By Harshal Kallyanpur
As a mobile ad exchange that serves more than 61 billion ad requests per month, across the globe, Vserv.mobi needs an IT infrastructure that is not only scalable as per demand but can also deliver the same level of performance across multiple geographies.
To achieve all this, the company adopted cloud computing. Explaining the need for such an infrastructure, Ashay Padwal, Co-founder and Chief Technology Officer at Vserv.mobi, says, “It is not possible for us to exactly understand and estimate the traffic. For us, the elasticity or scalability of our infrastructure, without any human intervention, was the key consideration while providing our service to the consumer.”
“We cannot wait to figure out how much traffic a consumer might bring to us and whether we have that infrastructure in place, or do we need to provision it. And therefore, cloud was a natural choice,” he adds.
Vserv therefore built its ad exchange platform keeping the cloud in mind. The company’s IT infrastructure is hosted across multiple data centres of Amazon in different geographies. This allows Vserv to eliminate latency issues and meet the response times that ad requests typically demand. “When an ad request comes to us, we have to respond to it within a 200 ms timeframe,” informs Padwal.
“Because everything is automated and I can keep adding and removing servers (as per requirement), we are able to manage 2.4 billion ad requests per day with a four-member team. Putting in this level of capex on a fixed infrastructure, and running it with 70% utilisation, and a three year depreciation on hardware—we see significant RoI on cloud in comparison,” he says.
Companies such as Vserv are prime examples of enterprises that are making the most of the cloud era. They have understood the benefits that cloud has to offer, and are taking a cloud-first approach, building or renting applications that are designed for the cloud.
Although cloud computing, with its ‘low/no capex, easy opex’ model, is especially appealing to startups and SMBs, its scalable and agile IT deployment model works well for large enterprises too.
That is why cloud computing is steadily gaining momentum in India among startups as well as enterprises. According to a recent Gartner estimate, the public cloud services market in India is on pace to grow 32.2% in 2014 to total $556.8 million, an increase from the 2013 revenue of $421 million. The research firm believes spending on software as a service (SaaS) will total $220 million in 2014, growing 33.2% over last year.
According to Gartner, SaaS is the largest overall cloud market segment, followed by infrastructure as a service (IaaS), totalling $78 million, and business process as a service (BPaaS), totalling $75 million. In terms of future potential, from 2013 to 2018, BPaaS is expected to grow from $62.3 million to $204 million; SaaS from $166 million to $636 million; and IaaS from $58 million to $317 million.
A lot of this growth can be attributed to the fact that when organisations turned to the cloud for saving cost on IT, the technology has delivered.
But is cost saving the only measure of success for cloud? What could be the other factors pertaining to RoI? Then again, are hard numbers the only way to make a business case for cloud? Let’s take a look.
Why go for cloud?
There is no denying that the biggest pull factor for a lot of enterprises is that the cloud promises to reduce their spend on IT. In most enterprises, IT is still considered a cost centre, and any saving on this front, however small, is welcomed by CXOs.
What the cloud allows them to do is they can do way with their age old infrastructure—without buying new infrastructure and being tied to it for the next few years. They do not have to worry about licensing any more, as much of it is handled in a consolidated manner in the cloud. Management costs are lower, too, as now it is the cloud service provider’s headache to ensure the infrastructure is always running, secured, patched and chugging along as smoothly as they want.
Says Monish Darda, CTO, Icertis, a Microsoft-based cloud solution provider, “If I store my files on Amazon or Azure or any cloud service provider and compare my storage costs, there are two big differences. The storage is, by default, disaster recovery(DR) and high availability enabled in the cloud. This is not necessarily true for all storage on-premise. People might use tape backups for large volumes of data but they may not have DR and high availability at all.”
He further says, “There may not be a geographical separation between two copies of the data. In the cloud, you cannot avoid this. You get it with your cloud storage, whether you want it or not. Even if the customer does not choose a geo-replication service (at additional cost), the data is still replicated at a different location for backup.”
