dell cloud solution for web applications revolutionary cloud platform

8/2/2019 Dell Cloud Solution for Web Applications Revolutionary Cloud Platform

1/13

Dell Cloud Solutions for WebApplicationsA Revolutionary Platform for Public andPrivate Clouds

A Dell Technical White Paper


2/13

Dell Cloud Solution for Web Applications: A Revolutionary Platform for Public and Private Clouds

Page ii

THIS WHITE PAPER IS FOR INFORMATIONAL PURPOSES ONLY, AND MAY CONTAIN TYPOGRAPHICAL

ERRORS AND TECHNICAL INACCURACIES. THE CONTENT IS PROVIDED AS IS, WITHOUT EXPRESS OR

IMPLIED WARRANTIES OF ANY KIND.

2011 Dell Inc. All rights reserved. Reproduction of this material in any manner whatsoever without

the express written permission of Dell Inc. is strictly forbidden. For more information, contact Dell.

Dell, the DELL logo, the DELL badge, PowerConnect, and PowerEdge are trademarks of Dell Inc. UNIXis

a registered trademark of The Open Group in the United States and other countries. Other trademarks

and trade names may be used in this document to refer to either the entities claiming the marks and

names or their products. Dell Inc. disclaims any proprietary interest in trademarks and trade names

other than its own.

May 2011


3/13


Page 1

ContentsIntroduction ............................................................................................................. 2Not All Clouds Deliver the Same Service (.Nor Should They) ................................................. 2

Software as a Service ............................................................................................... 2Platform as a Service ................................................................................................ 2Infrastructure as a Service ......................................................................................... 3

Limitations of Common Cloud Infrastructures ..................................................................... 3Public, Private, and Hybrid Clouds .................................................................................. 4Dells Approach to Cloud Computing ................................................................................ 4Dell Cloud Solution for Web Applications .......................................................................... 5Solution Architecture .................................................................................................. 5Software Architecture ................................................................................................. 7

Resource Bursting .................................................................................................... 9Flexible Configuration .............................................................................................. 9Enhanced Application Performance .............................................................................. 9Integrated Security .................................................................................................. 9

Logical Network Configuration ....................................................................................... 9External VLANS ....................................................................................................... 9Internal VLANs ........................................................................................................ 9Administrative VLAN ................................................................................................. 9

Hardware Components ............................................................................................... 10C Series .............................................................................................................. 10R Series .............................................................................................................. 10

For More Information ................................................................................................ 11Notes .................................................................................................................... 11

Figures

Figure 1. Converging Approaches ................................................................................... 4Figure 2. Up to 15 Compute Nodes per Rack ...................................................................... 6Figure 3. Taxonomy of Dell Cloud Solution for Web Applications ............................................. 6Figure 4. Architecture of SmartMachine Virtualization ......................................................... 8Figure 5. Three Networks ........................................................................................... 10


4/13


Page 2

IntroductionIn the past three years, the term cloud computing has become increasingly elastic. Developers,

industry analysts, and customers have all stretched and morphed the definition of the term to

encompass a broad range of technologies and products. While an expanding market offers businesses

more choice, it also complicates the rational analysis of the underlying technologies. For this reason,

companies evaluating potential cloud computing infrastructures should take a pragmatic, business-minded approach in evaluating competing cloud computing infrastructures. Put succinctly, companies

need to ask, What does this technology do for my business?

Not All Clouds Deliver the Same Service (.Nor Should They)Cloud computing can entail a simple, web-based application available over the Internet or an array of

high-powered, massively provisioned data centers capable of processing thousands of online

transactions per second for millions of worldwide customers. The only constant is the cloud itself

centrally located computing resources available over the Internet or wide area network. Reviewing the

broad types of cloud infrastructures and their primary IT and business functions helps businesses more

fully understand this bewildering span of cloud computing products and their capabilities.

