Server costs?

[nathan]

Just to clarify: this is a system similar to what we'd like to purchase, we haven't got as far as actually buying it yet...

 

That's two servers and all their storage / network hardware - each server is hefty enough to run multiple VMs, and should have enough capacity to support out entire network. If one server goes down the other seamlessly takes over, and any components are replaced under warranty - if something breaks, all we have to do is wait for someone to come and fix it. I guess, in total, over 5 years we'd be spending around £14,000 a year on hardware / support costs (technically a bit more as we still have to maintain an on-site server cupboard with cooling) - that's at least half the cost of employing someone to do the same job.

 

I can, in theory, go and buy a bunch of hardware and configure it, but it takes time to get the specifications correct and test everything. In practice, a school environment doesn't lend itself to sitting and working on server configuration for a block of time, I get pulled away to do other stuff, which results in systems being put in place where we aren't confident are absolutely solid. The school's money would be better spent paying for a tested, warrantied system that doesn't need anyone on-site to maintain.

 

You're quite correct in that £20,000 for a couple of servers is a sizeable cost, and if anyone from Stone or companies with similar offerings happen to see this then it'd be interesting to hear from them. I should add that £20,000 is quite reasonable compared to the pricing we've seen for some other similar systems!

 

What size school is this for?

[mavhc]

Just to clarify: this is a system similar to what we'd like to purchase, we haven't got as far as actually buying it yet...

 

That's two servers and all their storage / network hardware - each server is hefty enough to run multiple VMs, and should have enough capacity to support out entire network. If one server goes down the other seamlessly takes over, and any components are replaced under warranty - if something breaks, all we have to do is wait for someone to come and fix it. I guess, in total, over 5 years we'd be spending around £14,000 a year on hardware / support costs (technically a bit more as we still have to maintain an on-site server cupboard with cooling) - that's at least half the cost of employing someone to do the same job.

 

I can, in theory, go and buy a bunch of hardware and configure it, but it takes time to get the specifications correct and test everything. In practice, a school environment doesn't lend itself to sitting and working on server configuration for a block of time, I get pulled away to do other stuff, which results in systems being put in place where we aren't confident are absolutely solid. The school's money would be better spent paying for a tested, warrantied system that doesn't need anyone on-site to maintain.

 

You're quite correct in that £20,000 for a couple of servers is a sizeable cost, and if anyone from Stone or companies with similar offerings happen to see this then it'd be interesting to hear from them. I should add that £20,000 is quite reasonable compared to the pricing we've seen for some other similar systems!

 

Guess it depends how much you believe the people offering support are going to spend out of the 14k you give them, if they're proactive, working with you to customise everything, giving advice etc, great. If they're just taking the money, deploying a standard script and half finishing the job, like every other support contract anyone's ever used, not great.

[mavhc]

Sounds like a good plan keeping an old server.

 

Both sites are currenly linked via switch and fibre accross from one site to the other.

 

They are currently on seperate networks and only share the Internet connection. Router has the two ports configured for each nerwork.

 

So it's really 1 network already. One server, backup server made from the other servers.

 

Get SSDs in the server though, HDDs are so last millennium.

[dhicks]

What size school is this for?

 

Currently 400 pupils and 160 assorted staff, but we expect the size of the school to grow in the next few years. We actually plan to move a lot of functionality over to cloud-based systems, but we intend to keep some local server provision for backup / failover purposes. The school currently operates over several different buildings (we're in central London, real estate is very expensive but Internet connectivity is good), the plan is to use cloud-based services for the most part and avoid the need for local server hardware where we can.

 

Guess it depends how much you believe the people offering support are going to spend out of the 14k you give them, if they're proactive, working with you to customise everything, giving advice etc, great. If they're just taking the money, deploying a standard script and half finishing the job, like every other support contract anyone's ever used, not great.

 

That's a very good point, and the part that we need to investigate thoroughly. I was impressed by Stone's quality of work and attention to detail during their installation of some touch screens over the summer, but of course server provision is a different job.

[mavhc]

Currently 400 pupils and 160 assorted staff, but we expect the size of the school to grow in the next few years. We actually plan to move a lot of functionality over to cloud-based systems, but we intend to keep some local server provision for backup / failover purposes. The school currently operates over several different buildings (we're in central London, real estate is very expensive but Internet connectivity is good), the plan is to use cloud-based services for the most part and avoid the need for local server hardware where we can.