With the traditional approach to infrastructure, even if an enterprise decided to create a redundant and highly available environment, it would end up spending double the cost of its critical infrastructure in order to create a mirror infrastructure. This would further translate into additions in management, power and space costs. With a cloud service, this cost is taken care of by the service provider.
Jignesh Upadhyay, Director, Presales – Service Assurance Solutions at CA, shares a similar view. He says that offering the same level of service uptime turns out to be cheaper with a cloud service provider, because the cost of the infrastructure to offer that level of uptime is divided among all the customers and the service provider can take advantage of economies of scale.
“As an application developer, whenever I size my application, I always plan for certain buffer capacity. As we add up all the buffers, we end up with a 40 to 50% spare capacity. All of this risk can be passed on to the service provider, as long as I get the response time I signed for, as long as I can service the number of users I had signed for. And I am able to pay the service provider as per that usage,” he says.
Darda shares a somewhat different perspective on this. He says, “Running a Windows server in a data centre might cost around $150 per month. But the same server in Azure might start costing $200. Obviously, it is more expensive. However, enterprises need to think beyond cost in terms of whether the in-house infrastructure has a high availability server and storage to back it up, or if they have even invested in capabilities to make it highly available.”
As such, cost is not the sole reason that organisations look at or should look at when adopting cloud. With cloud, enterprises are able to provision resources in hours or days as opposed to months with traditional infrastructure. This makes it easier for enterprises to reduce the product development lifecycle and take it faster to the market.
Furthermore, by reducing resource provisioning times, enterprises can go back to the drawing board, make amends to their products, test them and roll them out—all in a time which was previously allocated to just procuring the very IT resources required for this process. As the improved product rolls out in the market, the RoI starts flowing in the form of higher customer satisfaction.
Prashant Gupta, Head of Solutions (India), Verizon Enterprise Solutions, shares a similar perspective. “What enterprises need to look at is the value they are trying to derive out of their cloud investment. Any RoI model cannot be defined just on cost. The real value comes by what is your expectation out of the cloud deployment,” he says.
This sentiment finds echo with some CIOs. Take Rajendra Deshpande, CIO of BPO firm Serco Global Services, for instance. He believes that if somebody is trying to make a business case for cloud only on cost savings, they cannot get their cloud implementation right. “The IT organisation needs to clearly understand whether their cloud effort is to address pain or gain. While they would obviously look at payback, if the cloud adoption does not solve a business pain area, then the move is of no use.”
Serco, for instance, has a survey application being delivered out of the cloud. To get a survey done across the entire customer base is a big challenge. Setting up the hardware and software just for a small period is quite a task. Doing this through the cloud drives the business benefit. However, what’s more, the company is able to reach out to a larger audience. For a performance application that it has also hosted in the cloud, it is able to deliver the application experience on mobile devices like the iPad.
“90% of enterprises talk about cost saving when it comes to the cloud. [But] if you go to your CFO with the cost saving option, he will not buy your business case for cloud. Cloud is no longer an IT play, it is a business play,” says Deshpande.
This is true in Vserv’s case too. While the company was able to eliminate the capex associated with setting up hardware at various locations, the business benefit that it gets far outweighs the initial cost savings. The company’s platform can respond to an ad request within the latency limits because these requests can be served from the nearest location (because the Amazon cloud is spread across multiple zones). When a customer brings in more traffic, he does not have to wait because Vserv is able to spawn additional infrastructure within a few minutes and service the additional traffic coming in. Furthermore, as the service is cloud-based, the company is able to deliver ads across a variety of applications on various mobile end points.
Going wrong with RoI
While cost is not the absolute factor in choosing the cloud or realising its benefits, it is an important consideration. There are several factors that an enterprise tends to look at when making assumptions on how it would size, build and operate its cloud infrastructure. And these are areas that can throw up hidden costs or anomalies that the organisation might not have factored in.
The first assumption that can go wrong is that the cloud would require the exact same infrastructure as an on-premise or traditional setup. This can result in over-provisioning IT, which translates into additional cost overheads.
