Network Redesign Through Servers Consolidation Computer Science Essay

Published:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

In this paper, we have described a redesign methodology for an existing network in the educational system of 6-9 levels in the State of Kuwait, through servers' consolidation. Our detailed study about the existing network showed that the existing servers are under-utilized by more than 44%, thereby increasing the tasks execution delay and maintenance cost. Our redesign methodology operates offline by taking a snapshot of the existing network topology with its clients, servers and their interconnections, and translating it into a model to be manipulated for better performance with reduced number of servers. Our technique searches for the best possible way to consolidate the existing servers, while maintaining an acceptable performance and increasing the utilization of remaining servers. The experimental results show improvement in the utilization from 44.2% to 73.67% with the performance sustaining 100% for a network originally comprised of five servers serving 40 clients and consolidated to three servers.

General Terms

Management, Measurement, Performance, Design.

Keywords

Server Consolidation, Network Redesign, Server Utilization, Server Sprawl.

INTRODUCTION

The enterprises need to deploy information technology applications to support the expanding business processes. Often, this growth was achieved in an unplanned way. Many enterprises require new applications while upgrading the networking equipments (both hardware and software). Each time a new application was added; a new server accompanied by a storage needs to be installed. This has led to server and storage sprawl, including many underutilized servers with heterogeneous storage elements [3]. Server sprawl creates difficulty of managing such environment. If the average usages of many servers are low, then there is wastage of resources, and requires more staff to manage large number of heterogeneous servers, thereby increases the total maintenance cost of the network.

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.

iiWAS'11, December 14-16, 2009, Kuala Lumpur, Malaysia.

In our paper, we have proposed a redesign methodology for improving the performance of underutilized servers by consolidating them. Our redesign methodology computes the average CPU usage over a period of time for all the existing servers. We have implemented the server consolidation technique by redistributing the files and rerouting the traffic of lowest utilized servers to the remaining servers. Then, we re-evaluate the CPU usage of the remaining servers after increasing the workload and we continue the server consolidation process until all the existing servers are optimally utilized.

Our existing network scenario reflects the network in computer laboratory for 6-9 levels in the public education system in Kuwait. Each school has its own computer laboratory connected to the Internet through Ministry of Education. The computer laboratory includes 40-50 personal computers clustered into four sub-networks with five servers. In this paper, we have examined many scenarios, such as reducing the number of servers from five to one, two, three, or four servers through merging the workloads of two or more servers. Thus our main objective is to increase the utilization of the servers, simultaneously reducing the power utilization and maintenance cost.

RELATED WORK

Various research works were done in creating and maintaining virtual network, with an objective to enhance power savings through virtual server consolidation techniques [1][2][3][4]. Kokkinos et al [5] examined the data consolidation (DC) problem in grid networks. The DC was performed, when a task needed two or more pieces of data for its execution, possibly scattered throughout the grid network. The authors proposed a number of DC policies, which vary in choosing the data replicas and the DC site depending upon on time, traffic, or they were randomly chosen.

Mehta et al [6] presented a planning tool called ReCon to analyze the historical resource usage in large multi-cluster data center by monitoring data collected from an existing environment. Constraints specified in ReCon are essentially to restrict the space of possible mappings between VMs (Virtual Machines) and physical servers. The authors formulated the problem in an optimization framework wherein valid recommendations are generated while the total number of physical servers is minimized, and the constraints are satisfied.

Gupta et al [8] discussed the problem of server sprawl. The authors modeled the problem of server consolidation as a variant of the bin packing problem, where items to be packed are the servers being consolidated and bins are the target servers. They developed a new heuristic algorithm for determining the number of destination servers in the presence of the incompatibility constraints including bin-item incompatibilities. Bichler et al. [9] proposed a dynamic approach for virtualized systems. The solved the optimization problem exploiting multi-dimensional bin-packing approximate algorithms. This methodology does not take into account performance aspects, e.g. constraints on response times. Shim et al. [5] described an optimum way to relocate the server in a distributed network by considering the total communication delays between clients and servers.