 

 

 

That's a very good point, and the part that we need to investigate thoroughly. I was impressed by Stone's quality of work and attention to detail during their installation of some touch screens over the summer, but of course server provision is a different job.

 

Best thing is to find someone else who's using them, everyone's super responsive until you commit to giving them money for 5 years, make sure the contract specifies everything precisely, with financial penalties that they can't get out of.

 

For every 1 min of downtime not caused by a mains power outage 1% of the yearly fee will be held back. Downtime is defined as not being able to read files at more than 10MB/sec, and access databases at more than 100 queries per second.

[theeggmaster]

That's a very good point, and the part that we need to investigate thoroughly. I was impressed by Stone's quality of work and attention to detail during their installation of some touch screens over the summer, but of course server provision is a different job.

 

We had two Stone Equinox servers installed in the October half term. I'm very impressed with the quality so far, and their support seem proactive. The installation and migration went with barely a hitch. The only things we had to do port migration was upgrade the Hyper-V Integration Services from 2012 R2 to 2016 to allow Veeam to create recovery snapshots, and Impero thought it wasn't licenced, but it just needed the licence key re-entering.

[Fazza]

We get our Dell servers mainly from Server Source - sign up for their 'spam' and you'll see they have quite a few BOGOF offers on servers with the spec you are looking for.

 

Joe is the Sales guy there and does some cracking deals on whatever you are looking for. They can pretty much spec out any Dell server you like from ones costing a couple of thousand to several hundreds of thousands if you have a massive data center to set up! The emails you can sign up for show a small offering of configurations that they have put together. All come with either a 3yr or 5yr onsite warranty.

[PotNoodleTech]

I've gone full opposite strategy, as well as only buying refurbished servers, I do not pay for any warranty on the kit either. Some people may gasp at that!

 

If a hard drive dies, I swap it out myself. Same for a power supply or fan. The extended warranties are just a rip off, you will only actually get any value out of them if something MASSIVE like a motherboard dies. In the last 15 years I have had 0 motherboard failures in servers, so for me the warranty is 100% not needed. If I had a catastrophic server failure I'd ring up one of the above mentioned suppliers and get a server delivered next day for under £2k. Touch wood, so far I have never had to do that.

 

In modern times, with virtualisation (VMware or HyperV) and epic backup software (Veeam) I can restore onto a new server in about 1 hour including installing windows (and I could automate that using MDT if I got around to setting it up). The hardware effectively doesn't matter any more, you could even restore your entire network to a core i5 workstation if you were totally desperate.

 

People pay SO MUCH for the what if's and but's but they never actually use them, or get value for money out of it. Ask yourself this, how often have you actually used a warranty on every PC or server?

 

So I've got an equivalent spec (if previous generation intel) to what Dhicks is looking at and yet for my 2 x VM hosts stuffed full of RAM and intel SSDs and 1 x Veeam staging server stuffed with 4Tb drives, I spend approx. £1500 per year averaged over 5 years and that's for a school twice the size.

 

If I spent £10,000 to £20,000 per year on servers I'd get sacked

Edited December 4, 2018 by PotNoodleTech

[matt40k]

I think the key is making sure you have backups that work and money in a pot you can spend if need be, otherwise you're playing russian roulette @AButter. I kinda agree about having spares over warranty, but each on to their own.

[theeggmaster]

If I spent £10,000 to £20,000 per year on servers I'd get sacked

 

I think if anyone spent that much, they'd get sacked. Our Equinox servers are that price in total, spread over 5 years.

[Fazza]

I've gone full opposite strategy, as well as only buying refurbished servers, I do not pay for any warranty on the kit either. Some people may gasp at that!

 

Not me! The servers we get from ServerSource tend to come with a 3 or 5 year warranty as standard. When the warranty runs out we continue to use the servers and if something dies, we buy a replacement and fit it ourselves! So much cheaper than extended warranties. Quicker too in some cases as most of our suppliers can delivery stuff next day. We have backups so worst case scenario we can recover from them. All critical servers are replicated (on to un-warrantied servers) so we can fail-over to them if required.

