Clearing misconceptions about the Cloud

The cloud panel at the BroadcastPro Summit brought together four experts in the field: Saleem Bhatti, CTO of STARZPLAY; Faraz Arshad, Director of Infrastructure and Platforms at STARZPLAY; Adriaan Bloem, Head of Digital Infrastructure at MBC Group; and Salloum El Dahdaah, CTO of ITWorksMe.

Saleem Bhatti set the stage for a fascinating discussion by exploring some common myths about the cloud and encouraging a debate on the varied benefits of moving to it, exploring multi-cloud environments, changes required within an organisation and more.

One of the first myths that the panellists dispelled was that cloud brings immediate cost benefits to a company. Rather, they explained how cloud offers the perfect opportunity to not just optimise a business but to also undertake activities that may have been prohibitive or difficult to achieve in the past, like analysing video streams for metadata and transcription. In some cases, the launch of a service would be impossible if a company relied on an on-premise solution.

MBC’s Bloem pointed out that “cloud is not so much about saving costs as enabling you to be a more competitive business”.

In explaining MBC’s move to the cloud, he said that one of the first things he did when he joined MBC Group eight years ago was “to aim to get rid of the servers and move into the cloud”, which the broadcaster did organically. Most of its digital infrastructure is now in the cloud, with just some hardware encoders left in a data centre in the MBC building. On the broadcast side of things, however, many things are still running on-premise.

“We don’t want to run any of our hardware. That’s not our core strength and not where we are adding value. What I would love for Technical Operations to move into the cloud is our archives, so we can run analytics on it, and use AI to generate metadata from it. That would make it easier for us to reuse it and extract more value from our vast library – something that a broadcaster rarely does if the content is on stacks of tapes in a room.”

For an SVOD service like STARZPLAY, Arshad said the requirements and benefits of moving to the cloud are different. STARZPLAY moved to a cloud solution five years ago, purely because an on-premise solution did not give it the ability to launch a great service that required efficient content processing workflows and hosting a complete applications stack, while considering the time to market (TTM) factor.

“When you have very intensive workflows, you need real elasticity to get your compute done. How can you do that with limited resource pools that you are running in your own data centres?

“One bottleneck we observed was when we started off initially on-premise. Almost immediately, we recognised that this was not going to work. This is because all the studios we work with are in the US and Europe, and for them to get their content delivered over Aspera and Signiant was causing a bottleneck. In essence, we were getting stuck. Then, once the content was ingested, efficient transcoding and packaging led to another bottleneck.

“That’s why we took the cloud approach. We shifted all our workloads with regard to content operations to the cloud, and within a couple of months we had a decent library to start the service with. Along with the content workflows, we also started hosting our micro service applications stack and CDN origins within the cloud, to achieve both scalability and performance.”

Of course, moving from on-prem to a cloud solution is easier said than done when the right skills are not available within an organisation.

Arshad said STARZPLAY built an internal digital transformation team led by him; he helped the team to not only engage with cloud transformation but also get onto the cloud adoption path, by providing the information they needed.

One company that has successfully enabled digital transformations at many regional TV stations – including LBCI, MTV, Saudi TV, Rotana, Al Arabiya, Alsumaria News, Al Jadeed, Rudaw, Kuwait TV and Sharjah Broadcasting Corporation – is Lebanon-based ITWorksMe.

CTO Salloum El Dahdaah says migration from a traditional broadcast environment to the cloud “requires a lot of integration at the back end”. To make it easier for his clients and to ensure they were able to operate from a single platform, ITWorksMe built a KWIKmotion platform and integrated the functions of KWIKmotion with multiple cloud services.

“At the back end, we undertook several integrations on multiple clouds, because there is no single cloud that offers all the answers,” he explains. “For instance, there are AI functions like transforming speech to text and enabling those translations. We used IBM Watson and Google Vision at the same time.”

The panellists also tried to ensure a clear distinction was being made between using multiple clouds and going with a hybrid cloud solution. They seemed to suggest they were using multiple cloud solutions for different tasks, and also went on to point out the subtle differences between the phrases ‘multi-cloud’ and ‘hybrid cloud’.

“We use multiple clouds for redundancy, back-up and additional services. We can’t use one cloud for different tasks,” explained El Dahdaah.

In the case of MBC, Bloem pointed out that the broadcaster was running all its backend processes and its own software on Amazon while its videos are served from Microsoft Azure.

“All of this has moved into a multi-cloud set-up, not because that’s the strategy, but because we need different capabilities from different cloud environments. That’s not nearly as clean as reading an article about multi-cloud, hybrid cloud and so on. Real-life set-ups are never that clean,” Bloem said.

Arshad also pointed out that a hybrid cloud delivery model is not for everybody.

“Hybrid cloud adoption requires a strong cloud centre of excellence (CCoE) focussing strongly on cloud governance. The workload readiness for a hybrid cloud ecosystem is really important, as not all the apps can work in a real hybrid environment. The question to ask when you are looking at adopting a cloud environment is whether it has enough resource pools to adopt the kind of scale that you are pushing towards it.