Jerger et al [11] studied the behavior of server consolidation workloads, focusing particularly on sharing of caches across a variety of configurations. The authors presented a study of a variety of last level cache sharing arrangements to illuminate some of the pressures felt by the cache hierarchy. They presented a simulation methodology which is designed to mimic a dynamically partitioned system running a hypervisor or virtual machine.

Spellman et al [12] presented a new way to analyze consolidation alternatives using performance modeling and stepwise refinement. The paper talks about consolidation opportunities categories, which are: centralization, physical consolidation, and data and application integration. The authors discussed the stepwise refinement, which is a process that lets performance analysts address the performance modeling challenge, and provides the practical methodology for modeling and analysis to support the consolidation effort for each category.

Our work focuses on redesigning the existing network infrastructure by analyzing the various possibilities of removal of under- utilized servers.

AN OVERVIEW OF ENTERPRISE EDUCATIONAL NETWORK

A typical information technology environment, which is utilized in the educational system of 6-9 levels in the State of Kuwait, consists of 5 servers and 40 personal computers, as illustrated in Figure 1. The four servers are hosting ten numbers of personal computers each, and the fifth server connects all the servers together and is connected to the Internet. We considered each group of server and their associated clients as single cluster, having five clusters totally. The existing four servers are: mail server, file server, web server, and DNS sever. The main server and the clients are connected to these servers via a network tree topology.

Figure 1. Network architecture.

Here the environment is heterogeneous, where the physical servers may have different resource capacities, e.g., CPU, memory, etc. We analyzed all servers and our outcome had shown server sprawl situation, wherein four servers are underutilized. Also, these four servers consume more physical space and waste the resources such as electrical power.

THE PROPOSED SERVERS CONSOLIDATION PROBLEM

We have concentrated on redesigning the infrastructure through servers' consolidation by reducing the number of servers (through merging the workloads) and maintaining an acceptable performance level. We examined the changes in performance, CPU usage, and response delay before and after carrying out the consolidation.

The inputs to the servers' consolidation problem are the number of servers, number of clients bound to each server, server's capacities (servers processing speeds, hard disks retrieve rates, file allocation within servers, etc) and locations, and traffic (client-client traffic, server-server traffic, client-server traffic).The expected outputs after the server consolidation are: number of servers less than initial, files reallocations, new traffic distribution, and performance of the new data infrastructure.

4.1 Objective Function and Constraints

The main objective function is to maximize network performance, subject to the reduction in the number of the underutilized servers. However, this reduction would affect the performance of the remaining servers. We planned to reduce the number of servers (high utilization through high CPU's usage)-simultaneously maintaining the performance issue. Hence we examined the delay of the redesigned network after consolidation, to ensure an acceptable level of performance in comparison with the original network. We also verified that all clients in the network must bound to servers after the consolidation.

REDESIGNING SPACE FOR SERVERS CONSOLIDATION

Here we have examined the redesign space for consolidating the five servers from 5 to 1, 2, 3, and 4 servers respectively. We added a constraint that the central server should not be replaced. During the consolidation process, we maintained the central server along with other remaining ones. In the first scenario, the removal of four servers resulted in a single server. In the second case, we reduced five servers into two servers, this resulted in 32 scenarios. Remaining categories, we reduced the number of servers from 5 to 3 and 5 to 4, thereby creating 54 and 16 redesign scenarios respectively.

The variation in the redesign scenarios depended on the workload distribution of removed servers. When we consolidate (remove) more servers, then the number of possible redistribution combination would increases with the expectation that the remaining servers should perform the tasks of the removed servers.

CONSOLIDATION POLICY

Our consolidation process involves collecting various measurements on the servers to find out their utilizations, by computing the average CPU usage over a period of time (for example 1 hour). This includes the number of requests received by the server and time taken by the server in processing these requests.

We selected a server with low CPU usage, removed it physically and redistributed its traffic to another server. The redistribution choice depended on the server usage matrix. We re-evaluated the CPU usage of the remaining servers, after consolidation-keeping in mind to observe the performance of the servers (because of increased workload) to avoid high response time, with the constraint that every client was connected to one of the remaining servers. We repeated the previous process until it was left with a single (main) server.

DNS Server

The performance of DNS server depends on Equation 1, which comprises of the send request time (t1), search time (t2), and send response time (t3).

