Cost Study: 3 Node Scale vs. 3 Node VMware IPOD
-
For VMware we will use the Dell R430 with 128GB per node, the same CPUs as above. Each node, just for the hardware, comes out to a list of $8,173 but with a real world discount bringing it down, at this time, to just $5,430 per node.
In addition to hardware, we need to license VMware vSphere Essentials Plus which is $5,619 for the entire cluster. So this is $16,290 for the servers, total.
Acquisition Cost: $21,909
Dell Support Cost: $5,377 ($1,792 per node)
VMware vSphere Support Cost: $3,772 ($943 for each additional year after the first)Total Pre-Storage Cost: $31,058
SAN Storage Costs:
Acquisition Cost: $49,756 EMC VNXe 3200 Hybrid 41TB
Support Cost*: $37,317
Total Cost: $118,131
*Unlike the product pricing, support cost are not listed on the EMC site and must be estimated. A very conservative number is 15% per year when purchased up front or as high as 30% when purchased later. That's percentage of purchase price, per year. So here we are estimating five years, at a low 15%, purchased up front. The real support cost could easily be double what we have here.
-
Reserved for SAN discussion...
Because the SAN component can vary so wildly we have a very difficult time doing a true comparison here. The hyperconverged solution is a scale out, ~40TB hybrid solution with node resiliency that cannot be matched in flexibility or safety, or often speed, by a traditional SAN. So we have to choose between pricing out multiple SAN nodes, different vendors, different approaches. It's not a one to one comparison so we have to assume reasonable ranges and work from there.
The EMC VNXe 3200 Hybrid iSCSI array is probably the most applicable comparison here. EMC is a division of Dell, the largest storage vendor, the VNXe line is the lowest entry point from them but good enough to be considered a viable candidate for a high availability setup such as this one.
EMC's stack setup uses a very different disk layout than our example but is very similar in overall capabilities and using caching instead of tiering. The capacity is nearly identical.
It is incredibly important to note that for this example the shared SAN storage pales in comparison to the hyperconverged storage in effectively every way. It is expected to be slower (at a minimum it has higher latency due to the storage being remote over iSCSI rather than local at least part of the time), it is a single point of failure (a good, well redundant single point, but the protection level is just not the same even so), does not scale well, requires additional network ports and is an entry point system rather than a high end SAN. In order to be even remotely in the ball park we are forced to cut corners on the storage leaving our inverted pyramid design rather inadequate in a direct comparison. But as close as we can reasonably get.
-
Similar information here: https://mangolassi.it/topic/11936/cost-study-3-node-scale-vs-3-node-vmware-vsan/4
Unlike the hyperconverged vs. hyperconverged scenario, the use of a SAN in an inverted pyramid design makes for a dramatic cost different between the two approaches. The inverted pyramid leaves us with less performance, less flexibility, less protection and dramatically more work to do while costing 30% more!
If we consider the skills needed for install (I saw install cost estimates that said that $8,000 was a starting point for the SAN portion alone) the inverted pyramid scale up very quickly in extra costs. Both installation and long term support are vastly higher even than the VMware and VSAN solution that we saw in the previous example. In house expertise and/or hired consulting skills must be much higher yet than before and over a solution lifespan (we picked five years here) the cost of supporting the inverted pyramid design is much higher with virtualization and storage skills needed, both of which are expensive and in demand.
We could look at a lower end SAN, of course, but anything lower in class than this device would lack the reliability needed to consider this a high availability setup in any way. Even as it the VNXe is only classified for five nines uptime making it officially only about the same as our normal server nodes and not as high as we hope for from an enterprise storage array.
We could look at higher end storage arrays to address the reliability concerns but this would only add costs making the solution less and less viable. Likewise we could move to a two SAN approach with the VNXe, but this would add more than $85,000 additional to the cost of the solution that is already too expensive to consider in the scenario!
This scenario does a great job of demonstrating how even cutting corners with the storage layer we are left with a lesser setup and higher cost when looking at the inverted pyramid compare to hyperconvergence at this scale.
-
@scottalanmiller said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
This scenario does a great job of demonstrating how even cutting corners with the storage layer we are left with a lesser setup and higher cost when looking at the inverted pyramid compare to hyperconvergence at this scale.
