Cloud Computing vs. Hyperconvergence
-
As IT departments look to move beyond traditional virtualization into cloud and hyperconverged infrastructure (HCI) platforms, they have a lot to consider. There are many types of organizations with different IT needs and it is important to determine whether those needs align more cloud or HCI. Before I dig into the differences, let me go over the similarities.
Both cloud and HCI tend to offer a similar user experience highlighted by ease of use and simplicity. One of the key features of both is simplifying the creation of VMs by automatically managing the pools of resources. With cloud, the infrastructure is all but transparent as the actual physical host where the VM is running is far removed from the user. With live migration capabilities and auto provisioning of resources, HCI can provide nearly the same experience.
As for storage, software defined storage pooling has made storage management practically as transparent in HCI as it is in cloud. In many ways, HCI is nearly a private cloud, but without the complexity of traditional underlying virtualization architecture, HCI makes infrastructure management turnkey and lets administrators focus on the workloads and applications, just like the cloud, but keeps everything on prem and not managed by a third party.
Still, there are definite differences between cloud and HCI so let’s get to those. I like to approach these with a series of questions to help guide between cloud and on prem HCI.
Is your business seasonal?
If your business is seasonal, the pay as you go Opex pricing model of cloud might make more sense as well as the bursting ability of cloud. If you need lots of computing power but only during short periods of the year, cloud might be best. If you business follows a more typical schedule of steady business throughout the year with some seasonal bumps, then an on prem Capex investment in HCI might be the best option.
Do you already have IT staff?
If you already have IT staff managing an existing infrastructure that you are looking to replace, an HCI solution will be both easy to implement and will allow your existing staff to change focus from infrastructure management to implementing better applications, services, and processes. If you are currently unstaffed for IT, cloud might be the way to go since you can get a number of cloud based application services for users with very little IT administration needed. You may need some resources to help make a variety of these services work together for your business, but it will likely be less than with an on prem solution.
Do you need to meet regulatory compliance on data?
If so, you are going to need to look into the implications of your data and services hosted and managed off site by a third party. You will be reliant on the cloud provider to provide the necessary security levels that meet compliance. With HCI, you have complete control and can implement any level of security because the solution is on prem.
Do you favor Capex or Opex?
Pretty simple here. Cloud is Opex. HCI can be Capex and is usually available for Opex as well through leasing options. The cloud Opex is going to be less predictable because many of the costs are based on dynamic usage, where the Opex with HCI should be completely predictable with a monthly leasing fee. considering further the Opex for HCI is usually in the form of lease-to-own so it drops off dramatically once the lease period ends as opposed to cloud Opex which is perpetual.
Can you rely on your internet connection?
Cloud is 100% dependent on internet connectivity so if your internet connection is down, all of your cloud computing is unavailable. The internet connection becomes a single point of failure for cloud. With HCI, internet connection will not affect local access to applications and services.
Do you trust third party services?
If something goes wrong with cloud, you are dependent on the cloud provider to correct the issue. What if your small or medium sized cloud provider suddenly goes out of business? Whatever happens, you are helpless, waiting, like an airline passenger waiting on the tarmac for a last minute repair. With HCI, the solution is under your control and you can take action to get systems back online.
Let me condense these into a little cheat sheet for you.
One last consideration that I don’t like to put into the question category is the ability to escape the cloud if it doesn’t work out. Why don’t I like to make it a question? Maybe I just haven’t found the right way to ask it without making cloud sound like some kind of death trap for your data, and I’m not trying to throw cloud under the bus here. Cloud is a good solution where it fits. That being said, it is still a valid consideration.
Most cloud providers have great onboarding services to get your data to the cloud more efficiently but they don’t have any equivalent to move you off. It is not in their best interest. Dragging all of your data back out of the cloud over your internet connection is not a project anyone would look forward to. If all of your critical data resides in the cloud, it might take a while to get it back on prem. With HCI it is already on prem so you can do whatever you like with it at local network speeds.
I hope that helps those who have been considering a choice between cloud and HCI for their IT infrastructure. Until next time.
