In my previous post (“Software is Eating the Data Center”), I discussed the fact that a modern data center should function and appear to users as a large computer that runs various workloads. The work Vapor is doing to integrate OpenDCRE into DC/OS to provide unified IT/OT management capabilities is exciting and groundbreaking. But we haven’t stopped there.
As it turns out, data centers and workloads are not all the same. The cost of running a workload in one facility or on a given set of hardware may be vastly different than running that same workload in another location, or on a different hardware configuration. Likewise, the cost of running a workload on-premise, versus in the public cloud (such as AWS, Azure), may vary over time, and there is 0 clarity around calculating and forecasting an optimal environment for that workload.
As part of Vapor’s broader announcement of Mist, we’re excited to be bringing currency-based scheduling to DC/OS. What this means is, for a given workload and set of resources, users can get accurate, real-time cost information related to running that workload on-premise and in the public cloud.
With Mist, performance characteristics for a workload may be rationalized against physical infrastructure costs – for example, operations/sec (transactions, requests served, reads/writes, etc.) against the hourly cost for the given resources. Using these currency-based metrics, it is possible to define policies for workload placement that can be used to run that workload most cost-effectively. In essence, users are finally able to get an “apples-to-apples” comparison from what has always been an “apples-to-oranges” issue of what it would cost to run a workload in a given location or on a given platform.
The work our team has been doing on integrating currency-based scheduling into DC/OS, while also unifying IT/OT management and infrastructure, has been exciting and fruitful. I believe that, with this work, we are helping to move towards a truly data-defined data center with reduced operational expense, efficiently-deployed workloads, fine-grained resource information and sharing, and a unified toolset for managing workload and infrastructure in the same place. A giant leap forward over the state-of-the-art!
We’re looking forward to sharing more on our ongoing work with the great folks at Mesosphere to make DC/OS the ultimate data center tool.