“Essentially, when considering hybrid cloud, you have several components to factor in – the bottlenecks, resource pooling, your key metrics when shifting towards the cloud, scalability or availability or throughput availability, or where you can have a hybrid cloud management layer and the orchestration of dockers between different clouds. That’s another challenge for DevOps, which is why infrastructure as code (IaC) is a key component to consider while taking the hybrid cloud approach.

“Once we have that per cloud performance and scalability matrix, you look at the percentage of traffic that is being served with the right KPAs (key performance areas) from each cloud, and which is doing better than the other. That is one dynamic approach required while adopting a hybrid cloud.

“The other approach could be a multi-cloud environment, where you have some stand-alone workloads running out of different clouds. Workloads may be run out of Oracle or Google, because we can afford to have those latency metrics for that particular region.”

El Dahdaah pointed out that one of the key benefits of the cloud is the speed of video processing.

“We have the ability to process around a hundred videos at a time on a local server, and to push the content onto different cloud systems and launch multiple instances at a time. This can be quite limited on-prem,” he remarked, adding that this is especially the case during the month of Ramadan. “During Ramadan, everybody is ingesting episodes at the same time. This year, we saw 90 episodes ingested at the same time. If we were not on the cloud, we could not have processed that data.”

The discussion came back to budget, and Bhatti asked Bloem to explain how the efficiencies gained by going on the cloud far outweigh the costs of maintaining and keeping them on-premise.

“Would you prefer to have one server and get your videos processed in 100 hours or have ten servers and get it done in 10 hours? The cost may be the same but it’s ten times faster to do the latter. Those are the real benefits rather than a simple cost equation. Cloud gives you that big peak scalability that you wouldn’t get on-premise, because no one will install that kind of compute power on-prem for using it once a day,” explained Bloem.

The panellists all agreed that moving everything entirely to the cloud may unfortunately not be possible at present, due to the massive bandwidth required and the lack of an affordable regional solution.

“You need multiple 10GB and basically the best you can get is probably multiple 1GB and it already costs a fortune. That’s not enough throughput, and then there are latency issues as well,” Bloem said.

They agreed that although cloud vendors are probably selling the ability to edit real-time in the cloud, it was still not quite there in the Middle East because of this.

They also agreed that cost is defined in terms of time and resources. In fact, this is where the real benefits of cloud can be seen.

“In five years, we will see that companies that don’t embrace this change and innovation will be lagging far behind, and then it may be impossible for them to catch up,” Bloem added.

El Dahdaah agreed that cost is “not just money but also time… as a broadcaster, sometimes the need to be the fastest to publish a news story means that you need an environment that can help you achieve that”.

However, Arshad pointed out that cloud solution providers are also looking at different pricing options for the kind of immediate scalability required to handle such workloads.

“Public cloud providers are willing to come up with better pricing models that may work for the broadcast and OTT industries, which require heavy and intensive resources. Otherwise, a broadcaster may eventually think of investing in a private cloud where they can control the cost. It needs to make sense operations-wise for them, so vendors should revisit the kind of packages they are pushing towards broadcasters for these heavy resources in terms of compute, storage and data transfer.”

When Bhatti asked if the panellists use cloud solutions as provided or customise them, each one commented that they add workarounds to ensure the cloud solution works for them.

Arshad described STARZPLAY’s journey. Initially, it had its CDN just-in-time (JIT) video streaming origins hosted on a cloud. In some cases, the public content delivery networks (CDNs) weren’t able to fetch the content, due to network latency and other performance-related challenges.

STARZPLAY introduced a hosted edge product called STARZPLAY Connect to host its JIT video streaming origins and edge application stack in different data centres localised in MENA regions; this eventually turned into a private CDN serving the video traffic along with other public CDNs. STARZPLAY is thus consuming both public CDNs using public cloud-based origins, and a private CDN to serve the video streaming.

To Bhatti’s concluding question, on which cloud solution is ahead of the rest, the panellists unanimously declared that AWS is the leader, though they pointed out the benefits of the others.

“AWS is the leader. IBM has better storage and speed. We don’t have any positive feedback on Azure. Google is impressive on the CPU side,” commented El Dahdaah.

Arshad said it “depends on which industry you are in and what you expect from each cloud-managed service”. He explained that one must measure the cloud from different angles – “resource pooling and scalability, CPU performance and its generations, and cloud network peerings with local ISPs, which help last-mile users to interconnect with cloud while consuming the offered services”.

“My peers often said they moved to the cloud because they had managed services from particular cloud services. I will go with AWS not because of their managed services, but because of their scalability and performance. They have now launched in Bahrain and the cloud footprint they have is fantastic.”

Bloem remarked: “Amazon is far ahead of other cloud providers in terms of managed services and serverless, to the point where our teams think it’s punishment if they have to run something somewhere else. But it depends on where you are in your cloud journey. For very specific purposes, if you are containerising things, Microsoft and Google are good options to consider, or maybe a combination of them. It also depends on your corporate culture. Microsoft does a lot of handholding. AWS is more of a knowledge transfer and Google is more experimental. It depends on what fits with your corporate culture.”