-
So what's holding HCI back? Anything? I know it takes time to break down the defenses of old-school thinking and resisting change. The market is saturated with the likes of Dell and their IPOD, and so things like brand recognition go a long ways before a new player can break into the forefront. I have a feeling that the price point is the only real stumbling block, as the technology seems almost too goo to be true. And that high price point may only be there at a very quick glance, also. I bet that if you look 5 years down the road, the TCO/ROI looks pretty good on HCI as a whole. Also, as mentioned, there is still a very strong case for cloud in many cases.
I know I'm kind of rambling here, but I wonder why we're not seeing more of HCI on the street yet, given its obvious strengths. Am I just not seeing it, or am I right and it's just not taking a big market share yet? If so, what's the big hurdle to overcome? -
@art_of_shred said in Cloud Computing vs. Hyperconvergence:
So what's holding HCI back? Anything?
Huge marketing and sales pressure from the non-HCI arena. There is so much money in the traditional approaches that it pays for a lot of marketing to keep people doing it no matter how bad it might be for them.
And in the enterprise space, few HCI solutions are ready for the 100+ server market. You move to cloud computing there, typically, today. HCI tends to be what the SMB turns to instead of cloud because they don't have a need for the automation of cloud.
-
@art_of_shred said in Cloud Computing vs. Hyperconvergence:
So what's holding HCI back? Anything?
In addition to what @scottalanmiller said, there is also the moving commodity line of IT infrastructure and services.
Fewer and fewer SMB need enough local compute power to warrant the cost of HCI hardware. The entry price point for @scale equipment is $25k. For what it provides, I believe this is a very good price point too.
The problem is my clients are reducing hardware to the point of only needing a single physical server to handle all of their local virtualization workloads.
A Dell from @xByteSean with a bunch of 4tb drives in raid10 and 128gb of ram handles so much and costs ~$6k.
-
If HA is not a concern @scale offers a single node appliance configuration (SNAC) which has all the features and capabilities of a three node cluster, sans HA and rolling updates. Additional nodes can be added if the environment grows and/or the technical requirements change at a consumable price point.
-
@Shizrah said in Cloud Computing vs. Hyperconvergence:
If HA is not a concern @scale offers a single node appliance configuration (SNAC) which has all the features and capabilities of a three node cluster, sans HA and rolling updates. Additional nodes can be added if the environment grows and/or the technical requirements change at a consumable price point.
What's the price for something like that? I would have to assume it would be roughly 1/3 the $25K pricetag @JaredBusch mentions. Does a customer gain the HCI abilities at 2 nodes, or do they have to fully jump to 3?
-
@Dashrender said in Cloud Computing vs. Hyperconvergence:
@Shizrah said in Cloud Computing vs. Hyperconvergence:
If HA is not a concern @scale offers a single node appliance configuration (SNAC) which has all the features and capabilities of a three node cluster, sans HA and rolling updates. Additional nodes can be added if the environment grows and/or the technical requirements change at a consumable price point.
What's the price for something like that? I would have to assume it would be roughly 1/3 the $25K pricetag @JaredBusch mentions. Does a customer gain the HCI abilities at 2 nodes, or do they have to fully jump to 3?
HA only at three, I asked.
-
@scottalanmiller said in Cloud Computing vs. Hyperconvergence:
@Dashrender said in Cloud Computing vs. Hyperconvergence:
@Shizrah said in Cloud Computing vs. Hyperconvergence:
If HA is not a concern @scale offers a single node appliance configuration (SNAC) which has all the features and capabilities of a three node cluster, sans HA and rolling updates. Additional nodes can be added if the environment grows and/or the technical requirements change at a consumable price point.
What's the price for something like that? I would have to assume it would be roughly 1/3 the $25K pricetag @JaredBusch mentions. Does a customer gain the HCI abilities at 2 nodes, or do they have to fully jump to 3?
HA only at three, I asked.
For that $25k entry point for the 3-node configuration, is Scale supplying just the 3 nodes themselves, or is it a package deal that also includes a backplane switch, cables, etc.?
-
@art_of_shred The setup I got when we went with them for a 4 node setup included the backplane switch and cables.
-
@dafyre said in Cloud Computing vs. Hyperconvergence:
@art_of_shred The setup I got when we went with them for a 4 node setup included the backplane switch and cables.
