Unigrid: An Emerging True Private Cloud Architecture
| Every Friday, the Wikibon research team convenes to discuss the turbulent intersection of technology, applications, and business value.
Our focus at a recent research meeting was hardware advances. How would the intersection of technology, application delivery, and business value change if technology fully exploited the intersection of SSD, faster interconnects like NVMe over Fabrics, and True Private Cloud (TPC), “software-defined” packaging? |
The historical relationship between business value, application development, and hardware has been dominated by the physical characteristic of hardware. As they crafted applications, application developers had to think “within the box” of slow disk (5K-to-15Kμs access times) and modest interconnect speeds (effectively 100s of MB/s). Consequently, business function and value was limited by the characteristics of the applications that it could run and weave into business capabilities.
This led to significant silos of application, data, and processing resources, with those silos typically reflecting – and reinforcing – strong organizational boundaries among business functions. If finance was political ascending in the business, the needs of its application silo turned into IT investments. If marketing was politically descending, guess what happened to marketing’s IT requests.
What about cross-silo/business function investments? Except in rare circumstances: Fuggetaboutit.
Thus, for most of the first 50 years of the technology industry, the state of IT was:
– Business functions battled to prioritize improvements to IT silos – and blamed IT when they didn’t get what they wanted.
– Application developers had to build applications based on infrastructure constraints.
– Infrastructure managers adopted specialized management practices to optimize silo performance.
The Emerging “UniGrid”
Every Friday, the Wikibon research team convenes to discuss the turbulent intersection of technology, applications, and business value. Our focus at a recent research meeting was hardware advances. How would the intersection of technology, application delivery, and business value change if technology fully exploited the intersection of SSD, faster interconnects like NVMe over Fabrics, and True Private Cloud (TPC), “software-defined” packaging?
To say it in different terms, how would a business use technology differently if any data (including apps) in a system could get to any other data in the system and pay only a 5μs penalty to do so?
These are no idle musings. For example, Micron recently introduced SolidScale, an interesting entrant into the TPC space, through a partnership with Mellanox and Exeleron. We’re going to see a lot more of these kinds of systems from component companies trying to vertically integrate and attract new enterprise customers (e.g., Micron or Mellanox); to traditional systems companies trying to provide a migration path and grow workloads for existing enterprise customers (e.g., Dell EMC ScaleIO); to start-ups trying to outmaneuver established players (e.g., HyperGrid); to cloud supplier companies that intend to use their scale to turn all hardware into a service (e.g., AWS).
These efforts are just getting started, but our research shows that they all vector towards a common architectural endpoint, which we call “UniGrid.” Why UniGrid? Well, “grid” because these future systems will be organized as grids that effectively share all computing resources in the system. “Uni” because the system will be:
Universal to business people. Business people will be able to conceive of systems based on any business data, not just “their” function’s data stored in “their” function’s applications. Negotiations for access can be a thing of the past.
“This is likely to catalyze a much more ambitious and aggressive embedding of new data sources into work. Folks are going to experiment with whole new classes of applications because the process will be faster and simpler.” Peter Burris, Analyst & Wikibon GM
Uniform to developers. It will provide a single application target for an app’s entire lifecycle. Generally, all infrastructure for all development tasks and actions would be the same and feature uniform performance, cost, and evolvability characteristics.
“It would mean the system of record can be just as fast as your OLTP database, which could be just as fast as your data science workbench and your OLAP database. All systems can be equally fast.” Jim Kobielus, Wikibon analyst.
Unitary to system administrators. It will employ cloud-like management interfaces capable of administrating any type of workload through common consoles.
“Lot’s of work required to make this real – for example, how to handle multi-tenancy data services and data protection – but all that has to happen anyway. One could call this ‘UniMin,’ for unified administration.” Nick Allen, Wikibon analyst.
As Nick stated, this is a technology vision that will take years to mature, but users need to start talking about this today: discussing the idea, adding it to IT strategies, factoring it in current purchases. Over the next month, Wikibon’s David Floyer will be publishing the first of many Wikibon research pieces on UniGrid that will look at the benefits and costs and opportunities and threats of UniGrid.
| What’s our recommended action item?
CIO’s need to start developing their IT strategies around true private cloud architectures as the basis for beginning to understand where workloads need to be placed contingent upon the underlying physical characteristics of the data that they need to perform the business services that increasingly digital businesses want. Taking this step, taking that into architectural premises, and the overall model or paradigm of IT, will dramatically facilitate the introduction of increasingly powerful hardware architectures including what we regard as the one grid platform, which will emerge over the course of the next few years. |
