This message was deleted.

Add better gear and consolidate hardware. Also depends on requirements.

1:1 server replacement is an antiquated thought process. I have replaced many clients with 20 servers, down to, say 4. It's all based on computing power.

@Ryan Gallier that was my thought. But do you consider it an ethical issue if you do a 1:1 replacement just to get added compute capability?

You could argue the other side of the ethical equation. If I'm replacing like for like, aren't I short changing the company because the other requirements have increased since the original equipment was purchased.

Wouldn't that be out of scope for a tech refresh? If other requirements have increased, then that's a new project or new requirement based on changing technology.

If you are selling them computing power they don't need, and they haven't asked for it, then yeah, that's not good. imo

We use zero based budgeting so there is no budget - it's 'what do you need' based. I can make the argument that if I want to turn off 20 old 8 core boxes, I do not need 20 new 28 core boxes

But take into account the reality of the situation, if you have 2 - 8 core servers, are you likely to get a server these days with that?

Exactly my point. I wouldn't purchase an 8 core, I'm purchasing a 28 core. But if my user requirement is for say, 100 users, and I have 20 servers now, that's 5 users per server. Do I still purchase a 1:1 server replacement with 28 core procs that can do 25 users per server? I would think I would only need 4 servers now.

But even then, the processors and memory are faster so it isn't truly like for like. It's like cars, in 3-5 years different things become standard. If you buy comparable level to me that's totally ethical even if it's 'more' than previous. Because the flip side could be true. With supply constraints now, you might not even be able to buy like for like for the same budget price. Not all up to you

Totally understandable and we expect some performance gain just because of technology improvements - SSD, NVMe, faster RAM, etc. I guess I had a hard time justifying the cost associated with saying "I need XXX to replace these servers, to keep XXX environment running" when in reality, you'll be getting X +30% extra capacity. It throws off the cost calculations for our VDI platform - I could have saved 30% had we not bought so much excess.

In my opinion the fact that you've even asked the question tells me you can sleep well at night with a clean conscience

LOL, the problem is that I have a team trying to 1:1 replace and I don't know if that's ethical from a fiduciary perspective. It might just be being stubborn!

heh, true

Fair enough, lol. That's why I asked. I didn't know if there was an industry standard of "we just replace 1:1" and I missed the boat

I'd say it's going to be hard to find a standard as it all is based off changing technology and changing market conditions.

Sure... True nerds will argue... "These 120 cores are WAY faster than the 160 cores in these old servers"

If physical servers mean nothing, then a 1:1 does NOT make sense. How do you justify replacing 320 cores with 1,120 cores?

that's what i mean.. the NUMBER of servers mean nothing

ok cool. Then it's NOT legit to just do a 1:1 - that's what I was thinking.

The way I look at it is looking at what did I buy then, and what is the comp to it today. Replace like for like. So if you bought a DL360 G8 with middle of the road proc, get a DL360 G10 with middle of the road proc. All good with that. Getting a 580 with top of the line proc, that's not comparable

That makes sense, but we have shifted our stance - instead of buying middle of the road and scaling out, we are buying a better performing host and scaling up.

Can you do an analysis to help right size what you need?

The issue isn't determining right sizing - we know we have much higher capacity on the new hosts. It was more of a 'what is proper and accepted' kind of question, I just didn't know if it was just a viewpoint, or if it was an ethical concern in trying to 'overbuy' by such a large margin.

Without analysing the current ressource usage, you can do a 1:1 core equivalence for starting , with CPU generation you will have an +30/50% perf per core . Intel 6326 (16 core) or 6342 (24 cores) have a good performance/price with a good frequency for EUC workload. So for your 320 core (20 servers/ 8core per CPU ), 8 servers with 48 core will do the job easily !