• Software that is decoupled from hardware creates new packaging opportunities.
• SDS + COTS will save me money. • Customers think they are Amazon, Azure, or Facebook.
• Software that is decoupled from hardware creates new economic models. But $ = what customers will bear. f(x) = value. • Customers lack a “BMaaS” function. • Some stuff sucks as SDS. • Control abstraction has MORE value than Data Plane abstraction.
“CONVERGED (CI) VS. HYPERCONVERGED (HC)” FALSEHOODS:
TRUTHS:
• CI is the answer.
• These are architectures, not products.
• HC is the answer.
• This *IS* about software (it all is!).
• One is more expensive (capital) than the other.
• HC can be packaged as software only, appliances, or engineered systems.
• HC comes in one form (appliances).
• Disaggregated Rackscale = webscale != HC.
• One is old, one is new.
• HC starts at lower $ than CI – but the price curves cross multiple times. • Because CI uses “traditional” ingredients it supports some workloads HC cannot.
• Everyone is doing it – but it’s early days in the enterprise.
• A data lake is a thing. • A data lake is easy. • A data lake will replace your RDBMS.
• A Data Lake is a composite – of multiple infrastructure layers and data fabrics = “polyglot persistence” • A data lake isn’t hard, but needs new skills, and we need to “industrialize” it. • Apps will span the data lake and traditional RDBMS.
• Public or Private Cloud is • Cloud is the answer. the Answer. • Economics vary by workload. The real • $of Hyperscale are lower. question is can IT get out of it’s own way? • Only “cloud native apps” can benefit from cloud.
• “Cloud native apps” benefit most from cloud models.