Wikibon recently completed a series of studies demonstrating the economic and business value of Single Managed Entities (SME), or enhanced Converged Infrastructure, as a best practice private cloud pathway to achieving public cloud economies. By adopting an aggressive on-premise IaaS with the minimum number of SMEs, savings of 41% can be achieved, as well as significant improvements in flexibility and availability. We chose Oracle as the reference database for our study, but Wikibon believes that the principles of SME will work for almost every application.
Figure 1 shows the details of the cost-related findings of the research.
One of the major findings from extensive discussions with successful implementers of this strategy is that the organization needed to change to support this business model. A Dev/Ops model was normally chosen, and Wikibon believes that this will morph into a Cloud/Ops model.
Single Managed Entities - The Business Force Behind Converged Infrastructure
Figure 2 illustrates the concept of a single "SKU", or more correctly a single manage entity (SME). Wikibon believes that converged infrastructure is about creating SMEs, with sufficient sales and maintenance volumes to drive economies of scale. This is true for both public and private cloud workloads.
The efficiency that cloud providers such as Amazon bring is not just about delivering low cost compute hardware. Rather it is primarily about providing the lowest total lifetime systems cost - including the costs of provisioning and maintaining compute, storage, networking and software over multiple locations, as well as providing infrastructure services for backup, security and disaster recovery. As Wikibon has pointed out in previous research, (”The Business Value of Converged Infrastructure Topologies” (Wikibon 2013) and ”Duplicating Public Cloud Economics for Oracle DBAs” (Wikibon 2013) the fewer the number of “Single Managed Entities” (SMEs) in a total stack, the lower the cost of management. Compare this approach to adding additional components requiring separate management and/or maintenance, thereby significantly increasing 1 operational management and maintenance costs.
The Wikibon definition of SME includes:-
- An integrated delivery of the solution as a single component;
- Regular singular, globally pretested updates to the solution;
- A single business partner for sales, deployment, maintenance and upgrades;
- A single hand to shake;
- A single throat to choke.
Applications that can define and maintain the minimum number of SMEs will decrease the lifetime infrastructure cost of supporting an application. In the public cloud, data center infrastructure companies such as Amazon are creating SMEs with greater and greater scope. Consider the progression:
- Initially the Amazon SME scope was storage only (S4);
- The SME scope grew to include server, networking and storage hardware;
- This is growing rapidly to include databases and middleware;
- Wikibon expects Amazon to offer an SME for integrated application stacks (e.g., Microsoft Exchange/SharePoint/Lync), including all hardware, software, networking and security.
The success of Amazon in the converged operational business model will drive a requirement for a number of second sourcing infrastructure, and the potential for many different approaches to providing cloud services. Oracle is an example of a company that can deliver SME application total stacks (database as a service) for mission-critical environments. The application of the SME concept to the private cloud is gaining momentum elsewhere as can be seen in the following developments:
- Smaller ISVs will partner with infrastructure firms such as EMC/Cisco/VMware (VCE VBlock) to provide converged infrastructure combined with application support.
- Larger ISV companies such as Oracle are creating Engineered Systems such as SuperCluster to provide a total maintenance model for key applications.
- With Satya Nadella taking the reins, Wikibon expects Microsoft (after joining the Open Compute Foundation) to define and support an SME package including OCP hardware, infrastructure software and applications software for most of its application suites. We expect it to provide SMEs to cloud providers, infrastructure partners and its own 365 cloud services.
- Companies and partnerships that successfully integrate the total stack with volume will be in the best position to apply automated break/fix algorithms to continually reduce the cost of operations and maintenance.
Figure 3 shows the result of Wikibon research into converged infrastructure economics. We interviewed a large number of customers with a variety of converged infrastructure solutions, including EMC/VCE Oracle converged infrastructure (Vblock Specialized Systems for High Performance Databases), Oracle Engineered Systems such as Exadata together with the Oracle Platinum service, and Oracle SuperCluster, SAP with EMC/VCE and cloud service companies such as ServiceNOW. This data was analyzed and synthesized into a simple data model shown in Figure 2.
Wikibon believes that enterprise IT must learn to implement and manage converged infrastructure stacks that have industry volume and a minimum number of SMEs successfully. This will be a prerequisite for successfully running in-house or collocated IT, and providing a cost-effective service to the lines of business. Internal IT must run its business at an equal or lower cost than cloud service providers and ensure that the same levels of business flexibility and availability are available to enterprise users.
Summary of Previous Research Methodology & Findings
As Oracle is the leading database, both in breath of function and in breadth of mission critical deployment, Wikibon has focused much of its recent research on Oracle database best practice. In infrastructure best practice for Oracle database systems Wikibon concluded that servers have become sufficiently powerful and virtualization management software sufficiently stable that best practice for deploying x86 Oracle databases is to create an optimized and virtualized server and storage infrastructure for each level of Oracle service required.
In the research, Wikibon recommends that the best quality servers be used for the database services, and that flash be aggressively deployed to minimize IO response time and variance. The overall findings of this research are that optimizing the server and storage over traditional approaches (and increasing the spend of flash storage and DRAM/server) leads to a significant reduction in the number of processor cores that are required to handle the workload. The traditional approach takes 60% additional cores (an increase from 120 to 192 processor cores). The cost savings in Oracle licensing far outweigh any increase in cost for flash. Figure 4 shows the financial impact of this approach - with the traditional deployment of 192 cores being about $1.4 million more expensive over three years (26% higher budget cost) than an optimized configuration with 120 cores.
In a very recent follow on study on DBA productivity, Wikibon investigated the impact of infrastructure best practice on DBA productivity. In the detailed analysis of Oracle DBA productivity shown in Figure 5, Wikibon shows that by adopting best practices in Oracle environments, database administrator (DBA) productivity is improved by as much as 50%.
