DISH & AWS versus 1&1 & Rakuten

Two different approaches to new networks

When the news broke in April that DISH would be hosting its RAN and mobile core on Amazon’s public cloud infrastructure, it sent tongues wagging about what a radical idea it was.

Now, a new announcement has come around: this time it’s happening in Germany, with operator 1&1 partnering with Rakuten to build and run a cloud native network.

@Rakuten’s building a new network in Germany. It will be Interesting to compare their approach vs @dish. Both use Open RAN, both are “cloud native”, but instead of using @awscloud (public cloud), Rakuten plans to build their own private cloud.https://t.co/C2YinYXOzN

— Danielle Royston (@TelcoDR) August 14, 2021

What’s so great about both of these announcements is that we now have two companies developing blueprints for cloud native networks, which can set a precedent for the rest of the industry.

On paper – and to many telco execs who’ve yet to learn about the powers of public cloud – these two approaches sound similar. However, each is approaching core, edge, applications (and everything in between) in very different ways.

While on the surface the plans may sound similar, we think the results will be vastly different. We think DISH will be able to build a new 5G network for a lot less CapEx, be able to offer lower prices to its customers, and provide a far more attractive tech stack at the edge for enterprises than 1&1. So let’s dig a little deeper and see what’s going on below the surface.

The RAN

Both Rakuten and 1&1 plan on building their RAN networks based on OpenRAN infrastructure. We love this approach. OpenRAN disaggregates the hardware and software in the RAN, enabling the software to be virtualized, with data processed in the cloud, as opposed to the traditional model of using dedicated equipment for each base station. This model opens the network interfaces, giving operators more choice around which vendors they use in the network. 

In an analysis of Rakuten’s approach to building its OpenRAN network, my friend Iain Morris at Light Reading cites that Rakuten’s OpenRAN network could be about 40% cheaper than constructing a traditional one. So for both operators, it’s fair to say that going down the OpenRAN route is a great move.

In our analysis, we are going to assume the cost of deployment of OpenRAN is roughly the same. DISH has selected Mavenir’s software and 1&1/Rakuten will be using Altiostar, Mavenir and Intel (read the recent news of Rakuten’s purchase of OpenRAN specialist Altiostar). For the RAN hardware, Rakuten is using a mixture of vendors, including Airspan, NEC, and Nokia. DISH, meanwhile, will just be using Fujitsu. Rakuten’s approach to using multiple vendors is truer to the objectives of OpenRAN – a multi-vendor network – but only time will tell if this approach is more cost effective. 

Building a Cloud-Native Network

There’s ‘cloud-native’ and there’s ‘public cloud-native,’ and for this we are really going to discuss each operator’s approach to using different software vendors.

For 1&1, we assume it will use Rakuten’s vendor ecosystem, Symphony, and the Rakuten Communications Platform (RCP) stack of access, core, cloud, and ops solutions. Symphony is Rakuten’s curated software stack, deployed into a single, private cloud environment. Even if this is built with the most advanced private cloud technology on the market today, this will mean that 1&1 will still be managing the usual IT software lifecycle of implementation, upgrades, bug fixes, and  change requests for its deployment. This complexity will be somewhat offset by the fact that the functionality will all be offered and managed by Rakuten. Assuming it has an updated roadmap, 1&1 will enjoy the simplicity of having one throat to choke if it is unhappy with anything in its environment. The downside of this approach is that 1&1 will not be able to get the best of breed in a particular product area, and will be 100% dependent on Rakuten’s pace of innovation and level of product investment.

DISH, meanwhile, has opted for a best-of-breed approach. It has selected all of its own vendors, making sure each passed its ‘cloud-native’ criteria. For example, it has selected Ciena’s Blue Planet for Network Intelligence and automation. From what we know of these vendors, each has a slightly different approach to being ‘cloud native,’ ranging from a lift-and-shift approach to the public cloud, all the way to public cloud native (refactored specifically for the public cloud).

For vendors that are truly public cloud native – building their code for a specific cloud – the TCO should be 80% less than the traditional on-premises approach. For those vendors that have not really refactored their software for the public cloud, there may not be much savings at all.

The approach that DISH has taken, using multiple software vendors, will see it struggling with the usual IT software lifecycle of implementation, upgrades, bug fixes, and change requests for any vendor deployments that are still doing things the ‘old way’ of managing software. This will be less so for those vendors that are doing things in the new ‘public cloud native way.’