Software as a ServiceAmong products now defined as cloud computing, software as a service (SaaS) is the most targeted in

terms of functionality. Companies rent time and space from online applications as diverse as sales lead

tracking software to web-based email. Enterprise-level SaaS providers deliver a wide variety of

sophisticated applications, such as product lifecycle management and supply chain management, and

many other vertical applications.

If a business needs these specific applications, the SaaS model can save the company the expense of

buying hardware, software, and long-term maintenance. Drawbacks include the inability, in many

cases, to fully customize the solution for individual business requirements, complexities in integrating

the SaaS application with the existing business IT infrastructure, and difficulties in predicting and

budgeting pay-as-you-go pricing. For many companies, however, these targeted applications work well

to solve specific business problems and are not meant to replace or significantly augment corporate

computing resources.

Platform as a Service

When businesses need an application development environment, platform as a service (PaaS) is an

effective and efficient type of cloud computing resource to consider. Most importantly, companies can

produce new applications more quickly and with a greater degree of flexibility than with older

development platforms. Programmers and development managers especially appreciate that the cloud

provider handles all the care and maintenance of the underlying operating system, servers, storage,

and application containers.

PaaS is especially useful when development teams are widespread geographically or when partner

companies or divisions share development efforts. Engineers can more easily share and back up a

central repository of application data as well as implement tighter version control and environment

variables. PaaS cloud segments for development that painstakingly mirror the actual deployment

environment of the application ensure a more stable and polished finished product.

Companies can build up and tear down PaaS environments on demand with no capital expenditures or

long-term investment. One drawback to many PaaS implementations, however, is the tie-in to one

vendors platform and infrastructure. Customers need to ensure that the platform allows for maximum


5/13


Page 3

portability of applications and data. Overall, companies can save considerable capital and operating

expenses using a PaaS solution.

Infrastructure as a Service

The most foundational use of cloud computing is infrastructure as a service (IaaS): the rental of a

complete computing platform for running applications, hosting data, or housing a companys entire

computing environment. The latter use is not as common, and the majority of companies deployingIaaS now use it as a means to expand their computing capabilities in targeted IT areas without

drastically increasing capital expenditures on new hardware and software. In fact, to better serve this

market, most IaaS vendors emphasize their server and storage space as opposed to providing complete

data center or application services. This server rack and disk space rental model is a natural outgrowth

of the industry.

Some of the most prominent IaaS suppliers are e-commerce and Internet information businesses that

first entered the cloud computing market as a means to recoup revenue from excess, unused hardware

in their data centers. Server rack space is a useful commodity for businesses that require more

computing power but want to avoid long-term capital outlays. However, many companies are

discovering that rack space and virtual server rentals do not provide a complete solution, and that

reliance on ever-increasing cloud-based hardware resources produces many of the same IT

management issues as outright server and network hardware ownership.

Limitations of Common Cloud InfrastructuresCloud computing was initially conceived of as a way to allow businesses to unshackle themselves from

the challenges associated with managing hardware while delivering applications. Unfortunately, many

of the cloud computing companies have made technology decisions that do not eliminate those basic

constraintsthey simply virtualize the problem.

The most general issue is that hardware-centric cloud infrastructures lack focus on solving common

problems in software development. Many of the cloud computing companies have a surplus of hardware

due to their existing operations, and are simply leveraging that hardware inventory as a means toaccomplish additional flow of revenue. This is not their core business model, and they have not focused

their innovation on building solutions that make it easier and more efficient to deliver software.

Consumers of these services should make sure that the hardware, software, and network infrastructure

have been built and optimized for application delivery.

In some extreme instances, an inflexible set of hardware or virtualization parameters may determine

the type of software or software development that is possible on the infrastructure, further restricting

customer options and even locking in data to that specific architecture. Moves and changes to another

system, platform, or architecture might be costly or extremely complex.

