Diversity at Box Inc - Case Study Example

Proposal for Diversity at Box Inc Managing a successful data center requires a significant investment plan for all the involved companies (Greenberg et al. 68). Such a factor makes the services produced by some firms unaffordable for a number of individuals and businesses. Accordingly, this paper will first discuss the cost of cloud data for most firms today. A breakdown of the costs involved reveals the need for optimizing work done per unit of investment. It is unfortunate that resources for the data centers operate at the lowest levels of utilization because of resource fragmentation and stranding. For such a reason, Box Inc. is a company that manages the costs of all the factor inputs to offset the costs of data storage for its clients. To tackle the problem of expensive data costs, this work proposes to the company to increase network agility as well as provision of incentives that will shape consumption of resources. If the company adopts the proposed policies, it will reach a diversified provision of cloud data storage for its customers and have a better competitive edge against firms such as EMC and Google. This work will also evaluate the current trend where most companies, including Box Inc. use geo-distributed networks of data centers. The move reduces latency to the customers, but improves reliability in case an outage happens on an entire site (Greenberg et al. 68). It is noteworthy that without proper management and design, the geo-centers lead to increased costs of data management. Additionally, leveraging such centers that firms design services to benefit from it. To solve the problem, there is a proposal that the company undertakes a joint network optimization for center resources and providing new units for geo-distribution state. This paper outlines various strategies to enhance diversity at Box Inc. and, consequently, increase customers for the company. Table of Contents Abstract 1 Differences in Cloud Data Centers 4 Types of Data Centers for Cloud Services 5 Server Costs 6 Reducing These Costs 7 Infrastructural Costs for Box Inc 7 Reducing the Costs of Infrastructure 7 Power 8 Reduction of the Costs of Power 8 Network 9 Reducing the Costs of Networking for Box Inc 9 The Rationale for Reduction of Costs 10 Geo-Distribution as A Way of Diversity 10 Conclusion 11 Works Cited 12 Introduction There have been massive investments in services that support cloud data in the recent past by firms such as Facebook, eBay, Microsoft, Google, Yahoo, and others (Greenberg et al. 69). This work attempts to explain the concept that these data centers have as well as identifying the opportunity areas for R&D impact in centers of data network. As such, determining the costs involved in operating a data center gives the background for diversity of services for Box Inc. because the costs of operation form determinant factors in provision of services to customers. The paper begins the study with the question: Where Does The Money Invested Go In Contemporary Cloud Data Units? The study considers a data center housing that has an order of 50,000 servers, such as the number that Box Inc. operated initially before upgrading (Greenberg et al. 70). Such data will help in quantifying the approximated costs of a data center. The center in question will have contemporary well-understood technologies and using high quality yet available material. Table 1 gives an outlook of the projected expenditures. There is a consideration for amortizing costs, i.e. the work amortizes one time purchase over considerable lifetimes with an assumption of a 5% cost of money (Greenberg et al. 69). Through amortizing, there results a common cost run metric that applies for both one-time purchases such as servers and ongoing expenses such as power. Somewhat, details may vary according to site and moments in time, but the table indicates the major costs. As much as networking is not the main cost category, this paper argues that systems innovation and networking is the fundamental for reducing costs and benefiting the most out of each invested unit of money. Table 1: a guide to the distribution of costs in a data center (Source: Greenberg et al. 68) Differences in Cloud Data Centers There is a need to ascertain why the existent enterprise data solutions cannot work successfully for cloud data centers. First, contrary to data centers, the leading cost for the enterprise concerns operational staff. In data centers, such costs are almost negligible, usually under 5% because of automation, which is the reason for its omission from table 1. In an enterprise that is well run, the ratio of IT staff workers to servers is 1: 100 because there is a partial automation (Greenberg et al. 70). Human error is the source of problems affecting the firms from performance. The case is different in the cloud data service, which has mandatory automation for scale and it forms a principle of design. Typically, a well-positioned cloud center has a ratio of 1: 1000 for the workers to servers. As such, Box Inc. operates at optimal automation, a factor that makes the management of costs a critical component of its performance. If there is no proper management of the costs involved, the company will meet too strong opposition to withstand operations (Greenberg et al. 70). Types of Data Centers for Cloud Services Box Inc. is a mega data center because it works on about tens of thousands servers that draw a significantly high amount of power at peak. There are massive data analysis applications that run to ensure that the company manages the array of services that it offers to its clients. Such technologies suit the operations of the company because some of the processes involved require large amounts of fast RAM; others need massive disk I/O bandwidth, yet other need large numbers of CPU cycles. Such problems need extensive communication appliances. However, constraints in the data network of the company cause barriers, which prevent agility such as ACLs, VLANs, Load Balancers, Broadcast Domains, and service-specific engineering (Greenberg et al. 70). The cost of communication tends to rise if the company spreads out the servers across the data centers separated over long distances because the