Public Cloud vs. Hosted Hyperconvergence Costing Project
-
@KOOLER said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
I don't get the "hosted colocation" idea. If I want to move move workload out of my own datacenter, why should I trust to somebody small running SMB solution on steroids, if I can go for AWS or Azure? Are we comparing costs only?
Well for AWS you don't get HA without extreme costs and there is so much "technical know how" needed on the customer end that many simply can't do it. Azure, you can't seriously compare to enterprise colocation; Azure outages are so common and extreme that if you have concerns about reliability you can't even have them in the conversation. AWS, Softlayer and such, yes, really great reliability if you know what you are doing. But that's just one factor.
Enterprise colocation is rarely someone small. There is a reason that every Wall Street and Canary Wharf firm uses that approach. Enterprise colocation have uptimes exactly the same as AWS, or better because they can't rely on site failover to cover for a site outage in the same way.
AWS and Enterprise Colocation are very much toe to toe competitors here. They are a bit different in what they do and how they do it and where the cost and the expertise has to exist.
Also, for tons of people AWS (cloud at all) is not an option. Colocation is actually the majority appropriate case for the SMB and SME markets.
-
@Dashrender said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
Why do you consider putting a HC into a colocation a somebody small running SMB solution on steroids? A level IV DC is equivalent to what AWS is running (I'm assuming), so the DC itself is the same.
Correct. Both are Tier IV. I can tell you from working with Tier IVs on Wall St., they are just like Amazon's datacenters.
-
Actually, you can do a 2-node cluster with Windows Server 2016 and S2D to get HA and redundancy.... node and disk redundancy. You can even get 2-disk redundancy using erasure, which gives pretty good space efficiency. I think upwards of about 75% on the high-end with all SSDs. Of course mirror or even 3-way mirror is slightly better performance than erasure, but it still beats RAID 5 or 6.
I took your life-like specs, and priced out an all-SSD R730xd server to fit the needs. It's a bit more capacity than required, but I priced it as if RAID10 will be used for the SSDs, as I had no idea about the server load, so I did that just to cover high-IOPS intensive loads. So you could subtract 2-4 grand if RAID5 will be used, depending on how much extra capacity you need.
It's about $13,000 per server, without Server 2016 Datacenter (assuming extra capacity for RAID10 needed). That's straight consumer pricing from their website. They give DC for about $4,000 through Dell. No SA, but still.
So you're looking at $17,000 for a 2-node HA cluster running 2016 Datacenter, using S2D.
Server specs (each):
- 24x 2.5-inch bay + 2x 2.5-inch flex-bay chassis
- 2x Intel
Xeon E5-2630 v3 2.4GHz,20M Cache,8.00GT/s QPI,Turbo,HT,8C/16T (85W) - 128 GB RAM
- 2x 120gb flex-bay SSDs
- 8x 1.6TB SSDs (about 13 TB raw capacity)
- Redundant 750w PSUs
- PERC H330 (pass-through will be used)
- Rack rails
- iDRAC enterprise
- Network: Intel X710 DP 10Gb DA/SFP+, + I350 DP 1Gb Ethernet, Network Daughter Card
Two of those will get you your HA and node redundancy, plus all the awesome benefits of S2D!
I didn't do the math per-VM and per-VM-specifications... but this setup would be great.
This setup seems like direct competition to Starwind vSAN, except you get the benefit of full Windows 2016 Datacenter as well. I guess Starwind could be better financially if Datacenter and the features of a full Server OS isn't needed. Then you could do this using Hyper-V Server and Starwind. I don't know the pricing of Starwind off-hand, though.
This would be an awesome "Cost + Benefit Comparison" for you do to SAM
The new Server 2016 Datacenter + S2D + SR, etc... vs Starwind vs ? This I'm sure would surely beat something like a Nutanix cluster or something similar (price-wise). -
@Tim_G
Oh, just in case someone reading this isn't aware of this type of setup:
You can definitely tell this video is with a "selling" it to you attitude, but if you can ignore that, the concept is true and accurate. I've been testing and deploying things similarly.
-
@Tim_G said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
Actually, you can do a 2-node cluster with Windows Server 2016 and S2D to get HA and redundancy...
is this production ready for real?
-
@Tim_G said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
Then you could do this using Hyper-V Server and Starwind. I don't know the pricing of Starwind off-hand, though.
The product itself is free, you only pay for support. So depends on your support desires.
-
@matteo-nunziati said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
@Tim_G said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
Actually, you can do a 2-node cluster with Windows Server 2016 and S2D to get HA and redundancy...
is this production ready for real?
Yes, this stuff has been production ready before it was officially released. What you need to make sure of is the hardware you use. I know Microsoft loves to use the R730xd with this stuff in demos, so I wouldn't go too far from something like that. I did notice in the video they do list a bunch of servers they are playing with:
- Cisco UCS C240 M4
- DataOn S2D-3110
- Dell PowerEdge R730XD
- Fujitsu Primergy RX2540 M2
- HPE ProLiant DL380 Gen9
- Inspur NFS280M4
- Intel MCB2224TAF3
- Lenovo X3650 M5
- NEC Express5800 R120f-2M
- Quanta D510B-2U
- RAID Inc. Ability HCI Series S2D200
- SuperMicro SYS-2028U-TRT+
-
@gjacobse looks like you quoted but forgot to write anything?
-
@scottalanmiller said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
@gjacobse looks like you quoted but forgot to write anything?
Was back reading - deleted as I wasn't intending to reply
-
@gjacobse said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
@scottalanmiller said in Public Cloud vs. Hosted Hyperconvergence Costing Project:
@gjacobse looks like you quoted but forgot to write anything?
Was back reading - deleted as I wasn't intending to reply
LOL, fail.
