Managing private cloud costs is a key challenge for the enterprise, as it involves a complex balance of startup costs, risk mitigation and support costs.
Two common private cloud options are Microsoft Azure and OpenStack. Azure Stack, which is still in preview stage, is a highly constrained, small version of the Microsoft mega-cloud. It's limited to specific hardware vendors and configurations, which bodes well for performance, but will come at a premium price when it's formally released in 2017. Microsoft's deployment plans may broaden, but there are also constraints on maximum cluster size for Azure Stack.
The other option is the open source cloud platform OpenStack. Deploying white box, commercial off-the-shelf (COTS) hardware, and using the OpenStack code, represents one of the lowest acquisition costs for private cloud deployments. The hardware cost savings can also be significant, as white boxes come from high-volume, low-margin suppliers.
The downside to the DIY approach with OpenStack is that an IT shop has to be savvy in both hardware selection and OpenStack configuration and tuning. This is a challenge for organizations used to being spoon-fed by major traditional vendors, and knowledgeable OpenStack professionals are in short supply.
There are, however, options to balance out the lack of knowledge. For example, organizations could use a managed OpenStack provider to get COTS gear with a solid software distribution, management tools and the expertise of an experienced supplier. This could increase your overall private cloud costs, but reduces risk and buys time to recruit an in-house team.
Tuning tips to reduce private cloud costs
It's worth looking at some tuning options before jumping into private cloud deployments at scale. Solid-state drives (SSDs) boost overall performance dramatically, as do all-flash arrays, which results in an organization needing fewer servers. SSDs from distribution are also now priced lower than enterprise hard-drive disks.
Using SSDs, flash to optimize private cloud
Private cloud storage is often shared among various users in an organization, which can lead to performance challenges. Learn how SSDs and flash technology can help overcome then.
Compression and deduplication are also important. These two data services reduce cold storage costs in a private cloud, and possibly allow the SSD farms to shrink. They can be found as free-standing code or as services built into appliances and all-flash arrays.
In-memory operation can also help organizations reduce private cloud costs. Depending on the use case, in-memory can reduce server counts by more than 90%, which pays for a lot of dual in-line memory modules (DIMMs). Within a year, non-volatile DIMMs using flash/X-Point/ReRAM technology will expand the effective main memory size by factors of 10, while reducing latency to persistent storage from milliseconds to fractions of a microsecond. This will reduce server counts even more.
Finally, plan for containers, as the efficiency gains could be more than 3X. This will help further reduce server counts to save cash.
Five open source options for deploying cloud
Explore these FAQs on private cloud planning
Discover the role of automation in private clouds