Thursday, July 31, 2014

The Future of eMarkets for Enterprise Compute Workloads

The industry is coalescing around standard design and run time platforms for next generation software applications.  These cloud-based platforms promise to deliver true workload portability across hypervisor options and private/public cloud compute capacity.  Once workload portability becomes a reality, variable workload pricing is the logical next step and markets for transacting workload capacity will follow.

It stands to reason that e-markets for workload pricing will emerge no different than markets today for securities, foreign exchange, and power (energy).  Enterprises will seek and are procuring fixed and variable capacity to run workloads, acquiring that capacity from internal IT or the public domain.  It also stands to reason if a cash/spot market has been established, a futures (and options) market will also evolve, providing enterprises with the ability to buy (or acquire the option to transact) technology capacity at future points in time.  These two markets will make enterprise technology far more economically efficient than it is today. 

Current generation outsourcers like HP, CSC, and Accenture will shift their models from building all of their capacity (and increasing that capacity in step-wise fashion) to owning a fixed portion (for certain workload types) of their capacity and acquiring variable (or all) capacity in the spot and futures markets.  Today’s service providers (e.g., Verizon, ATT, and Rackspace) will compete (via commodity hardware) with or resell web-scale capacity (from the likes of Amazon, Google, and Microsoft) to both outsourcers and large enterprises.  The web-scale providers will become wholesalers of compute capacity but will also sell direct. 

This development will result in the integration of fixed internal private capacity with variable external public capacity.  Enterprise IT will be challenged by their business users to optimize technology utilization by workload profile.  And, as workload portability becomes a reality, firms will need to balance the fixed/variable component.  IT risk managers will adopt hedging strategies for technology capacity acquisition.  And the logical evolution of that hedge is a futures/options market for technology capacity.

A Short History

Legacy workload processing was built around acquiring fixed capacities of technology (compute, storage, and network).  That same legacy workload processing had minimal pricing variability.  Enterprise IT bought/built capacity and charged it back to business to recover costs.  The emergence of faster and more flexible technology capacity provisioning, with more transparent (and variable) pricing, has had significant appeal to the business.  Capacity providers like Amazon Web Services and SaaS providers like Salesforce.com have emerged as attractive alternatives for variable or dynamic compute provisioning, application delivery, and pricing.

The Emergence of the Cloud

Cloud computing has created internal and public resource pools of technology capacity (private and public clouds).  Enterprises need the ability to transact workloads on a fixed and variable cost basis.  While workloads can be profiled, to date they are not truly fungible (company A reporting workload is NOT identical to company B reporting workload).  On the other hand, technology capacity is fungible (fixed amount of compute capacity for a finite duration).  Regulatory and legal requirements limit the ability of enterprises today to move workloads to alternative compute capacity venues.  Security and network concerns have also constrained production workload outsourcing.

Enterprise workloads today are designed for a specific technology capacity and specification.  Workload portability (brokering) has been challenged by the inability to move workloads (internally or externally) across fixed or variable technology capacity boundaries.  Technology capacity is pooling around internal (private) and public (cloud service providers) venues.  Public cloud services are offered by the likes of Amazon, IBM, Rackspace, Microsoft, VMware, and Google.  Enterprises want the ability to process workloads in the most efficient manner.  Finally, workload efficiency is influenced by a number of factors including size, performance demands, security requirements, and network transfer rates.

The X Factor

Workload portability is an aspirational statement.  Significant obstacles have existed which challenge this aspiration.  Technologies like BOSH from the CloudFoundry project change the rules of the game.  BOSH allows an operations team to describe the entire software stack (from OS to software packages to processes) and then permits the team to change the configuration to permit horizontal scaling.  And any of these changes can be deployed to a running system.  The implications for workload cost optimization are staggering.  This capability will theoretically permit real-time, in-flight workload portability and migration.  An enterprise could start a process in-house and then finish it at Rackspace.

Market Reality Check

Large enterprises are moving to second and third generation private clouds.  These clouds may be built around converged or hyper-converged infrastructure – built or bought.  Capacity, confidentiality, and complexity are the factors that will influence decisions around public or private.  Firms will mature their private/public cloud strategies and experience, and will then look to optimize.  Once true workload portability is a reality, price discovery will start as a bilateral process.  And once that matures, it will move to a multi-lateral process.  Enterprises will conduct reverse auctions to obtain time/price/size capacity.  Concepts such as market continuity and access, order flow, liquidity, trading costs, pricing dynamics, and transparency will all emerge in the market for transacting a compute asset.  The estimated timeframe for this evolution is 2-4 years.  Compute capacity (server, storage, and network) will emerge as a fungible, tradable asset.

(This was also published July 31, 2014 on Tabb Forum, QuantForum at http://tinyurl.com/k9eemgb)