Discovery at Carlson·12 Why dynamic data management matters
Soumya Sen: Why dynamic data management matters
The world is in the midst of the fourth industrial revolution, according to the World Economic Forum. With technology—and what’s expected of it—changing rapidly, businesses are working to determine how best to use these tools going forward.
Carlson School Associate Professor Soumya Sen has studied many aspects of this topic, including important decisions businesses can make when it comes to managing data, both for their customers and themselves.
“With nearly everybody getting information and using their cellphones all the time, companies have to evaluate how they can do things differently,” Sen says.
Incentives offer a win-win
The explosive growth of multimedia data traffic, particularly from smartphones, caused Sen to look at new pricing structures that internet service providers (ISPs) could implement for data consumers. As of now, ISPs use penalties such as throttling, capping, and overage fees to manage network congestion.
As an alternative, Sen and his colleagues propose an incentive-based solution. Options include the use of time-dependent pricing, a dynamic structure that manages congestion by offering discounts to users for shifting some of their data use to off-peak hours. This can be particularly useful in alleviating network congestion when the demand for bandwidth suddenly exceeds the available resources due to more users coming online at the same time (such as during the 2019 Novel Coronavirus (COVID-19) pandemic).
This would help both customers and ISPs, says Sen. Customers benefit from lower fees while ISPs avoid any slowdowns on its network.
“A lot of people think that in this struggle of managing data, either ISPs win or consumers win,” Sen explains. “But, with time-dependent pricing, we’ve created a situation where everyone wins.”
Sen has received two U.S. patents, ensuring the pricing structures are implemented and work as designed for customers and ISPs.
Infrastructure Investments
The growth in cloud computing and 5G technologies create new opportunities for businesses to evaluate whether they should deploy those kinds of services on a dedicated or shared infrastructure. A shared public cloud—such as Amazon Web Services (AWS) or Azure—allows companies data to live alongside that of others on the same infrastructure. A dedicated, or private, the cloud gives businesses complete control of a slice of infrastructure on which they store their data.
Conventional wisdom says businesses should favor shared infrastructures because it saves costs.
“When you have shared infrastructure, there are quite a few benefits and we would expect a lot of businesses to choose that model,” he says. “But that is different from many of the observations we’re seeing in the industry.”
To study this, Sen and his team developed analytical models that demonstrate these decisions are much more complex than they seem. They found that while lower costs always favor a shared infrastructure, having flexibility is a major benefit to organizations that design their IT infrastructure.
One important factor is reprovisioning, or the ability to provide more capacity quickly with technologies such as virtualization.
“Depending on other cost factors, such as the gross profit margin and return on capacity investment, we see that dedicated infrastructures may sometimes be a better option,” according to Sen. “This is consistent with what we’re seeing [in the market]. This doesn’t mean it’s the only factor— security could be another reason—but the ambiguous role that reprovisioning ability can play in favoring one choice over the other is something we revealed through our analysis.”
Looking to the future
With the vast array of internet-connected devices and the need for companies to constantly evaluate usage, change in the field is always happening.
Sen’s research on data pricing for ISPs and infrastructure deployment could have a tremendous impact on how millions of people pay for and use their mobile data.
“The current measures used by the ISPs are harmful to the internet ecosystem,” Sen says. “If we are able to make this shift, we can change user behavior and enable better revenue management in multimedia-rich networks.”
Soumya Sen：动态数据管理的重要性
根据达沃斯论坛的观点，第四次工业革命正在进行中。世界期待着技术突飞猛进，企业也在积极响应，准备充分应用技术。
卡尔森学院助理教授Soumya Sen在日常研究和商业合作中深入研究了这一课题，他的研究范围包括企业在数据管理方面可以采用的一些关键措施。
“几乎所有人都在随时时刻地获取信息和使用手机，企业需要考虑是否因此而调整运营方式。”
激励带来双赢
多媒体尤其是智能手机端的数据传输呈现了爆炸式增长，针对这个现象，Sen开始研究互联网服务供应商（ISP）的数据服务的定价体系。目前，ISP使用限速、流量上限、超流量费等惩罚措施来管理网络拥堵。
Sen及其同事提出，ISP可以采用激励性的解决方案。例如，按时间定价，通过折扣引导用户在非高峰时段使用数据服务，对网络拥堵进行动态调整。在大量用户突然同时上线（例如，2019年新冠疫情期间）、带宽需求超出资源总额而造成网络拥堵的情况下，这一解决方案尤其可行。
Sen表示，此举对于消费者和ISP都有好处。消费者可以节省费用，而ISP可以避免网络拥堵。
“在管理数据这个问题上，很多人认为ISP和客户只能有一方受益。但通过按时间定价，各方都能受益。”
Sen设计的使客户和ISP双赢的定价体系已经获得两项美国专利。
基础设施投资
得益于云计算和5G技术的发展，企业现在可以选择将服务部署在专用基础设施或共享基础设施上。亚马逊云服务（AWS）、Azure等公有云服务商为企业提供共享基础设施。而专有基础设施——或者说私有云，让企业能够完全控制存储数据的基础设施。
就传统观念来说，为了节省费用，企业会积极使用共享基础设施。
“共享基础设施能带来多项收益，因此应当有很多企业选择这个模式。但是，我们观察到的行业现实并非如此。”
为了研究这个问题，Sen和他的团队开发了分析模型，发现这些决策比看起来复杂多了。虽然共享基础设施能够降低成本，但是在设计IT基础设施时，企业更加看重灵活性。
一个重要的因素是资源重配，即通过虚拟化等技术快速扩容的能力。
“考虑到毛利率、产能投资回报等成本要素，我们发现在某些情况下专有基础设施可能是更好的选择。这也和我们观察到的市场现状一致。我们通过分析发现，资源重配的能力在很大程度上影响了企业的决策，当然这并不是唯一的因素，安全也是一个考量。”
展望未来
在当今社会，介入网络的设备种类繁多，企业需要随时评估用量，行业在不断进行调整和变化。
Sen对于ISP数据服务定价和基础设施部署的研究可能会带来深远影响，改变数以百万计的消费者购买和使用移动数据的方式。