[Tefters]

eBay is great for server spares. Running Gen7 server here for 6 1/2 years now and had 1x HDD failure, 1x PSU failure and 1x expansion NIC failure which totaled roughly £70 of eBay in total and caused 0 downtime.

Warranty is a gip and would never take it out past what gets thrown at me for free.

 

VMware is warning me everytime I bump up the version about my CPU being end of life and out of support but until it physically won't let me upgrade VMware anymore on it I will continue to run the server til it either A. Dies or B. Wont work with VMware.

 

My way of thinking and running is identical to that of butters. The amount of money saved every year from me doing minor upgrades to servers such as ram and SSD's with myself repairing them if anything goes wrong leaves enough in the pot to replace should anything go wrong in the future and have change to spare.

 

Even when replacement time eventually comes around I will simply buy a shell (when I say shell I mean CPU, rack case and mobo) and then buy/add the rest myself. It's like PC building just in a different case!

Might even build my own SAN's when that time comes too or go down the hyper converged route...

Edited December 4, 2018 by Tefters

[dhicks]

I think if anyone spent that much, they'd get sacked.

 

That's kind of the idea - that price is for hardware-and-support, at the moment I am spending a fair bit of my working time (i.e. the school's time, which they are paying me for) sorting out issues with our server setup. If I didn't have to spend time doing that I could be doing something else.

 

The main question, as raised earlier in this thread, would seem to be whether the support provided is going to be worth £10,000 a year. Interestingly, you seem to have bought just the hardware with, I imagine, initial configuration, and by the sounds of things that is working very well. If all we need to do is spend £20,000 now to get a couple of servers that simply hum away in the server cupboard for the next five years with no maintainance needed (maybe the occasional disk swap) then that sounds like a bargain.

 

I could, of course, buy some new (or quality second-hand) equipment and configure a suitible failover-capable setup myself, probably for a hardware cost of around £10,000. However, the total cost to the school is likely to be around the same by the time they've paid for my time to figure out how to set it all up and make sure I've tested every likly failure scenario (a backup isn't a backup until you've tested the restore process...).

[theeggmaster]

The main question, as raised earlier in this thread, would seem to be whether the support provided is going to be worth £10,000 a year. Interestingly, you seem to have bought just the hardware with, I imagine, initial configuration, and by the sounds of things that is working very well. If all we need to do is spend £20,000 now to get a couple of servers that simply hum away in the server cupboard for the next five years with no maintainance needed (maybe the occasional disk swap) then that sounds like a bargain.

 

I think the £10,000 per year is excessive, and maybe only worth it in the first year if the implementation was quite a big change and you needed supporting through the process to give you (and the school) peace of mind. After year 1 I'd doubt whether it was worth it.

[dhicks]

I think the £10,000 per year is excessive, and maybe only worth it in the first year if the implementation was quite a big change and you needed supporting through the process to give you (and the school) peace of mind.

 

Thanks - that does make sense, we'd be moving from a Xenserver setup, so maybe extended support at least for the first year would be a good idea.

[ITGURU]

I'm currently in similar situation replacing servers. I have a 50/50 split of HP ML350 Gen 5 and Gen6 servers. The first of the 2 oldest Gen 5's have failed in the last 6 months. Looking back these were purchased 2007/2008 so 11/12 years old :-)

 

No warning of failure, literally when the server died,a RED Led health light on the front and wouldnt boot again. Capacitors on the Motherboard had bulged and appears to be known online with the Gen 5.

 

The last servers I bought were in 2006 with their 3 year warranty. Have never extended beyond this, had the odd HDD failure but thats it, nothing else has failed.

The Gen 6's are still overspec'd for what we need and will run to their death.

I budget evey year £5k for a replacement server, but obviously depends on spec and Ram.

However it is good to shop around. I contacted HP and although i knew there was no trade in value for the GEn 5 or 6's with HP, they said if we didn't need the latest and greatest they could spec me up with a Gen 8 or GEn 9.

 

When they did, it actually would have cost me more than getting a shiny new Gen 10!

