by, Adam Bogobowicz, Sr. Director of Product Marketing, Parallels
Cloud infrastructure is now available in two different flavors.
– Small Cloud (virtual private server and cloud server), and
– Big Cloud (datacenter in the cloud)
Small cloud is all about delivering a flexible compute unit. Virtual Private Server enables small cloud with a compute instance available from a provider with predefined set of resources, and prepaid for a set period of time. Cloud Server delivers small cloud as an elastic compute instance, pre-paid or available on an hourly usage basis, self-healing, and highly available.
Big cloud, on the other hand delivers infinite scalability, complete pay per usage model, high availability and data center level durability. Big cloud mimics an on-premise datacenter with the added options of multi-tenancy and hybrid architecture (ability to stretch workloads between on-premise and cloud).
For most SMB and many enterprise IT needs you do not need an infinitely elastic and scalable, priced by the millisecond big cloud. The new breed of virtualization can deliver on most of the cloud server requirements, closing the gap between virtual private server and cloud server, further improving small cloud offerings.
– Most compute tasks and workloads have predictable granularity (size) and spikiness
– At low cost the value of elasticity goes down to zero
– You can get most of the elasticity benefits out of new breed of virtualization technology without invoking a cloud genie
– You do not need a cloud to convert CAPEX into OPEX
Granularity
We always knew that the cloud would be a perfect solution if your workload or application required a copious amount of compute power and you either could not predict exactly how much, or the “how much” wildly fluctuated. Saving just 10% of your hardware expense on these massive enterprise workloads is worth millions and justifies the corresponding cloud migration.
The reality of many applications, however is that they have predictable usage patterns and more importantly that they can be delivered from one or just a few pre-sized containers or VMs. With the growth of compute power in virtual servers, the size of the compute tasks that can be flexibly managed from within a single server has also increased. Because virtual servers have an increased compute power, a number of workloads and applications which once required a big-cloud solution, now work perfectly within containers and VMs.
Low cost
With container-based VPSs available to SMBs at $20-$50 range, the benefit of cloud usage and related cost savings are just not worth the trouble. 10% savings on 20$ investment simply does not matter unless you are desperate for a latte. If anything the opposite maybe happening with SMBs willing to invest in hosted dedicated servers, because value of the applications they are running is so much higher than cost of a hosted box.
Elastic Virtualization
Virtualization by itself is now cloudy enough to eliminate the need for the cloudiness at the next level of the stack. VMs can be migrated live to servers with adequate resources; containers can be scaled for memory and CPU without shutting down. With technologies like Parallels Cloud Storage, storage resources can be added to a container or VM as needed and on the fly. These technologies allow for flexible provisioning where elasticity is built into the containers and VMs directly. For more on that see my previous blog on new breed of container virtualization.
CAPEX conversion
And finally, do you really need a cloud to flexibly lease your compute resources? Hosters have been offering this option to SMBs for the last 10 years. A container or a VM can be rented from an infrastructure provider on a monthly basis and flexible container and VM payments are now extended to hourly granularity.
So who needs a big cloud? I do not think much changed here since the early years of the cloud wave. The above-virtualization-layer clouds are still needed in all massive data scenarios:
– Massive applications with unpredictable usage fluctuations
– Massive applications with unpredictable growth patterns and need for fully automated scaling
– Overflow scenarios or what industry our describes as cloud bursting
Big clouds are for big data. What really changed is applicability of small cloud to the needs of simple workloads in the enterprise and most of workloads of the small and business businesses. This means that until SMB IT needs can be fulfilled from a SaaS layer, small cloud will be the engine of growth for the cloud providers delivering infrastructure services in both SMB and the enterprise space.