Report examines the key costs in virtualising small cells
In its report, Business Case Elements for Small Cell Virtualisation, Real Wireless’ results challenged the traditional view that dark fibre is prohibitively expensive, focusing on sensitive small cell movements on virtualised network architecture within an urban area. It found that on a five-year TCO basis, dark fibre costs could be commensurate with managed fibre given recent falls in dark fibre prices. Another key finding was that most transport options can meet bandwidth and latency requirements up to and including a MAC/PHY split with latency requirements of 2 to 6ms if the centralised processing is done at a local macrocell.
“One of the interesting things of our study was the cost of dark fibre has actually come down quite a lot recently,” explains Julie Bradford, managing consultant at Real Wireless. “At the beginning of this study, we thought the dark fibre option was going to be extremely expensive compared to everything else. But if you look at the dark fibre option over a five-year total cost of ownership and compare it to a managed fibre option, then actually they start to come into the same cost region. It’s just the split of the CAPEX [capital expense] and OPEX [operating expense] is different between the two. The next step will be to focus in on one of the virtualisation splits that we’ve considered and potentially provide a more detailed and fuller business state around that split.”
Core to its assessment was considerations towards wireline and wireless transport options to small cell sites and how reductions and costs potentially increase. It was found to take place as architectures move from traditional distributed RAN (DRAN) LTE networks towards classical cloud RAN (CRAN) architectures. This is with remote radio heads and CPRI interfaces at the network extremes with strict latency requirements in the order of 250us and bandwidths in the order of 2.5Gbps.
“We looked at four different options for how much functionality you could take from the network edge,” explains Bradford. “You could centralise and look at the sensitivity of those virtualised network options against cost and can virtualise quite a lot of the network functionality with the transport options that are available to date without incorporating too much of a cost penalty. I think that was quite a surprising finding from our studies.
“The aim of this study initially was to go along side the work that was being done by the [Small Cell Forum] while they were assessing the different technical benefits of trying different virtualisation spits to small cells. So they were coming up with the trunk gains and the processing gains that you get from centralising a MAC/PHY split or having a higher split in the protocol stack.”
- The CPRI/ORI case of the most challenging virtualisation split considered is supported only by Sub-6 GHz and dark fibre in 2015 but could be supported by all other transport options by as early as 2020 (except copper). However, we note that most wireless options will only support these requirements over short links and with good line of sight and that managed fibre products do need to evolve from packetised services offered today.
- Across the transport options surveyed, the cost increase for supporting CPRI split beyond a MAC/PHY split was most dramatic for managed fibre and microwave (assuming these will support CPRI by 2020).
- Across the transport options surveyed, the lowest five-year TCO for a CPRI split was found for Sub-6 GHz with $32k versus a managed fibre CPRI of $95k in 2020.
- The virtualisation cost, although very sensitive to whether NFV is done at a macrocell or data centre, is still a very small cost in the bigger scheme of the total TCO of a small cell site and has only a marginal impact compared with the transport connection to the small cell site.
- The power cost is also very small (in the range of $100/year) and has little impact on the overall TCO and has little variation between transport options.
The full report is available here: Business Case Elements for Small Cell Virtualisation