So a single node would likely be a little less than the "1/3 of the cost of 3 nodes" since a single node doesn't require any of the peripherals.
-
@art_of_shred said in Cloud Computing vs. Hyperconvergence:
@dafyre said in Cloud Computing vs. Hyperconvergence:
@art_of_shred The setup I got when we went with them for a 4 node setup included the backplane switch and cables.
So a single node would likely be a little less than the "1/3 of the cost of 3 nodes" since a single node doesn't require any of the peripherals.
That'd be a question for their sales team, but I'd expect you would be right.
-
yeah, but how much is a backplane switch? I suppose it could be around $1500 (really I have no clue how much 10 Gb switches cost).
-
@Dashrender said in Cloud Computing vs. Hyperconvergence:
yeah, but how much is a backplane switch? I suppose it could be around $1500 (really I have no clue how much 10 Gb switches cost).
Um, not realistically. You want redundant 10GigE. That's not $1500.
-
@scottalanmiller said in Cloud Computing vs. Hyperconvergence:
@Dashrender said in Cloud Computing vs. Hyperconvergence:
yeah, but how much is a backplane switch? I suppose it could be around $1500 (really I have no clue how much 10 Gb switches cost).
Um, not realistically. You want redundant 10GigE. That's not $1500.
How much for 2 10 Gb switches? Also how many ports do they normally include in a standard 3 server setup? 12 per switch?
-
@Dashrender said in Cloud Computing vs. Hyperconvergence:
@scottalanmiller said in Cloud Computing vs. Hyperconvergence:
@Dashrender said in Cloud Computing vs. Hyperconvergence:
yeah, but how much is a backplane switch? I suppose it could be around $1500 (really I have no clue how much 10 Gb switches cost).
Um, not realistically. You want redundant 10GigE. That's not $1500.
How much for 2 10 Gb switches? Also how many ports do they normally include in a standard 3 server setup? 12 per switch?
Number per switch varies a lot. But for each node in your cluster you need four ports, dual front plane and dual back plane. So for a three node setup you need a minimum of eight ports for the LAN and six ports for the backplane. That's total ports, not per switch. But that's fourteen total ports between the two sides (whcih can be VLANed on just two switches.)
-
Am I the only one going "@scale isn't awesome because of the price, they take all of the additional knowledge needed for the hardware side of the HA equation off the table"? Add on the custom storage layer, which was already quick before SSD came into the picture, and it's just a lot of win for small to mid size companies that have a need for their own server infrastructure.
Sure, you could build it yourself for cheaper, but it's gonna take a lot more time. The real costs in a roll-your-own solution is the time sink.
-
@travisdh1 said in Cloud Computing vs. Hyperconvergence:
Am I the only one going "@scale isn't awesome because of the price, they take all of the additional knowledge needed for the hardware side of the HA equation off the table"? Add on the custom storage layer, which was already quick before SSD came into the picture, and it's just a lot of win for small to mid size companies that have a need for their own server infrastructure.
Sure, you could build it yourself for cheaper, but it's gonna take a lot more time. The real costs in a roll-your-own solution is the time sink.
Actually, they are not in the S side of SMB unless said SMB is doing things either wrong or is a very special case. We all know which one that likely is.
For the M side of SMB, it is a great solution for their needs.
-
I guess it really boils down to - do you need HA? If your SMB needs HA, then the cost of Scale is probably pretty great. Topped by the fact that it's an easy to use, turn-key or near turn-key solution and has full support from Scale.
But as we are constantly drilled by @scottalanmiller, most SMBs don't need HA. They can often afford hours, or even days of downtime.
Getting servers in the $6-8K range as JB mentioned earlier seems completely reasonable, and a significant savings considering that most SMB are rare if they need more than one worth of compute power these days.
-
someone sent me a message asking how to get the server I mentioned for $6k.
Here you go.
-
@Dashrender said in Cloud Computing vs. Hyperconvergence:
But as we are constantly drilled by @scottalanmiller, most SMBs don't need HA. They can often afford hours, or even days of downtime.
I generally agree with you. Our local 911 center was down for something like 36 hours a couple weeks back, made my head spin that. It's just the edge case to support @scottalanmiller's rule.
