In the service provider world of traditionally built, on-premises proprietary hardware, applying cloud-based virtualization to existing tools such as the EPC can help improve efficiency across multiple operator pain points.
Facing the inefficiencies
Today, service providers face natural pricing inefficiencies across the board. A few years ago, for instance, Japanese networks carried a reported 3:2 data-to-voice ratio, and yet data only represented 40% of those carrier revenues. Numerous carriers have tried turning to "fairer" pricing models, but this has not necessarily led to reconciliation between economic efficiency and network efficiency -- especially as the issue has become highly politicized in the wake of recent net neutrality debates and policies.
Enter the evolved packet core (EPC), a traditionally hardware-based solution using several network devices (such as routers, switches, firewalls and servers). Since its widespread introduction and adoption, EPC technology has made network architecture simpler, cheaper, flatter, and more streamlined. EPC carries voice the same way it carries data, treating it as IP-based. High packet performance is a necessity for EPC technology because of latency problems that are not well tolerated with voice transmissions.
The EPC has helped improve network efficiency, but the classic model based upon proprietary hardware has proven far from perfect. Even with EPC deployment, pricing and bandwidth inefficiencies can persist -- inefficiencies that service providers hope virtualization can solve.
"The EPC is typically configured to meet what the operator considers is likely to be peak load in the case of predictable events (e.g., sports match) or unpredictable events (e.g., death of Michael Jackson)," notes Heavy Reading analyst Gabriel Brown in an industry report. "This inevitably means that, 95% of the time, it is vastly overprovisioned."
Solving both problems
Conversely, that other 5% can be troublesome in the opposite extreme. According to a white paper by Ixia, "99% of all network failures occur only under peak usage."
A virtualized EPC in an NFV-based cloud environment, therefore, can solve both of these problems -- scaling out as bandwidth loads demand to avoid inefficiencies while helping to prevent network congestion.
"Although the traditional EPC may have been bandwidth constrained and energy inefficient, today the EPC is better optimized in both areas," AT&T network architect Margaret Chiosi told The New IP in an email interview. "Virtualizing the EPC will shorten the time to increase or decrease the EPC capacity when appropriate."
As an added benefit, a virtualized EPC (vEPC) may consume considerably less energy than classic EPC deployments -- minimizing both opex and carbon footprints.
Moreover, vEPC instances can enhance achievement of the pricing efficiency goals originally sought by classic EPC layouts because of the ease of adapting vEPC to current business models.
"As a rule of thumb, packet core revenue is comprised of 30% hardware (chassis, line cards, etc.) and 70% software licenses (e.g., by number of simultaneous connections)," reports Brown. "This means that operators and vendors are already familiar with a predominately software model, and that EPC can adapt to a virtualized sales model more easily than hardware-based parts of the network."
Unfortunately, broad vEPC deployment is still a long ways off. The technology remains relatively nascent; commercial vEPC is less than two years old. Only "a small group of very progressive [mobile] operators" are aiming toward fully virtualizing their current EPC functions in the near-term (i.e., by the end of the next calendar year), according to Brown, who predicts that overall adoption of wholesale vEPC deployments will be slow.
To be certain, most service providers seem to be taking a tack of "if it ain't broke, don't fix it;" they are already so heavily invested in their current hardware-based infrastructures that they are reluctant to fully virtualize EPC functions.
Brown, therefore, proposes a compromise in the interim: a hybrid solution combining virtualization with a traditional EPC chassis model, making use of virtualized instances to meet demand when systems are running at or near peak capacity. This will allow service providers not yet ready to fully replace their proprietary-box infrastructure with virtualization to continue to make the most of their existing on-premises hardware while beginning to leverage the efficiency benefits of vEPC.
Either way, although vEPC deployment is relatively slow-going, service providers will not long be able to afford to avoid the trend of virtualization.
— Joe Stanganelli,
Freelance Contributor, special to The New IP