Cloud on an Overdrive

As the need for high performance computing grows and cloud technologies mature, high performance cloud computing (HPCC) is finding a lot of takers among enterprises of different hues

By KTP Radhika

A few months ago, a leading advertising and content management company in the UK faced a somewhat unique problem. The firm looks after content management and delivery for about 90 club websites on behalf of a leading sports organization. Things were going okay, until the organization wanted to re-platform these clubs across the web and mobile devices. And it was only then that the management realized the sheer magnitude of the exercise. To execute the migration in a short span of time, without affecting the performance was a complex and powerful process. And the management knew there was only one solution: a high performance scalable architecture through cloud. They did just that; the cloud architecture was able to handle 4,000 page views per second and empower the organization to serve as many as 1.9 billion advertisements each month and helped it in driving strong outcomes in both customer experience and revenues.

This agency is no isolated example; many organizations are now facing similar high demand for their high performance computing (HPC) resources. Reports show that the global HPC market will generate $220 billion in revenues over 2015-2020. According to research firm IDC, the worldwide HPC sector has been one of the “lone bright spotsin IT spending, estimated at $20.3 billion in 2011 and growing at a compound annual growth rate (CAGR) of 7.6%. Market Research Media states that the worldwide HPC market will reach $44 billion in 2020. Thanks to large operational costs, implementation time and various other critical issues, the obvious solution for many of the high performance customers would be a cloud-based set up. Industry experts claim that availability of cheap cloud resources could propel the HPC sector further in coming days.

The cloud route
High performance computing is used in a remarkably broad set of industries, research organizations and academia to build all kinds of high resolution computing models that range from economics modeling to jet engines. In most cases, they need large amount of CPU capabilities, and these data-intensive tasks require huge data storage. Traditionally, customers used to invest huge amount of capital upfront in order to create HPC environments, which consist a large number of computers inter- connected in a network such as clusters or supercomputers, and spend a lot of time maintaining and operating these large supercomputing facilities.

Explains Matt Wood, General Manager, Data Science, Amazon Web Services (AWS), HPC workloads have an insatiable appetite for computing power, but even after this capital investment, the capacity is fixed irrespective of the amount of computational horse power you need to model a specific system.Quoting his customer Jason Stowe, who is the CEO of Cycle Computing, Wood says that supercomputing facilities are too small when you need them, and too big the rest of the time.Cloud facilitators such as AWS remove these constraints. With the advent of cloud computing, customers can spin up exactly the right amount of computational power for the task, be it 100 cores or 100,000 cores, and only pay for what they use. There are no upfront fees and subscriptions. So when a job is complete, customers can just switch off their resources, and stop paying for them. Applying that utility model to high performance computing environments is incredibly powerful,Wood adds.

HPC needs extensive capex, specialized skillset to setup, provision, monitor and manage, while they guarantee truly high performance all across. Using infrastructure as a service (IaaS) and platform as a service (PaaS), users are able to utilize their HPC applications in the cloud without worrying about the costs associated. Cloud also guarantees on the quality of service. According to Subram Natarajan, Executive, Deep Computing, Systems and Technology Group, IBM India/SA, traditional HPC has been a key resource for many companies over the years to help resolve a broad range of problems that need large amounts of computing resources. These systems are usually custom-made to address a specific task. This has led to issues related to the justification of the total cost of ownership of these systems. However, companies have started to take advantage of cloud technology from where HPC services are being used.HPC cloud (HPCC) offerings can enable clients to more easily manage and prioritize HPC assets on a global basis, while maintaining operations and data securely behind their company’s firewall. This is a key in expanding HPCC role within these companies and showing an increase on their return on investment.

Growth factors
The objectives of enterprise computing are efficiency and consolidation, reducing computing space as little as possible. That said, HPC applies large amounts of resources for solving big and complex problems. In many of the cases, computing demands will be fluctuating, resulting in periods where dedicated resources are either underutilized or overloaded. A cloud-based IaaS approach for high performance computing applications promises cost savings and more flexibility,feels Srikanth Karnakota, Director Server and Cloud Business, Microsoft.

Big data explosion is one of the biggest driving factors for high performance on cloud. As more
and more data become available, there is an increasing drive to turn this data into valuable, useful and relevant information. HPC on cloud is finding increasing relevance and use among large enterprises, as they look to solve complex business problems with reduced time-to-market. This is driving an increase in adoption of HPC cloud based applications for modeling, simulations, visualization and big data analysis across the business. We cannot do high performance analysis without a highly scalable infrastructure,opines Microsoft’s Karnakota.
Another factor driving cloud based HPC is the desire to optimize datacenter space and energy efficiency. From a business perspective, using HPC on cloud may prove to be useful considering the nature of usage of the systems. Karnakota explains, For example, there will be a lot of HPC needs in the media and entertainment industry, especially on the animation side of it. If they are using an on-premise solution, once the particular project is done with and until the second project arrives, the entire hardware infrastructure is unused. So a traditional on-premise solution is not at all a sustainable solution in these types of industries. A cloud-based architecture will resolve all the issues in these type of situations.

