R E P A P E T I H W
STREAMLINING SERVER DEPLOYMENT AND MANAGEM ENT IN THE D ATA CENT ER
Brocade Tapestry Tapestry Application Resource Manager (ARM) provides a wide range o operational advantages or enterprise environments.
As IT organizations continue to extend their server investments throughout the data center, they have encountered additional management costs due to the complexity o deploying and maintaining those servers. To help them overcome this challenge, Brocade® Tapestry™ Application Resource Manager (ARM) streamlines server deployment and management by leveraging existing Storage Area Network (SAN) environments. Compared to alternative methods, Tapestry ARM boosts productivity by automating common tasks and an d reduces costs by improving productivity, productivity, consistency, and resource utilization. This paper describes some o the key Tapestry ARM capabilities and processes, including usage scenarios and typical deployment environments.
THE EVOLVING SERVER MANAGEMENT CHALLENGE
As today’s IT organizations address an ever-growing range o business and regulatory requirements, they are redening their technology inrastructures. One o the most common trends in the last ew years has been the increased deployment deployment o x86-based rack-mount or blade servers in the data center. Several actors have driven this trend, such as the transition rom proprietary proprietary RISC-based servers, the need or increased server densities, and the continued drop in acquisition costs or more powerul servers. Unortunately, volume deployments o x86-based servers have also caused additional data center complexity and higher management costs, especially the human cost o management due to the sheer number o servers deployed. In addition, nding cost-eective ways to keep these servers highly available continues to be a signicant challenge. Fortunately, this management challenge coincides with the continued growth in Storage Area Network (SAN) deployments. By merging the parallel growth trends o x86-based servers and SANs, IT organizations can utilize the unique value proposition o their SAN environments and extend a wide range o benets to their data center servers. In act, the ability to deploy hundreds or thousands o x86 servers at an attractive price/perormance ratio in a SAN environment can signicantly help these organizations maximize the value o their server investments. AN INNOVA INNOVATIVE NEW SOLUTION SOLUTION FOR RAPID SERVER DEPLOYMENT
In the past, most traditional server deployment tools have taken a piecemeal view o management by associating specic servers with their sotware and storage counterparts in a tightly coupled manner. However, by reducing the physical system dependencies and improving the mobility o the entire application resource set (server/application/ storage/data), IT organizations are better able to ocus on complete business solutions rather than low-level system details.
One o the rst tools to enable this type o capability is Brocade Tapestry Application Resource Manager (ARM), which utilizes the fexibility o the SAN to provide a wide range o business benets (see Table 1). This solution is composed o two key components:
R E P A P E T I H W
apestry y ARM Appliance: A hardware device based on the Brocade SilkWorm® • Tapestr Fabric Application Platorm
• Tapestry ARM Service Processor: Sotware running on an x86-based server
Higher productivity
• Automates repetitive tasks to minimize administrative time and eort • Provides easy point-and-click actions actions
Table 1. Key Tapestry ARM business benefts
• Utilizes template-based template-based operations Increased control
• Automates common server deployment and management tasks tasks • Provides a bound relationship or key elements in application server environments • Masks storage complexity complexity
Faster ROI
• Minimizes human touch to help eliminate eliminate errors • Reduces deployment time to improve data availability availability • Optimizes resource utilization utilization
Greater fexibility
• Extends IT inrastructure agility agility • Enables more efcient server liecycle management • Leverages the consolidation consolidation benefts o the SAN
At the center o Tapestry ARM is its ability to reduce complexity and provide a unique view o the complete server, operating system, and storage relationship by integrating management tasks or all o these thes e elements. ele ments. To accomplish accom plish this, Tapestr apestry y ARM manages these elements as workload that can be redeployed and quickly replicated across the data center via a SAN. The innovative use o the SAN or management and deployment can provide enormous benets or organizations that are deploying new blade or high-density 1U server environments, anning out their server resources, or scaling down in shared storage environments. TAPESTRY ARM WITHIN THE DATA CENTER
Figure 1 shows a typical data center with a SAN that integrates the Tapestry ARM Service Processor (x86-based (x86-bas ed sotware) so tware) and the Tapestry Tapestry ARM Appliance. By leveraging the capabilities o the SAN, Tapestry ARM can automate key management unctions and streamline procedures such as server reconguration down to minutes (rather than hours). This is in stark contrast to today’s time-consuming manual processes or tasks such as physical server reconguration reconguration and Logical Unit Number (LUN) mapping.
