinning disks in RAID, especially if it's for a Terminal Server for 15+ users, when they need quick access to applications.

It's truly baffling me.

I would assume by now, that only storage would be on Spinners, and boot media and software media would be on SSDs, but I can't get anyone where I work to see the logic in that, so a lot of our customers call in asking us to speed up their servers. (I work in Remote Support) And in the end, all we can do, is tell our customers... "Sorry, this is kinda the way your hardware was built" making my entire organization look like we're stuck in the late 90s.

[LogicalDrm]

-> Moved to Servers and NAS

[Electronics Wizardy]

2 minutes ago, UNSC Trek said:

but the Host systems are all Spinning HDDs

Welp thats your problem.

 

3 minutes ago, UNSC Trek said:

I would assume by now, that only storage would be on Spinners, and boot media and software media would be on SSDs, but I can't get anyone where I work to see the logic in that, so a lot of our customers call in asking us to speed up their servers. (I work in Remote Support) And in the end, all we can do, is tell our customers... "Sorry, this is kinda the way your hardware was built" making my entire organization look like we're stuck in the late 90s.

I can't convince management for you, so I don't think there is anything I can do to help.

 

 

But I worry about a company that hasn't switch to SSDs in 2022, is the rest of the company competent? 

[UNSC Trek]

14 minutes ago, Electronics Wizardy said:

Welp thats your problem.

 

I can't convince management for you, so I don't think there is anything I can do to help.

 

 

But I worry about a company that hasn't switch to SSDs in 2022, is the rest of the company competent? 

That's exactly my worry.

 

I guess my biggest request is this:

 

Does anyone know of any online sources that promote current, 2022-2024 standardizations that I can bring to management's attention that promote SSDs over Spinners for new and existing server builds? Like actual documentation, online resources, not just opinions.

[wseaton]

Promoting SSD's over spinners is like promoting 1Gb switches over 10Mb hubs. There's no promotion here. SSD are mandatory in any server or desktop for the operating system and main application drive. SSD's are not mandatory for bulk storage like big filer services or data stores. The reason being is the network is often the bottleneck. With 500GB data center grade SSD's costing well south of $500 that's plenty of space for the OS, lots of Terminal Server profiles, and healthy database app or two.

 

Terminal  / RDS / Citrix servers benefit the most from SSD. They are basically just desktops people remote into and are very, very sensitive to local storage latency (spinners are horrid in this dept).

 

There is also no definitive guide or class for servers. Pretty much any HP or Dell or Lenovo Server is "business class".  Even the cheaper entry level ones with single PSU last forever in my experience. What you typically pay more for is very large memory expansion options and then storage. The CPU is becoming increasingly the least expensive component.

 

However, there are some customers that we have who experience server reboots of nearly 120 minutes.

 

Bare metal Servers take longer to boot because they have to chew through iDRAC / iLO checks, then the RAID card spin up, and usually a more exhaustive hardware checksum by default. This at most adds a minute or two to boot up. 120 minutes indicates a severe hardware problem, or a Domain Controller with some serious configuration or service dependency issues. Something is broke on those servers and you need to look at it ASAP.

 

I've been told that this is "typical" for business class servers.

 

Dual socket servers have been in decline for about 10 years. Single CPU's are getting stupid fast with high core counts, and they are significantly cheaper than dual or especially quads, which are now becoming reserved for large data centers that need very large memory pools. Most smaller businesses can get away with the smallest single socket server offered. 

 

It sounds like you work for an MSP with clients paying you money to support them, right? So, the job is not for you to convince your management, but the customers paying you for support. That's a salesman's job. MSPs often don't have the best rep because they charge serious margins for hardware. Most customers are going to balk at $6,000 for a basic domain controller and small terminal server when a small entry level server at 1/5th the price is more than adequate for the job. The cloud then looks plenty tempting.

[UNSC Trek]

12 hours ago, wseaton said:

It sounds like you work for an MSP with clients paying you money to support them, right? So, the job is not for you to convince your management, but the customers paying you for support. That's a salesman's job. MSPs often don't have the best rep because they charge serious margins for hardware. Most customers are going to balk at $6,000 for a basic domain controller and small terminal server when a small entry level server at 1/5th the price is more than adequate for the job. The cloud then looks plenty tempting.

You're correct. However, it's not $6,000 for a basic domain server. Its $45,000. And the Customers are never properly informed, and the Sales team is not tech savvy at all, as they just sold 5 firewalls to a company that has one, and only 2 inbound internet connections. So it's a matter of true competence.
I appreciate the responses though.

 

Everyone here is rather awesome. (Wish I could hire from here for our company, enthusiasts know far more about hardware than our current sales team.)