One of the most common issues is that the infrastructures are not truly elastic in nature. They do not

have the ability to scale vertically by providing immediate access to resources beyond a subdividedphysical environment. Instead they manage scale with rapid provisioning of additional virtual servers,

using horizontal scale to provide a limited form of elasticity. Customers using this type of cloud

computing notice that adding more processes and users requires the manual addition of more virtual

CPUs and the startup of new instances of the application, tasks that their cloud vendors cannot

accomplish instantly. Performance suffers, but more importantly, the customer confronts the sprawl of

more and more added resources. While these newly added servers are virtual machines, they are


6/13


Page 4

nonetheless additional resources to pay for and manage. Ironically, physical CPUs may become heavily

utilized while running underutilized virtual machines.

Many cloud customers face continual issues of sizing and capacity planninghow much disk, CPU, and

RAM must they purchase? How will new software, more users, or increased application demands affect

the cloud installation?

Finally, because they use hardware simulation on a per-machine basis, many cloud computing

infrastructure providers are not able to take advantage of resource pooling in their billing. Without this

advantage, they effectively must pass through the underlying hardware costs and management

overhead in order to maintain profitability. The cloud providers should use resource pooling as a

mechanism for smoothing consumption, so that all complex infrastructure costs that are related but

disassociated from the application business model are not transferred directly to the software

provider. This is especially important for applications that receive bursts of traffic or that are public-

facing and subject to complex traffic load.

Public, Private, and Hybrid CloudsCloud computing offered as an IaaS solution is typically either public or private. Public clouds usually

entail a number of companies sharing the same cloud hardware at the same time, with reasonable

separation of data and network segments to preserve security. Private clouds are infrastructures

hosted entirely in privately owned company data centers. Virtual private clouds, which are becoming

increasingly popular, are cloud resources that a vendor hosts on dedicated hardware for a customer.

Hybrid clouds consist of some combination of public, private, and virtual private solutions.

Generally, public cloud infrastructures are the least expensive, but the lack of control and resource

sharing may make them less desirable for certain types of business computing. Private clouds, while

offering greater control, are a more expensive proposition, as the company must purchase, house, and

maintain all the hardware and software for its cloud. Many businesses find virtual private clouds the

best solution. In this scenario, the cloud host maintains the hardware and software, while dedicated

hosting and maintenance ensure greater security and segmentation.

The DellApproach to Cloud ComputingCloud computing is a business model driven by the need for more efficient use of IT resources on a pay-

as-you-go model. Recent progress in two critical technologiesnetwork services and virtualized

infrastructurehas increased server utilization multifold. Impending refinements to cloud technologies

should accelerate this trend to drive significantly higher ROI from IT infrastructure investment.

Figure 1. Converging Approaches


7/13


Page 5

Dells experience working with both hyperscale and enterprise customers shows us that there are

multiple right ways to create cloud solutions. Hyperscale cloud providers have created revolutionary

application architectures that leverage network services to distribute their applications over thousands

of servers in parallel. Enterprise customers avoid disruptive changes to their applications by evolving to

use more virtualized infrastructure. These approaches are converging as we introduce new cloud

applications and operational solutions that leverage the best of both approaches.

Regardless of their approach, customers are seeking the same core values from their cloud initiatives.

They want to consolidate resources (infrastructure, multi-application), improve efficiency (resource

utilization, density), respond quickly to changing needs (elastic, pay-for-use) and reduce administrative

burden (standardized and automated). On the surface, these values appear to be a return to

mainframe computing. Looking deeper, cloud architecture breaks single system and vertical scalability

limits. Rather than timesharing the resources of a single massive supercomputer, cloud applications

decompose into distributed units across multiple systems, effectively creating resource pools of

commodity components, such as a RAID implementation for generalized computing.

The technologies underlying a cloud are complex, but the core components are simple: applications,

networked services, and virtualized infrastructure. These three components, in varying proportions

garnished with management APIs, form the basis for all cloud solutions.