Says Padwal of Vserv, “What a lot of enterprises do is, if they had 10 servers on a fixed infrastructure, they come to the cloud and ask for 10 servers. They then try to estimate RoI on it. That’s where they see their estimations go wrong.”
In Vserv’s case, the company had started right out of the cloud with its applications. It did not really have a legacy to deal with. However, there might be a case wherein an enterprise is trying to migrate an application or some part of its application stack to the cloud. If this migration process is not estimated and planned right, the enterprise might start incurring costs that it had not factored in previously.
Other assumptions or events can give a false picture too. Says Vishnu G Bhat, Senior Vice President and Global Head, Cloud & Big Data, Infosys, “When an organisation looks at realising RoI on the cloud, it assumes that it would be moving certain workloads to the cloud over a particular period. If it is not able to move all of these workloads, it’s a big risk. The benefits start kicking in only after the workloads have been moved. There could be various factors for this derailment. It could be business readiness, or technology or regulatory readiness from a cloud perspective.”
Deshpande of Serco agrees that enterprises tend to forget the factors associated with migration or whether an application is to be cloud-enabled. This is where their underlying assumptions could go wrong.
“If you’re migrating something to the cloud, you need to factor in the time in which you will phase out your current hardware or software running that application. Your transition time has to clearly articulate this, else you end up keeping paying for both your existing hardware and software and your cloud adoption.”
However, if the enterprise does get its migration right and moves and starts operating from the cloud, there are other aspects of cloud which could throw surprises when the organisation gets its cloud bill.
According to Darda of Icertis, everything that enterprises use or don’t use in the cloud—but provision it—is chargeable. “I’ve seen people running into bills of thousands of dollars per month, without realising that they have left instances on—the ones that they did not need,” he says.
“From an ROI perspective, enterprises think that it’s great that they can provision resources in the day and turn it off at night and cut costs by 50%. But they tend to forget to turn it off at night, because they don’t have the process or they do not have the automation in place,” he explains.
If an enterprise wants to leverage the volume and the cost parity of the cloud, this process cannot be manual and has to be automated with tools, many of which are already available with the cloud service provider.
Furthermore, a lot of enterprises tend to think that they can easily carry their governance rules and compliance processes as applied to the on-premise infrastructure over to the cloud. Many of these rules and processes are manual or time-consuming and may not lend themselves to the more agile nature of the cloud. While organisations need not throw all such procedures out of the window, they should tweak and adjust them as appropriate.
“They should have a trust but verify mechanism in place, where they can have the users provisioning resources. But it should be monitored and verified to check if what they asked for is really what they need,” says Darda.
Nitin Mishra, Senior Vice President – Product Management, Netmagic, talks about how organisations tend to overlook costs they might not have incurred with their on-premise infrastructure but are likely to bear in case of cloud. He explains, for instance, that when enterprises host applications in-house, the power cost is a part of the administration bill and is not given to IT. However, with a cloud service, those costs are factored in.
Furthermore, when the IT infrastructure is managed internally, any additional management workload is often taken as an incremental work for the people already deployed to manage the infrastructure—not so for the cloud, in which such work is done by the external service provider.
Likewise, when the enterprise data reaches a certain threshold, in case of in-house infrastructure, the enterprise just ends up buying more hardware. Internally, such additional expenses are kept in check by questioning the need for additional storage and asking the staff, from time to time, to clean up data or delete what is not required. This frees up some space that can be re-utilised.
“But with the cloud, storage is, in principle, available infinitely. Therefore, users keep on utilising it, and may end up storing more than they should—and the enterprise keeps paying for it,” explains Mishra.
“In many cases, organisations don’t even know why some virtual machines (VMs) were switched on and nobody wants to risk switching them off, as they do not know how it will affect the business,” he says.
He also highlights other cost surprises. For instance, an overseas cloud provider may not show the TDS cost in its catalogue or the costs change when exchange rates change.
Doing it right
To get the RoI estimation right, an enterprise adopting cloud would first need to clearly outline the objective of its initiative and the business outcome it expects. Going with an assumption that it would be cheaper to run an application on the cloud as opposed to running it in-house may not necessarily be the right approach or yield the desired results.