Total time = t1 + t2 + t3 (1)

Where t1 = packet size * network delay,

t2 = 0.5 * quarter number in the lookup table, and

t3 = response size * network delay.

The DNS server utilization is calculated by dividing the total time of processing all requests from the client over 3600 seconds (1 hour = 3600 seconds). The total CPU usage is low with the low utilization of 16%.

File Server

The performance of the file server depends on Equation 2, which consists of four parameters (t1, t2, t3, and t4).

Total time = send request time (t1)+ search time (t2) + copy file time (t3) + send response time (t4) (2)

Where t1 = packet size * network delay,

t2 = 0.5 * quarter number in the lookup table,

t3 = file size * server CPU rate, and

t4 = response size * network delay.

The file server utilization before consolidation is low and is about 40%.

Email Server

The performance of the email server depends on Equation 3, which includes send request time (t1), search time (t2), copy message time (t3), and send response time (t4).

Total time = t1 + t2 + t3 + t4 (3)

Where t1 = packet size * network delay,

t2 = 0.5 * quarter number in the lookup table,

t3 = message file size * server CPU rate, and

t4= response size * network delay.

The utilization of the file server is low before server consolidation and is about 38%.

Web Server

The utilization of the web server is formulated in Equation 4, which comprises of t1, t2, t3, and t4.

Total time = send request time (t1) + search time (t2) + copy page time (t3) + send response time(t4) (4)

Where t1 = packet size * network delay,

t2 = 0.5 * quarter number in the lookup table,

t3 = web page size * server CPU rate, and

t4 = response size * network delay.

The utilization of web server before the consolidation is about 33%, which is also found to be low.

Main server

The main (central) server is involved in the network communications when a client from one cluster communicates with any of the servers in the remaining clusters.

Server-server traffic shows the server-server traffic observed in one hour. We described the server workloads in terms of the percentage usage of CPU. For example, the requests from the DNS server to the web server represent approximately 11.25 % of the web server workload. The main server workload is the summation of the workloads, which are result of the requests from one server to all other servers. For example, the summation of all requests from the file server to the other servers (DNS, email, and web) is the workload of the main server results from the clients in the file server cluster to these servers, and it approximately represents 21.845% of the main server workload.

Server-server traffic shows that the main server is working up to 93.8% of the time. This high percentage is considered as high utilization, so we exclude the main server from the consolidation process, to avoid high delay time (low performance level).

RESULTS AND DISCUSSION

Removal of Four Servers

In this experiment, we removed four servers except the main server. The entire workloads of the removed servers (DNS, file, email, and web) were added to the main server and we estimated the new workload and delay of the main server. The CPU usage of the main server was increased to 127.15%, showing an over-utilized situation. In addition, this selection resulted in high delay time and low performance level, as shown in table 1 and figure 6.

Figure 6. Consolidation by removing 4 servers.

Removal of Three Servers

Here we removed three servers, and distributed their workloads to the remaining two servers. The main server and one of the remaining four servers were retained. Table 2 displays two of possible ways of this approach.

Table 1. Consolidation by removing four servers

Server

Status

Distribute to

CPU usage %

CPU Idle%

Delay (sec)

Delay (min)

DNS

Removed

Main

 

 

 

 

File

Removed

Main

 

 

 

 

Email

Removed

Main

 

 

 

 

Web

Removed

Main

 

 

 

 

Main

On

On

127.15%

0%

977.51

16.29

Table 2. Consolidation by removing three servers

Server

Status

Distribute to

CPU usage %

CPU Idle %

Delay (sec)

Delay (min)

DNS

Removed

File

 

 

 

 

File

On

On

127.15%

0%

977.51

16.29

Email

Removed

File

 

 

 

 

Web

Removed

File

 

 

 

 

Main

On

On

93.87%

6.13%

0

0

Server

Status

Distribute to

CPU usage %

CPU Idle %

Delay (sec)

Delay (min)

DNS

Removed

File

 

 

 

 

File

On

On

89.14%

10.86%

0

0

Email

Removed

Main

 

 

 

 

Web

Removed

File

 

 

 

 

Main

On

On

131.88%

0%

1147.78

19.13

According to first category in Table 2, the workloads of the removed servers (DNS, email, and Web) were distributed to the file server. This distribution resulted in overloading of file server showing a CPU usage of 127%. This caused an increase in the delay over 16 minutes. However, in the second case, the workload of the email server was redistributed to the main server. This improved the situation and the CPU usage decreased to 89.14%, which is an acceptable level.