Dells job as a solution provider is to help match proven core cloud technologies with customer

business needs. To accomplish this, we have built and are constantly expanding our cloud tool box.

Clouds rapid rate of change makes locking into a proprietary solution especially risky. Dells core

values drive us to invest in technologies that are open, capable, and affordable.

Dell Cloud Solution for Web ApplicationsDell Cloud Solution for Web Applications (DCSWA) provides a revolutionary cloud platform that supports

web-based applications throughout their life cycles. The platform is a combination of Joyent Smart

Datacenter cloud software optimized on Dell hardware and supported end-to-end by Dell services. It

provides a scalable hardware architecture managed by the Joyent Smart Datacenter. An applicationdeveloper can scale for both performance and redundancy by scaling out the application to multiple

nodes in the cloud.

Solution ArchitectureThe solution comes with prescriptive reference architecture (RA) that defines the design and minimal

requirements around the networking infrastructure. It defines a cloud as a collection of interconnected

pods, which are a collection of up to 12 racks. Each rack contains up to 15 compute nodes to handle

web applications. Each pod has a provisioning server (PS) to manage compute nodes. The Joyent

software management layer resides within one of the pods provisioning servers, and the software

agents reside within each compute node. See Figure 2.


8/13


Page 6

Figure 2. Up to 15 Compute Nodes per Rack

Figure 3. Taxonomy of Dell Cloud Solution for Web Applications


9/13


Page 7

Within the infrastructure reside the Joyent SmartMachines: end-user virtual machine applications built

on the Joyent Unix operating system virtualization model. Within these SmartMachines, the customer

runs the components of their web applications. The SmartMachines are connected via Zeus Load

Balancer, which automatically balances load and enables the scaling of the web application across

multiple SmartMachines. When considering a traditional multi-tiered web application, load balancers

can be inserted between all three tiers (web, applications, and database layers) to provide

performance and redundancy.

The solution supports open source tools and development libraries and comes loaded with basic

SmartMachine templates, such as Zeus Load Balancer and MySQL, to facilitate rapid deployment of

database components for applications. Administrators have the flexibility to create additional

customized templates based on business requirements. This set of tools represents the basis of the

PaaS environment. See Figure 3.

The solution allows administrators to carry out routine system management tasks from a cloud control

administrative portal. It allows delegation of basic provisioning activities to end-users through a

dedicated self-service end-user portal. The management component provides operating, reporting,

monitoring, and diagnostic functions at both the cloud administrator and end-user levels for their

respective domains. Administrators can also identify underutilized resources, reassign them to optimize

output, and create utilization reports at overall system and individual components levels.The solutiondefines a backup methodology for end-user datasets. And it provides security features and

methodologies that ensure privacy and security for end-users assets in either a multi-tenant public

cloud hosted by a Telco/IT service provider or a private cloud hosted by an enterprises IT

organization.

Software ArchitectureIn a traditional network, the developer must manage the entire server environment, including all the

hardware (CPU, storage, RAM) as well as the operating system. If a system needs to scale horizontally,

the operator purchases more hardware. Networks accomplish vertical scaling by over provisioning

installing enough hardware to accommodate worst-case, high-traffic situations, leaving the excesshardware resources idle most of the time. These practices result in server sprawl, wasted resources,

and complex management. And being unprepared for load spikes can result in costly lost opportunities.

By contrast, standard virtualization based solutions allow the abstraction of the hardware and the

sharing of its resources among multiple virtual machines. If managed correctly, this can ensure higher

utilization of hardware resources. But it cannot provide instantaneous vertical scaling because

resources are still provided in fixed sizes to the virtual machine, and the developer must still manage

multiple machine instances with over-provisioning.