At a basic level, the cloud provides good returns if: the dynamics of the business supported by the application have changed such that the workload can no longer be predicted; the outreach of the application would be across multiple users spread spread out locally or globally; or the enterprise is looking at new delivery channels to expand business.
However, there are a few things that enterprises can take cognisance of, when making their RoI assumptions. Bhat of Infosys believes that an enterprise should first carry out a current state assessment. They should assess the current state of their infrastructure and applications, and whether these applications can adapt to the cloud environment. If they require extensive re-architecting, the cost of transition can be quite high and have a significant bearing on RoI.
“Most enterprises only take an infrastructure view in the current state when doing an assessment. It is very important to do an assessment of the workloads of the application itself and the effort to make these applications cloud-ready,” says Bhat.
Gupta of Verizon says, “A cloud readiness assessment should be done before you start moving to the cloud. Based on the assessment, you can decide which application, what time frame and which cloud provider offers you an environment that is closest to your requirements.”
According to Deshpande of Serco, IT generally tries to make a business case for cloud by measuring the cost of cloud versus that of running a traditional infrastructure. However, the IT assets will always be at different phases of their lifecycle. Therefore, enterprises need to segregate them into different categories and decide what can be moved to the cloud.
Besides this, if the enterprise has a lot of data to migrate to the cloud, it has to plan the migration well—else the transformation could take so long that the organisation will end up bearing the cost of running it in- house as well as the cloud costs.
“Start with something which is absolutely not integrated with the core infrastructure, which can be tried and tested and understood well. It is really not worth comparing your current spend versus what will you be spending in the cloud. You will need to clearly articulate the business benefit,” says Deshpande.
The way an application is built to run on the cloud is different from the traditional IT infrastructure. Workloads need to be handled differently, the I/O aspect needs to be looked at. Sizing requirements have to be tighter and, once hosted, running and maintaining the application also needs a slightly different approach. All this is necessary for the organisation to realise full benefits of the cloud.
Padwal of Vserv shares a similar opinion and says, “What is the cost of the talent that would be required to adapt to the cloud. If the cloud effort fails, what would be the cost? It’s a good idea to go ahead with a small PoC, and go in a phased manner so that even your talent pool starts understanding and moves in that direction.”
Furthermore, some organisations have highly sensitive data. They need to evaluate the cost versus risk. According to Sridhar Iyengar, Vice President – Product Management, ManageEngine, the decision is never across-the-board and some services or applications may have to stay in-house.
“The risk perception and cost benefit could be different for each service. There is no single formula that will say ‘move everything’ or ‘don’t move everything’ to the cloud,” he says.
Mishra of Netmagic, on the other hand, believes that a lot of returns on the cloud are based on how well the enterprise is able to utilise the underlying infrastructure.
For instance, in the returns calculation, it is important to look at the per core, per GB cost rather than the absolute cost of a hosted machine (considering that different solutions may allow different physical-to-virtual ratios). Many people look at the overall features of the cloud, the DR part, but they miss out on such nuances wherein they can get better physical-to-virtual ratios.
“On a per-unit basis, they might end up paying more for a technology, but in reality they might need fewer units and it would work out better from an RoI perspective,” he says.
Taking stock of the bigger picture, Mishra says, “Even after the RoI calculation is done, it is important to check it on a monthly basis and keep validating the cash outflow to the service provider—whether it is in the same line or if there are gaps.”
Bhat of Infosys says, “In the cloud world, the cost of migration is the primary investment. You are not buying anything. You are moving workloads, and what enterprises need to look at is how much money will be spent on migration tools, the migration effort and when will the effort start showing returns: what is the payback period?”
Once the cloud adoption is beyond the payback period, there will be a net differential in the cost of running applications. The RoI needs must include a sufficient time frame that takes into account the cost of transitioning to the cloud. Once the transition is made successfully, the spend is primarily operational expenses. There are no assets that depreciate. The licensing changes to a consumption-based model. And once the enterprise reaches a certain point in that transition journey, it will start to see returns.