New Research Methodology & Findings
In our most recent research on this topic, Wikibon focused on the financial impact of converged infrastructure (CI) on budgets. In setting the stage for this investigation, Wikibon conducted a series of in-depth interviews with its membership who were using public clouds, including small, medium and very large organizations in both the private enterprise and government sectors. We found that the potential savings of taking this approach for new projects was on the order of 40-50%.
In addition, we found that the significant investment in management time and IT organization will sometimes be required to adjust to new ways of providing IT. For instance, there is often strong resistance to new methods of service delivery from lines-of-business due to compliance and data security concerns. The study also found a level of frustration from the business that in-house IT could not provide the same level of service and agility provided by public clouds and that it is imperative that IT find more responsive and cheaper ways of providing IT services.
For this project, Wikibon interviewed Wikibon members who have adopted converged infrastructure solutions, such as VCE VBlock and Oracle Engineered Systems, providing highly converged infrastructure solutions. In particular, Wikibon looked at the impact of converged infrastructure on both direct and support costs, as well as the indirect impact on performance, agility and availability.
The top half of Table 1 shows the budget items that were used in Wikibon's previous work on infrastructure best practice for Oracle database systems. To extend the model to include the financial impact of both DBA and converged infrastructure productivity, three more budget lines were added to the financial model as shown in the bottom half of Table 1.
The same Oracle best practice elements used in the previous research were used to assess the impact of converged infrastructure on overall IT financial costs, as shown in Table 2.
During the extensive interviews of Wikibon members, we assessed the impact of Oracle best practice elements individually and collectively on each budget line item (Table 3). The first column shows the budget line items, and the second column the percentage contribution to goal budget. The largest contribution is Oracle database software licenses and maintenance (67%). Columns 3 - 9 show the percentage impact of each best practice element on each budget line item. The methodology behind the detailed calculations for column 10, and the "Total 3-year Cost" & "Percentage Contribution" rows are shown in Table 3 in the previous DBA productivity research entitled "Duplicating Public Cloud Economics for Oracle DBAs".
The results of the analysis in Table 3 is shown in terms of budget costs in Table 4. The overall reduction realized by applying all Oracle best practices to a traditional Oracle environment is 41%. The largest component of the traditional environment is Oracle database licensing and maintenance ($4.7 million of $7 million, 67%). This is reduced to $2.2 million in the optimized environment (a 53% cost savings). The biggest increase in cost is storage ($0.34 million to $0.48 million, an increase of 42%), which enables the improvements in IO latency and consistency. Together with the optimized servers, this reduces the number of processor cores and thus reduces Oracle licensing costs.
Figure 6 answers the question, "Which best practice elements contribute most to the 41% overall savings?". The biggest contributor is the overall converged infrastructure (database highly converged infrastructure plus converged infrastructure), which together is 36%. Next is flash memory and server optimization at 27%, and virtualization at 20%.
The Business Case for Oracle Best Practices
The final part of the research is the business case (Table 5), where we compare the budget profile of refreshing with Oracle best practices, using database converged infrastructure, optimized storage/servers and virtualization vs. the traditional Oracle infrastructure case. Our analysis shows a $2.7 million net cash flow over 3 years and an annual ROI of 361%.
Conclusions & Recommendations
Figure 7 is the same chart as Figure 1, with the addition of the detailed data in the lower part of the table. Figure 7 shows that the overall IT savings from converged infrastructure for an Oracle database, based on the assumptions in the study, are about 41%. From these findings and other research, Wikibon has calculated that IT executives can expect savings of 40% to 50% in costs or equivalent value by adopting an aggressive on-premise IaaS with the minimum number of SMEs, even after taking into account the higher costs of highly converged solutions. This is in line with the expected savings from using public cloud.
Moreover, users reported many additional benefits of converged infrastructure in our research. These included:
- “I had been working with five vendors previously. Now I have only one throat to choke. My downtime has been significantly reduced."
- "Ninety percent of IOs were served from flash."
- "We experienced big performance boosts - both in terms of speed and consistency. Our online transactions went from 3 secs to ½ sec for invoice bookings. Turnaround for our short reports went from 8 minutes to 1 minute."
- "There are fewer bugs and quicker resolution because it’s the same stack top to bottom. Our DBAs now have a view of the entire stack vs. multiple siloes. This has been an important benefit for us."
- "All the patching is now automatic. I have one DBA now. If I were doing this on the old platform environment I would have to have six. Single support from the vendor solves a lot of finger-pointing and upgrade related compatibility problems. I’m thankfully done with those sorts of problems."
The additional benefits of reducing the number of SMEs is increasing availability, decreasing the time to solution and time to value, and improving the overall flexibility of IT. These are profound reasons for adopting converged infrastructure for Oracle databases and positioning IT to be competitive with pubic cloud IT services.
Action Item: One of the constraints to achieving the value of fewer SMEs is organization. The traditional organization of separate AD, AD maintenance, server, network, storage and DBA teams is a significant obstacle to optimizing around lower SME counts. The eventual organization adopted by successful installers of highly converged infrastructure are organized around an application-focused Dev/Ops or Cloud/Ops model.
Definition of SKU The term "SKU" (stock keeping unit ) is often used to indicate a unit that is sold and serviced (or replaced) as a unit. A Single Managed Entity (SME) takes the SKU concept a little further.
Note2 Platform assumptions in Figures 4, 5 & 7 above:
This analysis was based on using a converged infrastructure reference model of a VCE Vblock system, which included virtualization, single sku up to the middleware, flash and memory optimization, array based replication and a "single hand to shake and a single throat to choke." At the time only Vblock had such attributes. Since the original publication other vendors have come out with similar capabilities.