10/13


Page 8

Figure 4. Architecture of SmartMachine Virtualization

A complicating factor in this scenario: additional time is required to boot up a virtual machine,

resulting in an ongoing need for predictive over-provisioning to maintain performance underunexpected loads. Also, with resources bucketed to each virtual machine, idle VM resources cannot be

leveraged by running VMs. A final disadvantage to virtual machine administration is that developers

must also manage the guest operating system, creating another layer of complexity in application

production and stability.

The Joyent SmartMachine differs from traditional and virtual machine architectures. Primarily, it goes

further in abstraction of the hardware by presenting the hosted application with access to a pool of

combined resources, rather than control of a fixed resource. The SmartMachine is designed to be

transparent to the underlying operating system, Joyent SmartOSall part of Joyent SmartDataCenter

design. SmartOS uses this transparency to provide all SmartMachines with as-needed access to a large

pool of available resources, while still providing each SmartMachine with minimum guaranteed access

to resources based on a pre-established fair-share schedule.1 The transparency also allows Joyent

SmartOS to automatically identify underutilized resources and repurpose them to provide enhanced

performance. The RAM and CPU resources are either used directly used by applications, or used by the

operating system to optimize disk I/O and provide other performance enhancements to the

SmartMachines.2

The SmartMachines lightweight operating system virtualization means that a highly tuned operating

system (Joyent SmartOS) is available within the SmartMachine, reducing the need for the developer to

manage multiple guest operating system images. When using a SmartMachine, the developer writes the

application and manages only the application logic to enable horizontal scaling. This simplifies

management compared to traditional hardware or virtual machine based architectures, where the

developer must also create and manage multiple abstractions of physical servers with operatingsystems and other software.

In addition, pooled hardware allows for easy and dynamic vertical scaling. Developers can concentrate

on application production rather than abstraction layers, operating system management, or virtual

machine resources. The simplicity of this solution greatly reduces management complexity for system

administrators as well. Now administrators can easily create and manage accounts, monitor

performance and resource utilization, and troubleshootall from one central console.

App

Code

App

Code

App

Code

App

Code

Joyent SmartOS

Resource PoolPhysical Hardware

Operating System +Virtualization Layer

SmartMachines


11/13


Page 9

Because the SmartMachine is built on Joyent SmartOS, it incorporates the following core features.

Resource Bursting

SmartMachines have the ability to tap the expanded pool of CPU, RAM, and network bandwidth through

the Joyent SmartOS, providing needed capacity during bursts of activity or usage. Joyent SmartOS also

supports more CPU cores and a larger memory footprint than other virtualization operating systems, so

the resource pools grow larger as underlying hardware is improved.

Flexible Configuration

SmartMachines come preloaded with a variety of popular software environments, including a vanilla

Joyent Unix installation (SmartOS), development platforms (PHP, Java, Ruby, Rails), databases (MySQL,

Oracle), web servers (Apache), load balancers (Zeus), and more. Developers are not forced into any

specific development platform and readily get the tools they want for application development.

Enhanced Application Performance

Joyent SmartOS provides a system-wide RAM cache for all disk I/O, which significantly improves read

and write performance for applications. Within a resource pool, any memory and CPU not directly

utilized for application business logic is dedicated to providing improved performance. This approach of

drawing on idle resources and providing them directly to applications is unique to Joyents SmartTechnologies.

Integrated Security

The SmartOS operating system adds additional security and stability to the SmartMachine application

environment. Per-machine administration is performed as root user with no ability to manipulate the

base operating system kernel. SmartOS isolates memory, network, and processor by SmartMachine, and

blocks network reconfiguration and traffic sniffing, which keeps even public cloud applications

completely segregated from one another.

Logical Network ConfigurationThe solution relies heavily on virtual local area network (VLAN) technologies to separate the different

networks. There are three networks configured by default and all exist in all configurations.

External VLANS

These VLANs are typically addressed with Internet-routable addresses and are used for connections to

devices that are external to the cloud infrastructure. These include externally visible services such as

load balancers and web servers. There can be one or more of these networks, depending on the need

to segregate traffic among groups of SmartMachines.