Figure 1. A typical Tapestry ARM data center confguration
Tapestry ARM Service Processor
Servers
SAN
Tapestry ARM Appliance
Storage
Server Instances
To minimize the manual eort and time required to complete common system-related activities, Tapestry ARM provides a ully automated graphical user interace, easy point-and-click actions, and template-based operations or managing the complete server environment. Tapestry ARM manages the server/operating system/storage data as a single entity represented as a server ser ver instance on the SAN. As a result, a given application can be independent o the physical server that hosts it. The ecient management and deployment o this “server “server instance” is the key objective objective o Tapestry ARM, and this simplied approach can reduce previously complex and time-consuming tasks to a matter o minutes. In act, Tapestry ARM can automatically deploy server instances on multiple systems, as business needs require. The fexibility and transparency o system deployment enables administrators to ocus on key data center issues rather than on the details o managing individual hardware hardware and sotware relationships. To ully understand these concepts, it helps to identiy some new terms: • Processing element: The traditional hardware server system • Server instance: The entire sotware sotware entity that runs r uns on a processing element • Golden image: The template or a server instance stored on the SAN or uture deployment Figure 2 shows an example o Tapestry ARM deployment with these items in a typical SAN environment: • Processing elements: x86-based servers • Server instance #1: Web server on Red Hat Enterprise Linux and related storage inormation • Server instance #2: Database on Windows Server 2003 and related storage inormation • Golden image #1: Red Hat Enterprise Linux image • Golden image #2: Windows Server 2003 image These components play a key role in Tapestry ARM usage scenarios such as rapid server ailover and ast hardw hardware are upgrades or x86 servers in Linux and Windows Windows environments. environments.
Figure 2. Tapestry ARM, processing elements, and server instances in a SAN environment
Tapestry ARM Service Processor
Tapestry ARM Appliance
Processing Elements
SAN fabric
R E P A P E T I H W
Linux or Windows Windows or Linux Red Hat and Windows golden images #1 and #2
Red Hat Enterprise Linux (server instance #1) Windows Server 2003 (server instance #2)
Rapid Server Failover
Tapestry ARM provides a ast, yet inexpensive high-availability ailover capability suitable or many types o applications. applic ations. For instance, it can take idle or less critical c ritical systems ofine and rapidly redeploy them to address ailovers or accommodate workload spikes. Ater the workload spike subsides, the redeployed systems can be restored to their original states. Compared to the hours required or rebuilding and restoring a conventional server with local disk storage, Tapestry ARM can provide substantial timesavings and higher application applic ation availability availabil ity.. The key dierentiator di erentiator or Tapestr apestry y ARM is that it deploys dep loys server ser ver instances by utilizing the golden images on the SAN. With a rack-mount x86 server independent rom its storage, the SAN storage can be dynamically recongured— enabling either the Linux-based Web server ser ver or the Windows-based Windows-based database (in Figure 2) to boot on the rack-mount x86 server. No urther manual conguration is needed. Because Tapestry ARM can rapidly deploy either type o server instance, IT organizations can support key scenarios such as the rapid manual ailover o a deective server (see Figure 3). Ater detecting an immediate ailure state, a system administrator can use Tapestry ARM to take the deective server ofine and reboot the SAN-resident server instance on a dierent server. The total time rom ailure detection to reboot on a new server is typically 10 minutes or less. This type o rapid server deployment can especially benet organizations that want higher availability levels without resorting to clustered systems and their associated cost and complexity.
Processing Elements
Linux
Linux
Windows Failure: Reboot new server
Shut Down
Idle
Windows
Figure 3. Rebooting a server ater a ailover
Rapid Hardware Upgrades
Another possible Tapestry ARM usage scenario is a le/print server ser ver that requires an ofine o fine hardware h ardware upgrade. upg rade. With Tapestr apestry y ARM’s ARM’s rapid ra pid deployment d eployment capabil capabilities, ities, a system administrator can simply restart the le/print services on another server within 10 minutes or less. This minimizes disruption during the hardware upgrade (which can sometimes take hours). Ater the hardware upgrade is complete, Tapestry ARM can restart the le/print services on the newly upgraded server (see Figure 4). This approach increases the availability availability o le/print services during normal system upgrade activities.