market price for such surpasses the intra-building links. Cloud services usually pile on one another, and therefore, having a relatively large number of servers in a central position eases the system design and lowers efficiency costs. Such a scenario supports applications with an array of dependencies and related communication requirements. Box Inc. has a specialized area of innovation, which is the same across the industry, which is designing and deployment of micro-data centers that work on about thousands of servers that draw power peaking to 100s of kilowatts. There are highly interactive applications such as office productivity applications and query or response applications, which suit the company’s geo-diverse centers. Considerably, the company should strengthen that area because of the associated challenges. In the wake of stiff competition from other firms in the industry, there is a need for Box Inc. to diversify into the geo-diverse centers. The company already utilizes micro-data centers primarily as nodes for distribution of networks such as email (Greenberg et al. 71). Cost Breakdown and the Rationale for Diversity This section explores the details of costs of running a data center as outlined in table 1. Server Costs Like indicated in the table, the largest data center costs for the company are for the servers. For instance, if there is an assumption that 50, 000 servers have a relatively aggressive prices of say $ 3000 for every server. Such a price could also be a cost of 5% of money and an amortization of 3 years, which brings the amortized cost of servers to $ 52. 5 million per year. With the prices of operation this high, there is a need that the company considers attaining a high utilization for every dollar invested. There are cases when the rate of utilization may turn out to be too low and mean that the company risks a great deal of its investment. The causes for such could be one or a combination of the factors discussed: Uneven application fit: a server combines CPU, network, memory, and (often) components of storage. There are cases when the application feet of the servers fail to utilize one or more of the components. Uncertainty in demand prediction: there are chances that demands for cloud services may spike quickly, as is the case for new servers far beyond the prediction of the company. Long time scales for provisioning: there is a tendency of large purchases for either upgrades or new builds, which may involve buying proponents in bulk. For such a case, the company expects servers and associated parts to last for long, which helps it to carter for the demands over a long period. Risk management: if successful enough, the demand of a given service may surpass the projected demand levels, as suggested. If there are no measures instituted, the case could result in harm for the operations of the firm. Inability for the server to meet the prevailing levels of demand causes failure when there is looming success. Hoarding: it is easier to get buy in from a team that provisions new resources than returning them because of the factors already discussed. Such inefficiencies multiply over service instances. Reducing These Costs A key element of the solution for the above-mentioned problems is agility. Agility will enable the company to increase utilization of resources and in a way, result in improved efficiency. Agility means the company should poses the ability to shrink and expand resources to meet the prevalent demand levels and draw the same resources from is optimal location. Currently, the principle barrier to agility is the network because it causes fragmentation of resources, which results in lowered server utilization. Infrastructural Costs for Box Inc Infrastructure implies the components of the system dedicated for to supply the unit with power consistently and to evacuating heat. To some sense, infrastructure forms the overhead of data centers for cloud services. Like indicated in table 1, that there is a substantial aggregate cost for the infrastructure, the costs are one of the main players in the unit. Reducing the Costs of Infrastructure There are considerations for some basic ways of cutting down the costs of power and other infrastructure for the company. First, the projections of power to high levels are because of the rising demand to sustain the unit. If the company were to relax the power requirements, it would result in scaling-out designs of data center using huge numbers of commodity and low costs servers even though their rate of failure is relatively high. For such a case, there would be problems that would result in the loss of productivity of the company. Another way of cutting down on the usage of power for the system is using larger numbers of networks that have smaller data centers with a target of 1: N resilience at the level of data center. That would mean that the failure unit translates to an entire data center. If the company attains resilience at data center level, it can strip out layers of redundancy for each data center. For instance, the unit will not require the UPS and the generators (Greenberg et al. 71). Power There is a need to establish where power goes to in the system because it is one of the costs incurred by Box Inc. achieving such will require an application of state-of-the art practice relating to modern well understood techniques as well as implementation basing on excellent quality, but widely available material. The Green Grid [6] gives a metric to explain power usage efficiency (PUE) for data centers. A modernized facility will attain a PUE of ∼1.7 that places it far below the average for the globe but among the best. An estimation of the power usage of a mega data center such as Box Inc. uses the assumption of a PUE of 1.7 as well as a reasonable utility price of approximately $ 0.7 for every Kilowatts. 