Data center typically holds racks of servers, storage arrays and switches. Traditionally, additional capacity would mean another rack of equipment. But, that is changing now. Information Technology Industry is going through a disruption caused by hybrid cloud. Hybrid cloud envisages use of private cloud for sensitive data and public cloud for the remaining data.
Private/Public Cloud Technical Trends:
What constitutes a Private Cloud? At the bottom of the infrastructure stack is hyper converged arrays which provides cluster based architecture for the nodes to scale horizontally. Nodes can be added to the cluster as you consume more storage and compute. Flash arrays also are independently part of the bottom layer of the stack. Most flash arrays scale up and hence need a scale out file system at the upper layer to provide petabyte capacity. However, flash arrays have the benefit of low power consumption, less cooling requirements and small form factor. Legacy arrays and secondary storage also form part of this layer. On top of this layer is the Software defined Storage layer (SDS). This layer provides a simple abstraction and hides the complexities of the Storage arrays. User of this layer need not be aware of compression, thin provisioning, deduplication or any other features provided by the array. In parallel to SDS, network and compute virtualization provides Software Defined Network (SDN) and Virtual Machines (VMs) respectively. SDS, SDN and VMs together are bound in Orchestration layer to provide Software Defined Data Center (SDDC). The orchestration layer is software provided by OpenStack or VMware. IT-As-A-Service layer is the uppermost layer of the infrastructure stack that makes infrastructure provisioning to IT user in a highly automated manner. The entire infrastructure stack is hosted within a private Cloud.
On the other hand, Cloud Service Providers (CSPs) use proprietary technology to host public cloud. CSPs deploy custom hardware typically supplied by ODMs at significantly lower costs. Their proprietary Hyperscale software binds the servers into a cluster that provides high availability through redundancy and also is tightly coupled to the hardware. Amazon, Microsoft, Google, IBM and other CSPs have built their Hyperscale cloud infrastructure using proprietary technology.
The key benefits of cloud are elasticity, scalability, SLAs and usage based costs as compared to upfront investment for a fixed capacity and performance. So, Enterprises are deploying their new applications in the public cloud rather than on the premise. Most Original Equipment Manufacturers (OEMs) catering to these enterprises are transforming their products to be cloud-enabled. A few use cases supported by the Cloud are IOT, IT-As-A-Service and Big Data. IOT deals with data collected from various sensors for further analysis, IT-As-A-Service supports business driven architecture and Big Data Analytics provides highly personalized customer experience through the insights gained by analytics on large sets of unstructured/semi structured data. The CSPs appeal to the Small and Medium Business (SMB) segment and many enterprises are still experimenting with the cloud or have few applications running on it. Security perceptions weigh very high into their decisions to move to the public cloud. However, CSPs do employ very strong security mechanisms internally and provide audit reports to assuage customer fears.
Does that mean the end of on-premise data centers? The answer is Yes and No. Not all new technologies have replaced old ones. Take for example, Fiber Channel vs. Ethernet. The data centers still host significant Fiber Channel based equipment and it looks like, this technology will not go away. Tape was also supposed to be extinct by now. However, data centers still use them for low cost archiving. Also, low cost tape services are being provided by Cloud providers using a new file system layer. Also, Flash vs. HDD is another example where we have observed that there is opportunity for all. Another perspective is that a person stays in a hotel (cloud) or lives in a house (own data center) depending on his scenario. Considering electric power usage, most small consumers buy from the grid while large enterprises have custom arrangements. Similar to these examples, companies will choose between cloud or on-premise data center depending on the various factors such as size, performance, longevity of capacity requirement, security perceptions, regulations and cost benefits.
This industry trend is also influencing the mergers and acquisitions in the IT industry. There are few patterns emerging such as large companies with synergies have merged. Dell and EMC or Broadcom and Avagotech mergers are some examples. Also, few companies have demerged to provide lean companies to focus on specific markets. For example, HP split into HP, Inc., HPE and IT Services division (merged with CSC). All of these are leading to consolidation among the OEMs supporting the on-premise data centers.
At the end of this journey, we expect that there will be many large enterprises hosting their on-premise private cloud and the remaining ones will be hosted by huge Cloud Service Providers. OEMs will have to support these enterprises but there will be space for only some of them. During this journey, service providers can expect business from both OEMs and enterprises. As most OEMs cloud-enable their products in-house, they would carve out packages and partner with Service Providers to deliver and thus enable quick time to market. Also, enterprises will SaaS-enable their applications in the cloud and will need consultancy and development services from the Service Provider to re-architect their software.
Author: Ravindra Muddinagiri, Senior Director, Capgemini Product Engineering Services, firstname.lastname@example.org