For the software, we will chalk this up to a wash. We don’t think either is advanced enough or doing enough to take full advantage of the public cloud and it’s hard to make a judgement if one approach is better than the other. If it were up to TelcoDR, we’d suggest looking for tools like Totogi that are webscale and managed for you, so that operators are just focused on the business logic and not on the implementation details of the software.

The big difference: the approach to data centers

The big difference in the approaches between DISH and 1&1/Rakuten, though, is in their data center plans. DISH is using AWS data centers and 1&1 is building its own data centers.

For 1&1 and Rakuten, the plan is to build four big data centers and hundreds of smaller ones for its edge. The core network will be spread across the four data centers, connecting to hundreds of decentralized centers throughout Germany. Those will then be connected to thousands of antenna locations with fiber optics. Data centers and fiber optic lines will be provided by 1&1’s sister company, Versatel.

Using TelcoDR’s Data Centre Calculator, we can predict the five-year TCO for an on-premise data center, as well as what the cost would be if a company ran the exact same infrastructure on AWS. Assuming 1&1 is building its data centers for 15-million subscribers (based on 1&1’s financial reports, it has 15-million mobile and broadband subscribers), TelcoDR’s Data Centre Calculator estimates that 1&1’s deal with Rakuten will cost a total of $467 million over five years for data centers and edge locations. That puts the TCO per year at an eye-watering $93.5 million.

This approach puts 1&1 on a treadmill to upgrade, manage, and keep up with capacity planning on its own and on its dime. It will be responsible for security of the data center and for maintaining the security of the OS patches – and it will need to think about failover and redundancy of its systems, too.

DISH, meanwhile, plans to use AWS’ existing data centers and AWS Outposts and Local Zone for edge computing. It will be able to pay for the exact amount of capacity it requires. Using our calculator, our estimates show that if 1&1 were to go in the same direction as DISH and use AWS, it would cost $271 million over 5 years, or $54 million per year – a saving of approximately $39 million per year. This means that 1&1’s current plan with Rakuten is 1.7x higher than what it would have cost to run its data centers and edge locations on AWS.

Furthermore, using AWS means DISH gets access to the latest hardware (including custom chips with 40% price performance improvement over Intel chips). Not only that – as new hardware becomes available, DISH will be able to access the latest and greatest offerings by moving workloads via a few keystrokes. Security, additional capacity, disaster recovery, and failover: all built in and managed by AWS. New custom chips like Graviton2 (and future chip advancements) give immediate price performance improvement – DISH doesn’t have to invest any CapEx, it just needs to make the decision to use it, if it wants to.

Place your bets

At the end of the day, each of these operators is making a big bet on its vendor partners. 1&1 is going to have to spend a lot upfront in building its own data centers – something DISH will shift to AWS and pay as it goes. 1&1 faces an additional risk by choosing Rakuten. As Karreta Advisors points out in Seeking Alpha, Rakuten has a 12- to 16-month window to increase customer acquisitions for its mobile business to generate profit. It also has a high credit risk that may negatively affect its ambitions to build credibility among telco customers. Rakuten may be running out of runway, and 1&1’s project could take years to finish, putting 1&1’s deployment at risk.

DISH is putting a big bet on AWS. With Amazon there is not the financial risk that Rakuten may be facing. And it has hedged its bets on the vendor side by spreading the love around. It has reduced total CapEx outlay by not building its own data centers, instead relying on the already invested (and significant) CapEx of AWS. It is taking a risk that AWS may have outages and issues not within its control, but as we have seen over time, the public cloud has become safer and more resilient. While not a zero chance of this occurring, it is reducing as the public cloud becomes more robust.

Two new, cloud-native, OpenRAN networks being built in two high ARPU countries, with similar enterprise customers, each with very different approaches. It’ll be interesting to see how things play out, including the ability of DISH and 1&1 to sell to enterprise customers.

I know which one I’ve got my money on: it’s clear that DISH’s approach is smarter, cheaper, and faster. If it were me, I’d bet on the public cloud.

Recent Posts

Get my FREE insider newsletter, delivered every two weeks, with curated content to help telco execs across the globe move to the public cloud.

"*" indicates required fields

Email*

Name

This field is for validation purposes and should be left unchanged.