One reason enterprises might eventually rethink why, where and how they use public cloud--including access service providers--are the normal economies of scale that apply to enterprise computing at scale. In almost every case, small volume tends to support the economics of a cloud approach that comes with leased access to compute cycles.
But in almost every case, very-high volume tends to shift economics back to ownership. In other words, pay-per-use is affordable at low use volumes but increasingly expensive as volume grows. At some point, as the number of instances, licenses, subscriptions or other usage metrics grow very large, ownership starts to offer lower cost.
Private Cloud, Public Cloud, On-Premises Stack Considerations |
Issue | Private cloud | Public cloud | Public cloud on-premises stack |
Financial treatment | Capital expenditure | Operational expense (opex) | Operational expense |
Capital investment | Required | None (costs were opex) | Commodity |
Operational expense | People, environmental, space, maintenance | Consumption-based | People, environmental, space, maintenance |
Existing infrastructure | Sunk cost, repurpose | Excess | Excess |
Consistent operations | High with control of hardware and software | Variability when using multiple public clouds | Variability when choosing more than one solution |
People | High | Low/medium | Medium, depending upon support |
Capacity provisioned/scaling | Provision to peak demand. Slower to scale beyond | Public cloud provider can absorb surge in traffic | Provision to peak demand. Could burst/ scale into public cloud. |
Life cycle | Control selection and upgrades of hardware and software | Controlled by cloud provider | Controlled by cloud provider |
Demand variability | Predictable, less variable demand; must plan capacity | Variable demand supported by cloud infrastructure | Less variable demand to fit capacity; could burst to public cloud |
Cost per traffic volume | Cost per gigabyte can be lower for high volume | Cost efficient for low volume or varied traffic patterns | Cost per gigabyte can be lower at high volume |
Application dependencies (e.g., multicast, real-time, network fabric, VxLAN) | Adaptable, customizable to meet complex requirements | Most are not supported. Standard compute, memory, storage, network. May offer telco services. | Most are not supported. Standard compute, memory, storage, network. |
Compute intensity | GPU, CPU pinning under control | Extra costs for GPU, no CPU pinning | Extra costs for GPU, no CPU pinning |
Storage intensity | Control storage costs and performance | High cost of storage services | High cost of storage services |
Data privacy and security | Control | Some providers may collect information | Some providers may collect information |
Location and data sovereignty | Control | Country- and provider-dependent | Control |
Regulatory compliance | Built to requirements | May not meet all requirements | May not meet all requirements |
Portability and interoperability | High within private cloud environment | Lower when public cloud(s) used with private cloud | Lower when public cloud(s) used with private cloud |
Intellectual property | Owned, controlled, managed. Can differentiate offers. | Subject to terms of service. Can add cost, limit differentiation. | Subject to terms of service. Can add cost, limit differentiation. |
source: Red Hat
The key, though, is whether the “owned” capabilities are functionally similar to the leased facilities. And that might be the key insight about access service providers using public cloud for their computing infrastructure. The advantages of using public cloud might, in fact, not be scale advantages at all, but the ability to take advantage of the most-advanced and up-to-date computing architectures.
Flexibility and agility or skills availability might be cited as advantages for using a public cloud, but those same advantages might be attributed to sufficiently-large private clouds as well. The issue there is “might be.” If there are advantages to scale for cloud computing, then perhaps private cloud as built by an enterprise. never can approach the scale advantages of a public cloud supplier.
But again, scale advantages might not be the main issue. Even when reliance on public cloud costs more than an equivalent private cloud, the perceived upside of using public cloud might lie in its expected advantage for creating new applications and use cases at scale.
In other words, customers might trade higher cost for higher agility. Those with long memories will recall the operational problems caused by many decades of accumulated proprietary telco operations support systems, for example.
The shift to reliance on public clouds for OSS type functions is a move to eliminate the downside of proprietary approaches and legacy solutions.
No comments:
Post a Comment