50, 000 servers each drawing about 180 watts amounts the costs to about $ 9.3 million per year (Greenberg et al. 71). Reduction of the Costs of Power Box Inc. works to cut down on the costs of power supplied to each server because it forms the most significant way of reducing the demands for power to the infrastructure. Such a move will also help the company in provision of just enough equipment that will avoid loss power and cause an increase in the expenses incurred. Such improvements are likely to result from hardware innovations that may include the use of high efficiency supply of power as well as modules for voltage regulation. There is a consideration for running a data center hotter as a way of innovation influenced by networking. Such an innovation would cut down the costs of power involved in cooling and save money on cooling equipment as well as the power that it consumes. However, there are initial indications that higher temperatures cause a higher rate of failure of equipment. Therefore, the innovation team at the company has a challenge of ensuring the best methods of doing so. Network The capital costs incurred in purchasing a networking gear forms a significant fraction of the costs that Box Inc. and other firms meet while struggling to offer cloud data services to its customers. The firm has its costs of network concentered primarily in routers, switches and load balancers. The remnants of networking costs distribute in a wide area of networking such as peering. Peering involves handing off traffic to Internet Service Providers, which deliver packets to end users of the company’s services (Greenberg et al. 71). Another area could be inter-data center links that carry traffic between centers that are geographically distributed. Back-of the-envelope calculations for networking costs for a wide area proves challenging because the costs defy a breakdown into costs such as traffic volumes or fiber miles. Rather, there is a variation of costs according to sites, and they vary in time according to industry dynamics. Such dynamics may include tariffs, varying options for wide area transport and regional peering facilities. There have been significant drops in the costs over the recent years, but they remain significant for the company. Reducing the Costs of Networking for Box Inc Costs for wide area networking form sensitive details for site selection and the dynamics of the industry. Accordingly, clever designs of transit strategies and peering combined with maximum placement of mega and micro data centers all help in reduce the costs of networking. Box Inc. could still consider maximizing data usage through a better design of the systems involved such as partitioning of their functionality as well as their inter-center state. The Rationale for Reduction of Costs Box Inc. requires having the lowest costs incurred possible, but still needs to make profits because it is a business. There are stronger forces of competition in the current market, with most of the competitors being prominent in the market. The company needs to make enough sales that will warrant enough resources for offsetting the costs of operation. The businesses maintains a sharp competitive edge because it offers its customers free cloud storage up to a certain limit. Such an incentive gave the corporation a ground for competing the rest considering that they have better prominence in the market. The law of demand and supply holds that if other factors remain constant, reducing commodity prices results in higher demand for products. This factor would mean that Box Inc. should offer competitive prices to the customers. As such, the lowered prices will only remain meaningful if the company utilizes the best approaches for competing the rest of the industry. As mentioned in the introduction, the company requires diversifying its mode of approach to the competitive nature of business. One such method of diversity is geo-distribution, which involves spreading of data centers across larger geographical regions. Geo-Distribution as A Way of Diversity The company considers that speed and latency matter because there is substantial empirical proof that performance affects revenues directly. For instance, Google recorded a 20% loss in revenues because of a specific experiment, which increased the time of displaying search results for customers for as little as 500 microseconds (Greenberg et al. 72). Similarly, Amazon lost 1% of its revenues because of an additional delay of about 100 microseconds. Such data provides Box Inc. with a strong motivation for geographically distributing its data centers around the globe to minimize the speed-of-light delays and open more chances for additional research challenges. Such challenges could involve determining the locale for a data center, how large they should be, and how they could utilize the geographic diversity as a source of redundancy to improve system availability. The company will benefit from an improved customer base and have a strong competitive reputation against other firms in the industry. Conclusion This study has analyzed the costs of running cloud data center systems. Such was necessary because there was a need to establish the best approach that Box Inc. would use to edge out competition in the industry. It was therefore worthwhile considering the effects of prices on the quality of services that the company would provide its customers. The paper proposed that the firm geo-diversifies, which implies that it starts newer centers across different regions of the world. As such, the company would meet a completely new set of costs to start the new centers. Box Inc. should venture in diversity because it will make it have a closer relationship with new markets. The work revealed that the most significant costs of running data centers go infrastructure and network, which therefore calls for the best approaches for minimizing them. Such is the rationale for discussing the methods of cutting down on the costs of operating the entire system. The costs also present an entire approach to management of resources, which will help the firm to minimize on losses. It is worthwhile noting the effects of underutilization of resources to the overhead costs of the business unit. Therefore, this work simply outlined the best approaches of managing the company’s resources in case it considers starting newer centers as a part of geo-diversity. Works Cited Greenberg, Albert., Hamilton, James., Maltz, David and Patel, Parveen. ‘The Cost of a Cloud: Research Problems in Data Center Networks.” ACM SIGCOMM Computer Communication Review 39.1 (2008): 68-73. Read More