Internal VLANs

These VLANs are typically addressed with private addresses and are used for communications between

SmartMachines. SmartMachines that perform front-end and back-end functions should be configured to

communicate via these VLANs. There can be one or more of these networks, depending on the need to

segregate traffic among groups of SmartMachines.

Administrative VLAN

This VLAN is used for administrative functions such as cloud control, KVM, IPMI/BMC, system logs,

backups, and other monitoring or administrative processes. SmartMachines do not have access to this

network. There is only one VLAN set up for this function and it is spanned across the entire data

center.


12/13


Page 10

Figure 5. Three Networks

Hardware ComponentsThe Dell prescribed reference architecture draws from both the Dell PowerEdge R and C series

hardware families. Specifically, there are two hardware sets for the solution:

C Series

Infrastructure Server: PowerEdge C1100A data center/cloud-optimized dual-socket Intel Xeon

server with six 2.5-inch HDDs, two LOMs, and one PCIe slot in 1U form factor.

Compute Nodes: PowerEdge C2100A data center/cloud-optimized dual-socket Intel Xeon server with

six 3.5-inch HDDs upgradable to twelve 3.5-inch HDDs, two LOMs, and two PCIe slots in 2U form factor.

The solution requires one additional dual-port PRO1000 Intel PCIe NIC and one LSI 9260-8i RAID

controller.

R Series

Infrastructure Server: PowerEdge R610An enterprise-class dual-socket Intel Xeon server with six

HDDs in a 1U form factor.

Compute Nodes: PowerEdge R710An enterprise-class dual-socket Intel Xeon server with eight 2.5-

inch HDDs, four LOMs, and four PCIe slots in 2U form factor.


13/13


Page 11

Each pod requires a network infrastructure. The size and expected growth of the pod determines the

size and complexity of the network infrastructure. The hardware components used are:

PowerConnect 6248

PowerConnect 6224

Arista Model 7124

Servers and networking are racked in the following Dell PowerEdge rack enclosures:

Dell 42U Rack (4220)

Dell 24U Rack (2420)

Within the rack, unmanaged power distribution units (PDUs) provide power to the hardware

components. The number of PDUs is determined by the size and power redundancy requests from the

customer.

Overall, the solution offers a turnkey integrated cloud solution that is optimized for performance and

built with scalability in mind.3

For More InformationTo learn more about the Dell Cloud Solution for Web Applications, contact your Dell representative or

visitwww.dell.com/dcswa.

Notes1) DCSWA SmartMachines can burst dynamically up to 32x CPU capacity* to support unexpected

traffic spikes. With low overhead, DCSWA is designed for average 70% utilization with > 95%

memory resource availability to support sudden load spikes.

* Configuration dependent. On a 16 vCPU PEC 2100 server, with 32 similar machines, each

machine can burst 32x if other machines are idle. Bursting takes place in real time

(sub-second) and does not require any manual intervention or third party software.

2) DCSWA delivers 14x faster disk I/O performance, 3x faster Memory I/O, 5x faster CPU and can run

3x more database requests per second on MySQL database as compared to Amazon EC2 machines

in a test conducted by public cloud benchmarking:www.joyentcloud.com/resources/benchmarks

3) Customer quote: The Dell/Joyent cloud solution gives us credibility. Its established, its proven,

its reliable, its scalable. It has all those things covered. Mike Schmidt, President and CEO

Uniserve.
http://www.dell.com/dcswahttp://www.dell.com/dcswahttp://www.dell.com/dcswahttp://www.joyentcloud.com/resources/benchmarkshttp://www.joyentcloud.com/resources/benchmarkshttp://www.joyentcloud.com/resources/benchmarkshttp://www.joyentcloud.com/resources/benchmarkshttp://www.dell.com/dcswa

dell cloud solution for web applications revolutionary cloud platform

Documents