Fireside Chat, Interview
a16z Podcast | Mobile Invades the Data Center
- Mobile phone components, including ARM chips and Flash, are projected to become the foundation of next-generation data centers by offering inexpensive, low-power hardware that requires minimal floor space and cooling, orchestrated by sophisticated software.
- Storage purchasing models are shifting from long-term procurement cycles of three to five years to an immediate acquisition strategy based on business needs with visibility extending only one year into the future.
- Incumbent vendors face pressure to transition value from hardware customization to commodity components and software capable of orchestrating seamless scaling, allowing customers to add a single box or node when needed.
- Innovation is expected to accelerate as companies contribute to open source platforms, commoditizing lower application stack layers and driving proprietary technology to become open and available for broader innovation.
- Industry adoption will occur at varying rates, with technology expected to coalesce differently within two years, and cloud-inspired architectures involving scale on demand, perform on demand, and multi-tenancy fundamentally changing the data center over time.
- Customers will increasingly consume technology on an as-needed basis, anticipating that infrastructure requirements five years from now will differ significantly from current needs.
- Reliability strategies will shift from preventing failure in tightly coupled systems to managing recovery in loosely coupled architectures built from approximately 20 components where six are expected to fail.
- System designs will address the higher failure rates of Flash compared to 50-year-old discs by utilizing scale and software innovation to create super reliable systems, while mobile phones are predicted to achieve better Mean Time Between Failures than mainframes in stable environments.
- Newer networking companies like Cumulus are expected to succeed with thinner business models by stripping away excess value, enabling supply chain partners to provide value-added services, whereas incumbents like Cisco may retain control by keeping specific Linux versions and patches secret despite using open source.
- Security and patching timelines highlight a divergence where new solutions can provide patches within three hours of vulnerability discovery compared to incumbents who may take about three weeks, a factor cited as a reason customers are suffering from lack of transparency regarding security issues.
- OpenStack is projected to emerge as the primary platform for cloud and application innovation, potentially surpassing the influence of VMware World, while the next generation of VMware products is expected to be open source even if not immediately announced as such.
- Enterprises and mega-scales are expected to transition away from existing incumbents that provide little value, with mega-scales already making this decision contrary to incumbent claims, prompting other enterprises to adopt new technologies based on observed trends.
- Competition between young companies and incumbents will see incumbents sometimes win by adopting challenger messaging and approaching price points, while the market war between public and private cloud is expected to settle in a middle ground rather than becoming exclusively public.
- The public cloud component of major providers like Microsoft and Google is forecast to be limited to approximately 50% of their total infrastructure, constrained by the math of their underlying profitable businesses and the finite nature of public cloud resources.
- Every data center, whether on-prem or public, is expected to evolve into a cloud-inspired environment that breaks silos to run both I/O hungry and lazy applications on the same storage, driven by a new generation of administrators accustomed to pay-as-you-grow economics.