Figure 7(a). Consolidation by removing 3 servers.

But the main server has a delay of 19.13 minutes. The delay timings approximately equal to zero shows that the timing are very low and difficult to measure. From this, we can interpret that the first case is better than the second one. But both the choices are not fair, because they decreased the performance level of one of the remaining servers. Figures 7(a) and 7(b) represent the status before and after consolidation by removing 3 servers using the two cases.

Figure 7(b). Consolidation by removing 3 servers.

Removal of Two Servers

Here, first we removed the DNS server, having lowest CPU usage and distributed its workload to the file server. Then we removed and distributed the workload of web server to the email server, which also showed lowest CPU usage. The results were summarized in Table 3 and Figure 8. From Figure 8, it is observed that the CPU utilization increases and the CPU idle time decreases to minimal value.

Figure 8. Consolidation by removing 2 servers.

Removal of one server

Here, we removed only one server having lowest workload, the DNS server, and distributed its workload to the web server. This resulted in four servers, with high performance levels, and low utilization, because of low workloads. The results were summarized in Table 4 and figure 9.

Table 3. Consolidation by removing two servers

Server

Status

Distribute to

CPU usage%

CPU Idle%

Delay (sec)

Delay (min)

DNS

Removed

File

 

 

 

 

File

On

On

56.00%

44.00%

~0

~0

Email

On

On

71.16%

28.84%

~0

~0

Web

Removed

Email

 

 

 

 

Main

On

On

93.87%

6.13%

~0

~0

Table 4. Consolidation by removing one server

Server

Status

Distribute to

CPU usage%

CPU Idle%

Delay (sec)

Delay (min)

DNS

Removed

Web

 

 

 

 

File

On

On

39.99%

60.01%

0

0

Email

On

On

38.02%

61.98%

0

0

Web

On

On

49.15%

50.85%

0

0

Main

On

On

93.87%

6.13%

0

0

Figure 9. Consolidation by removing 1 server.

CONCLUSION AND FUTURE WORK

We presented the results of network redesign in the schools (6-7 levels) in the State of Kuwait with underutilized servers. We focused on the servers' consolidation approach to solve the problem. We analyzed various possibilities while performing server consolidation and end up with good solution in removing two of the five servers. The results showed that after consolidation, the network consists of three highly utilized servers (utilization improved from 44.2% to 73.67%) with acceptable performance level.

Our present work can be improved by fully automating the server consolidation process and examining more scenarios for performing the redesigns concurrently in the network and the servers.

Writing Services

Essay Writing
Service

Find out how the very best essay writing service can help you accomplish more and achieve higher marks today.

Assignment Writing Service

From complicated assignments to tricky tasks, our experts can tackle virtually any question thrown at them.

Dissertation Writing Service

A dissertation (also known as a thesis or research project) is probably the most important piece of work for any student! From full dissertations to individual chapters, we’re on hand to support you.

Coursework Writing Service

Our expert qualified writers can help you get your coursework right first time, every time.

Dissertation Proposal Service

The first step to completing a dissertation is to create a proposal that talks about what you wish to do. Our experts can design suitable methodologies - perfect to help you get started with a dissertation.

Report Writing
Service

Reports for any audience. Perfectly structured, professionally written, and tailored to suit your exact requirements.

Essay Skeleton Answer Service

If you’re just looking for some help to get started on an essay, our outline service provides you with a perfect essay plan.

Marking & Proofreading Service

Not sure if your work is hitting the mark? Struggling to get feedback from your lecturer? Our premium marking service was created just for you - get the feedback you deserve now.

Exam Revision
Service

Exams can be one of the most stressful experiences you’ll ever have! Revision is key, and we’re here to help. With custom created revision notes and exam answers, you’ll never feel underprepared again.