LOL cutting corners...
I just want to toss out there that while yes, this is a less effective solution by probably all counts, it barely actually cuts corners on the SAN solution compared to what people would typically consider cut corners
yeah I also know that we're not going for what's typical, because we all know typical is bad or worse, but instead we are going for idealized if not perfect.
-
@Dashrender said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@scottalanmiller said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
This scenario does a great job of demonstrating how even cutting corners with the storage layer we are left with a lesser setup and higher cost when looking at the inverted pyramid compare to hyperconvergence at this scale.
LOL cutting corners...
I just want to toss out there that while yes, this is a less effective solution by probably all counts, it barely actually cuts corners on the SAN solution compared to what people would typically consider cut corners
yeah I also know that we're not going for what's typical, because we all know typical is bad or worse, but instead we are going for idealized if not perfect.
It really is cutting corners. You are adding in "what is typical" but that is not applicable. It is cutting corners compared to
1 Any need that would you have buying gear of this nature
2 Any need that would justify paying for HA features
3 The hyperconverged alternative that we are comparing against specificallyThe idea that "many businesses won't buy this" isn't important because this is a comparison for those who are buying it. It it wasn't cutting corners by not meeting the expected uptime and not delivering the features of the comparable solution, we could make that point.
-
What are you options for backup for the Scale side?
-
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
What are you options for backup for the Scale side?
What do you mean?
if you're asking if you can use Veeam, probably not because Scale is KVM.
If you're willing to use an agent though, then any system that uses agents will backup those VMs just like any other VM host.
-
@Dashrender yes, I was thinking of agentless solutions like Veeam. So if it has KVM support it will work with Scale?
-
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Dashrender yes, I was thinking of agentless solutions like Veeam. So if it has KVM support it will work with Scale?
That I don't know.
-
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Dashrender yes, I was thinking of agentless solutions like Veeam. So if it has KVM support it will work with Scale?
I wish I could help you with this. No place that I've worked at has needed something that big
-
@wirestyle22 said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Dashrender yes, I was thinking of agentless solutions like Veeam. So if it has KVM support it will work with Scale?
I wish I could help you with this. No place that I've worked at has needed something that big
I came really close - I just missed the Scale boat. 3 years ago when looking at a replacement EHR I posted about some ridiculous needs. Many conversations with Scott - and Scale never came up. Looking back, I have to assume that Scale wasn't something we knew about quite yet. Instead I was looking at a $100K two server setup with something like 20 disks each (mainly for IOPs - this was pre acceptable SSD pricing). Management went with another solution (one they hate today) because the startup costs where so high.
Found my old crazy thread.
https://community.spiceworks.com/topic/310103-new-greenway-install?page=2 -
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
What are you options for backup for the Scale side?
Image backups are built in, totally free. Outside of that all of the normal cross platform players.
-
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Dashrender yes, I was thinking of agentless solutions like Veeam. So if it has KVM support it will work with Scale?
Does anything have KVM support?
-
@wirestyle22 said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Dashrender yes, I was thinking of agentless solutions like Veeam. So if it has KVM support it will work with Scale?
I wish I could help you with this. No place that I've worked at has needed something that big
Agentless is for small. It doesn't scale well. You normally move to DevOps at scale today.
-
@Dashrender said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@wirestyle22 said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Mike-Davis said in Cost Study: 3 Node Scale vs. 3 Node VMware IPOD:
@Dashrender yes, I was thinking of agentless solutions like Veeam. So if it has KVM support it will work with Scale?
I wish I could help you with this. No place that I've worked at has needed something that big
I came really close - I just missed the Scale boat. 3 years ago when looking at a replacement EHR I posted about some ridiculous needs. Many conversations with Scott - and Scale never came up. Looking back, I have to assume that Scale wasn't something we knew about quite yet. Instead I was looking at a $100K two server setup with something like 20 disks each (mainly for IOPs - this was pre acceptable SSD pricing). Management went with another solution (one they hate today) because the startup costs where so high.
Found my old crazy thread.
https://community.spiceworks.com/topic/310103-new-greenway-install?page=2Yeah I was reading your thread earlier. It's interesting.