 

As they say if it aint broke, why fix it :-)

[mdrabble]

I have/had a VeryPC 1U server AMD Processor, 64GB Ram, etc- no idea how old it is (over 5 years at least and was in place when I started)

 

 

Seems to randomly crash with CPU cycle error (think that was the error as it came up briefly before going black screen - only managed to see it display the message once)

 

Since it was the RDS server, I temporarily created a VM but staff are complaint about sluggishness (well 1 CPU and 8Gb ram - what do they expect - no intention of adding more resources to it as it only temp)So I will be on the lookout for a new RDS server - @ICTDirect_Dave and anyone else who wants to send specs/prices for a suitable refurb RDS server.

[mavhc]

HP's whole "you have to pay for support to get critical fixes for bugs in the code we wrote" and "you have to buy our drives" policies are annoying me, so next servers won't be HP

[ITGURU]

The only drivers requiring PAID support are the BIOS upgrades now I believe. Whenever I've needed to download drivers or HDD Firmware, these have all been publicly available.

[Tefters]

Mav’s point is valid though, HP have got too greedy. I use to swear by them back when server were physical or even virtualisation started mainstream but nowadays I couldn’t give a flying monkeys.

 

Hardware is hardware regardless of the badge on the case, as long as it’s a octo-core Xeon processor or above and has 2 hard drive bays for raid 1 OS the rest is meh!

 

Everything is located on my SANS and all my hosts are used for are CPU and RAM boxes.

Obviously there’s the 2 PSU hotswap etc... (no SPOF) but I run a 3 host cluster that can quite easily run across 2 hosts while 1 is being fixed up.

 

Warranty and support mean nothing nowadays + the costs they ask for doing nothing is laughable.

 

If you were a very high profile business then yes go branded because they want/need those reassurances even if we all know their crap, me, I’ll be building my own new servers from shells.

Edited February 15, 2019 by Tefters

[ICTDirect_Dave]

I have/had a VeryPC 1U server AMD Processor, 64GB Ram, etc- no idea how old it is (over 5 years at least and was in place when I started)

 

 

Seems to randomly crash with CPU cycle error (think that was the error as it came up briefly before going black screen - only managed to see it display the message once)

 

Since it was the RDS server, I temporarily created a VM but staff are complaint about sluggishness (well 1 CPU and 8Gb ram - what do they expect - no intention of adding more resources to it as it only temp)So I will be on the lookout for a new RDS server - @ICTDirect_Dave and anyone else who wants to send specs/prices for a suitable refurb RDS server.

 

Morning! I had friday off so didn't get chance to reply!

 

Let me know what kinda specs you're after in this thread or by PM and I'll get a quote together, Sir.

[Fazza]

I temporarily created a VM but staff are complaint about sluggishness (well 1 CPU and 8Gb ram - what do they expect - no intention of adding more resources to it as it only temp)

 

If you change the CPU from 1 to 2 you will notice a huge difference - I know you say you dont want to, but it only takes a few moments to do and in my experience makes a huge difference in performance.

[mavhc]

https://www.google.co.uk/url?sa=t&source=web&rct=j&url=http://download.microsoft.com/download/3/D/4/3D42BDC2-6725-4B29-B75A-A5B04179958B/WindowsServer2016VirtualTech_VLBrief.pdf&ved=2ahUKEwjBzZ_dhP_eAhWOUBUIHfgEB6oQFjAMegQICBAB&usg=AOvVaw2MAuoRl1OedUkmHupjF1OX

 

 

 

 

 

(My bold)

 

FWIW I have mine set up with the core edition for both the reasons you stated and also because it was the only free option when I virtualized, but I'm aware it changed and some people prefer it because you get the GUI.

 

Edit: From the same document, I learned that you if you want to fail over to a different host, you don't need to license it on both hosts as long as you only run the load on one server at a time.

 

https://www.altaro.com/hyper-v/virtual-machine-licensing-hyper-v/ explains it quite well.

 

You need 1 Windows Server licence, with enough licences to cover all the cores, for the physical machine, but you don't need to actually install it. Standard then offers 2 licences for VMs, Datacenter unlimited, to run on that physical machine. Want to live migrate to another machine? Need 2 licences. Or you can transfer to another machine offline, every 90 days. Or instantly if the host fails.

 

So you can install Hyper V Server, or ESXi, or anything, then buy 1 Windows Server Standard, and install 2 VMs