Like the media and entertainment industry, other data intensive industries like scientific research, genome analysis, design and manufacturing etc are finding high adoption rates for HPC in cloud. A lot of agencies, with high-end research work that needs huge computational hardware with an elastic and highly available software stack, have explored HPC in cloud and have been hugely successful too. For example, CERN, the premier nuclear research agency in central Europe, has extensively utilized OpenStack to run huge data crunching tasks. This has actually cleared the air and myth that HPC may not be very effective on virtualized infrastructures as they are too tied at the hardware and OS level tweaking and tuning,explains Prasanna Venkateshwaran, Global Head Application Services & CSS Corp Labs. Given the growing needs to have a resilient HPC with the benefits of cloud computing, HPC in cloud is evolving as a commercial reality.

“With the increase in the proliferation of machine-to-machine (M2M) communications technology, HPC in cloud is getting adopted in telecom sector too,explains Vishwanath Alluri- Founder & Executive Chairman- IMImobile. This is an emerging area which will potentially generate even greater volumes of data than what is being generated at present from the activities of human users. There will be far more connected devices in the future than people. The analysis of all this data will further drive the need for HPC and cloud based architecture will save a lot of infrastructural cost.Cloud-based HPC is also largely being adopted by small and medium enterprises, academia, those who are exploring investments, and researchers.

New technologies
Not all clouds are similar, though. Clouds can come in many forms and shapes. However, to deliver high performance, clouds should optimize the system utilization when compared to a dedicated application system,opines Venkateshwaran of CSS Corp Labs. Cloud providers are now responding by offering new compute resource types that can be a better fit for a broad range of HPC work loads. Vendors are continuing their focus on improving the performance HPC systems on cloud. Along with that specific technological advancements are also taking place to improve the space and power efficiency.

For instance, Amazon offers Cluster Compute, which includes eight instances running on memory-rich Intel Sandy Bridge nodes. Amazon’s Wood says, We have some great partners and customers building excellent software to help build large scale, high performance environments.StarCluster from MIT, an open source project, helps making high performance cloud computing extremely simple to spin up a fully functioning cluster with all the libraries, configuration and management tools customers need, in just a few key presses.
Likewise, Microsoft is now offering Big Compute on its Azure public cloud. It comes with eight compute cores with 60GB of RAM and 16 compute cores with 120GB of RAM both run on Intel Sandy Bridge nodes with DDR memory, five 1TB disks, 10 GB for network and storage communication and 40GB InfiniBand for internode communication. IBM’s Natarajan feels that this booming area provides good opportunity for vendors to come up with advanced technology. For example, iDataPlex offering from IBM is designed to offer the best power and space efficiency. The vendor is also offering HPC management suite, intelligent cluster solutions, etc. on cloud. We continue to improvise on this technology by bringing in advanced cooling techniques and better system packaging,he affirms. Many large and small vendors have understood the opportunity in this area and have started to come out with smart tools to enable rapid provisioning of HPC clusters at a very high scale, across the globe.

Cloudy parts
HPC on cloud is a natural evolution. To combine the computational power offered by HPC infrastructure along with the utilization and management benefits that cloud offers is a very powerful value proposition that one cannot ignore. Cloud computing delivers lots of benefits to the HPC world, however, one must think through the potential issues that may arise out of the usage patterns, application domains, connectivity and security considerations.
HPC on cloud, is today in its infancy stages. While this is a highly potent solution, challenges associated with management and provisioning are somewhat amplified due to the size of the clusters. Also, there could be challenges associated with usage and application specific requirements. But all of these are surmountable,Natarajan observes. The lack of massive cloud orchestration software is also posing a challenge from HPC cloud. But with cloud, integration and orchestration services will see a lot of traction in coming years and that will eventually promote HPCC. Galloping technology advances such as accelerator based computing, high performance networks using InfiniBand and storage devices, all will inevitably contribute to the growth of HPC. But in order to truly flourish, it needs to find new usages and markets.
Another key issue comes from regulatory bodies. Explains Karnakota: For instance, when a pharmaceutical company does clinical trials, it includes a lot of process analysis. It involves tremendous amount of data. However, due to compliance issues, these cannot be run on public cloud environments.Also, there would be certain cases that would need special proximity, security, integration, dependencies, etc. Therefore, sometimes not all HPC workloads can activate cloud economics. Return on Investment of cloud in HPC is heavily dependent on how the application scales up and how rapidly a user can enter and leave the cloud.

Superfast future
How can HPC in the cloud get beyond the initial hiccups? To start with, issues such as connectivity and security should be addressed well in time. Industry experts hope that cloud will be the future of HPC in many cases. We believe HPC in the cloud is very viable in Indian context for almost all cases and cloud computing is the step ahead toward the vision of utility computing. The industries that could benefit using HPC cloud include, bio-chemical, medical, life sciences, pharma, academia, weather, space, gas, oil, financial, automotive/engineering, gaming/digital content, etc,assures Samuel Madireddy S, Senior Solution Architect, CSS Corp. Also, as InfiniBand gain more acceptance, many of the cloud network issues will be solved.
With the emerging technologies and solutions, cloud can serve more users with more applications, and provide benefits for more HPC applications and HPC organizations. Here? how it is. For your in-house power needs, says for an expert, you do not want to buy a power plant. Similarly, with cloud getting more popular in the HPC environment, a customer will be able to connect the terminal to a compute outlet and gain the needed compute power with all the security and support wiithout shelling out too much money. That sounds business, indeed.