Backhaul connections for small cells are seen to be a costly component, especially for outdoor/public access where dedicated wireless equipment is required. The higher bandwidth demands from LTE and Wi-Fi place ever greater loads driving further investment. Data caching, storing frequently accessed data within each small cell, could significantly reduce backhaul bandwidth required and also improve the quality of service to customers. We look at two small cell vendors who champion this approach.
The concept of data caching
Many involved with computers will be familiar with the idea of a data cache, where frequently accessed data is held more locally and can be very quickly retrieved. This could be physically integrated into a microprocessor, held in main memory rather than on disk or downloaded to your browser rather than retrieved from a website every time.
Small cells can also be equipped with a data cache which holds frequently downloaded data, avoiding the need to retrieve a large video file or commonly accessed webpage from the server each time. Instead, the data is directly available to be transmitted at whatever speed the radio interface and remote terminal can sustain.
A server located within or near the small cell gateway co-ordinates which datasets can be served locally by the small cell, making the caching operation transparent to other core network nodes and services.
Benefits to end users and to the operator
Watching a streaming video, especially HD, requires a consistently high data rate end-to-end. Delivering this locally avoids many of the choke points, whether from the originating webserver or Content Delivery Node, all the way through operators' firewalls, gateways and policy filters. This results in fewer glitches, stuttering and timeouts.
Traffic served locally from the cache frees up capacity on the backhaul, significantly reducing the cost and capacity of expensive backhaul networks.
Where caching won't help
Some types of traffic aren't suitable for caching. VPN sessions, which are encrypted end-to-end into an enterprise IT network, can't be replicated.
3G systems use soft-handoff, where the same data is sent from more than one cell. Caching isn't really compatible with that mode, and is more appropriate when the user is locked onto a single small cell. For clusters of small cells, such as found in some enterprise systems, it may be feasible to hold the cache within the cluster controller.
Caching adds a few milliseconds latency to the original datastream, as the data is written into the cache. This is typically in the order of 5 to 10ms, and (perhaps un-intuitively) is lower for higher throughput. Long latency is more of an issue for the older/slower 3G Release 99 than the latest 4G/LTE, making it quite difficult to cache Release 99 traffic due to a latency in the region of 30-40ms. 3G HSPA, with a latency of less than 10ms works much better.
Different ways of doing it
There are two vendors most publicly promoting the technology for Small Cells today.
Small Cell data caching was perhaps first championed by Intel, who of course wanted vendors to adopt their chips for the purpose.
- Ubiquisys did so with their SmartCell in 2012 and later announced a partnership with Saguna, who brought extensive CDN software expertise to the table.
- Altobridge have developed their own byte-level caching technology, originating from the need to optimise remote basestations connected via expensive and slow satellite links. They are keen to resell their caching technology to other small cell (and large basestation) vendors, under the brand name DatE. [Ed note: Altobridge went into receivership in May 2014, with a view to selling this IPR and/or other assets].
The larger the better
As you would expect, the statistical gain when more users share the same small cell increases because the chance of two or more users requesting the same data improves.
The cache can be shared between 3G, LTE and Wi-Fi – the bigger the cache the better.
In Residential Femtocells, which are very low cost consumer devices, there isn't much room for adding software or features of significant value. The operator isn't paying directly for the backhaul and the number of users sharing a residential unit is very low.
Public Access small cells, both indoor and outdoor, found in urban shopping malls, conference centres and Enterprise small cells are the most likely to show the greatest benefits.
Some good results from the field
Altobridge, who have deployed their DatE caching technology, report backhaul benefits of between 40-60% from a 3G network. These are using 16 active user small cells with up to 100 registered users, but these results show it also works well when lightly loaded and so may be worthwhile to do for an 8 channel unit. The cache hit-rate reported is very good.
Reasons behind this relate to the behavioural patterns of end users. There is a community effect based on location, such as a university campus, a corporate environment or commuting route where users want to access the same content. Often friends, colleagues, want to view the same information. Today a popular piece of content can go viral very quickly – consider the Gangnam style YouTube video, or a dramatic news story.
Analyst forecasts suggest that a high proportion of data traffic is (and will continue to be) video related – YouTube and the like. It's these large files which make the most impact on total data traffic, and drive the greatest benefits from a caching solution.
Caching also works for upload too
Ubiquisys demonstrated at Mobile World Congress a live application which sent photos or video content to a cloud service to be shared, as many do today using Facebook or Flickr. The data was sent very quickly over the radio interface to the small cell, then more slowly passed back to the cloud servers at whatever rate the Internet and servers could accept. This frees up the smartphone to move elsewhere and/or do something else, leaving the data to be delivered at its own pace.
Those who have Satellite or Cable TV services with a set-top box may already be familiar with some of the techniques used to download programs during off-peak hours. Popular films and TV shows may already be available on your box, available for on-demand viewing and reducing the strain on backhaul networks.
A similar technique can be used to pre-populate content into the small cells based on forecasting the type and source of data that would be most commonly used. This will differ for each small cell and should be based on the historic traffic pattern.
This technique continues to evolve and a lot of R&D dollars are being spent in this area. Today's fixed network CDN (Content Distribution Networks) should bring a lot of experience here – they are already doing their own analytics, pre-populating their caches, prioritising some content based on commercial agreements. Potentially, the CDN could extend right out to the base stations, delivering even better QoS to the end users and opening up new commercial arrangements with mobile operators.
Altobridge forecast that backhaul savings of over 70% may be possible using these and related new features.
Avoid breaking the system
Data flowing through an operator's network is processed for a number of services which should not be adversely affected. These include:
- Billing, where the data consumed needs to be reported and allocated to the subscriber. Prepaid systems may use a different method and need the ability to terminate service when the balance is exhausted. Postpaid systems need to avoid bill shock for roaming users.
- Lawful Intercept, where the data traffic sent to an intercepted party should be copied to the authorities, but the interception should not be noticeable by the end user.
- Policy Management, where the data rate may be reduced for some subscribers, due to their tariff plan, exceeding an allowance or having used excessive amounts of data.
- Content Filtering, where certain websites may be blocked due to corporate or personal rules. These should still be unavailable even where the content has been cached locally for use by others.
To satisfy these requirements without making intrusive changes to the core network, data delivered to the user must also flow through the core network elements. This can be achieved using transparent caching techniques, such as byte level caching, in the backhaul segment.
Recent developments in Service Provider Wi-Fi, such as HotSpot 2.0, will encourage more Wi-Fi data to be sent via the operator's core network. The nice thing about the EAP-SIM architecture developed by the WBA is that everything goes over the same pipe, same core network, allowing the same cache sharing.
With more data being delivered from a consolidated tri-mode (3G/LTE/Wi-Fi) small cell in the longer term, the use of data caching could be a useful feature to constrain the ballooning cost of backhaul.
Our thanks to Richard Lord, CTO, Altobridge for his help in writing this article.
Ubiquisys demonstrated their caching features at Small Cell World Summit 2012 and at Mobile World Congress 2013. View our YouTube video of their SmartCell Demo.