Figure 4. Rebooting a server ater a hardware upgrade
Processing Elements
Upgrade Windows hardware: reboot to new server
Shut down and upgrade
Idle
Windows
Idle
Windows
Upgrade complete: reboot original server
Idle
ADDITIONALL USES FOR TAPESTRY ADDITIONA TAPESTRY ARM
Key Advantages over IP Network-based Deployment Tapestry ARM provides signifcant advantages over traditional IP network-based server deployment, which is characterized by: • Slow network response times due to local operating system disk images being transerred over LANs • Running out o local disk storage (Tapestry ARM can dynamically expand its SAN storage) • Limited system utilization with disk images tied to specifc platorms
Tapestry ARM has many additional practical uses, ranging rom basic license management to server startup and shutdown. Key eatures include the capability to discover and add new systems to the hardware pool. This provides the ability to create golden images or fexible deployment o server sotware environments independent o a specic hardware server platorm. The golden image is the total sotware environment captured as a single entity and stored on the SAN storage. When Tapestry ARM is ready to discover hardware on the network and prepare the necessary server instances or uture deployment, deployment, it utilizes an inter nal database that tracks system conguration details such as hardware discovery discovery and ava available ilable SAN storage. The Tapestry ARM Appliance, based on the SilkWorm Fabric Application Platorm, manages the total available SAN storage. This device provides the ability to virtualize the initiators to the storage arrays and the targets or the host ser vers. Storage is simply presented as a collection o LUNs regardless o the brand or model o the storage arrays. As a result, Tapestry ARM helps mask the complexity o the various storage arrays by presenting a single pool o av available ailable storage to the host servers. When new servers (processing elements) join the network, they advertise their availability by sending a Preboot eXecution Environment (PXE) request to Tapestry ARM, which responds by sending a discovery network boot agent. This agent is temporarily installed to discover the available hardware resources, such as type o Host Bus Adapter (HBA), system identication, network network ports, por ts, and so on. The agent also congures c ongures HBAs to ensure that they are available to connect to the SAN. However, the agent does not persist during runtime. The discovered discovered inormation inor mation is then transmitted back to the Tapestry Tapestry Arm Service Processor (an x86 system running Tapestry ARM sotware), which then records the inormation in an SQL database or uture use in creating and deploying server instances. Common usage areas or this type o capability are golden image creation, virtual machines, and multipathing environments—as described below.
Golden Image Creation or Increased Resource Utilization
R E P A P E T I H W
Tapestry ARM has an initial scripted installation process or Linux or Windows based on vendor-provided vendor-p rovided Linux or Windows media. Ater the Linux or Windows Windows operating ope rating system is installed, the system can set additional application and conguration parameters. The nal image can then be captured as a golden image and used to propagate similar servers in the data center. With the elimination o individual system dependencies, the golden image can be used to create a server instance that can be deployed within minutes on similarly congured systems. The resulting deployment creates a server instance that consists o an operating system, application, and storage data.Tapestry ARM automatically manages this relationship. Using Tapestry ARM, administrators can select a golden image (Linux or Windows), a processing element (any x86-based rack-mount or blade server), and the required storage (available LUNs on the SAN)—conguring them into a single entity. Ater the golden image is stored, it can be used to create a server instance by adding the “personality inormation” that can be deployed to any server o compatible base hardware conguration (see Figure 5). Because a server and its workload/operating environment are no longer tied together physically or geographically, IT organizations have much more fexibility in how they utilize their server resources. Virtual Machines and Multipathing Multipathing
Tapestry ARM provides the same type o conguration fexibility or IT organizations that utilize Microsot Virtual Server Ser ver 2005. Supported Supported capabilities include: • Treating a virtual server as a physical server in the virtual server host space • Moving a virtual virtual server running on a virtual server host to another virtual server host • Booting a virtual server host rom various server instances on the SAN For example, the virtual server runs on a host that in turn runs on a processing element. In this way, the virtual server can be thought o as a server instance, and the virtual server host as a virtual processing element (see Figure 5). The entire sotware stack runs on top o an actual processing element.
Microsoft Virtual Server #1 (VS1) Microsoft Virtual Server Host #1 (VSH1) Processing Element #1
Microsoft Virtual Server #2 (VS1)
Figure 5. Virtual servers and virtual server hosts
Microsoft Virtual Server Host #2 (VSH1)
Processing Element #2 VSH1 VS1
Mobility across hardware platforms VSH2 VS2
The benet o this approach is that a virtual server host can move with the server instance to provide even greater fexibility fexibili ty.. As a result, virtual servers can move between virtual server hosts located on dierent processing elements—enabling elements—enabling organizations to ne-tune their deployments or optimal rollout and system utilization. For example, this increased fexibility could be used or improved improved application load balancing, workload isolation, or other purposes. The ability to move a virtual server across virtual ser ver hosts ollows the same concept as detaching or attaching a server instance on a processing element (see Figure 6). A virtual server ser ver can move move rom one virtual host server (on processing element #1) to another virtual host server on a dierent processing element (#2). This provides even greater mobility options or virtual servers.
Figure 6. Moving virtual servers across virtual server hosts
Microsoft Virtual Server Host #1 (VSH1) Microsoft Virtual Server Host #2 (VSH1) Processing Element #1
Microsoft Virtual Server #2 (VS1) Microsoft Virtual Server #1 (VS1) Processing Element #2
VSH1 VS1
Mobility across hardware platforms VSH2 VS2
Because hardware processing elements are independent o the server instance, the virtual servers act like processing elements, since they are independent o the virtual server hosts. This provides virtual servers with similar benets to server instances— with mobility and agility extended across the virtual space. Tapestry ARM also supports Multipath I/O (MPIO) or LUNs in specic disk array/HBA combinations. The advantage o MPIO is the ability to provide redundancy i one o the data paths goes ofine or ails. Using two Tapestry ARM appliances signicantly increases the reliability o the overall system. For instance, i a data path is taken ofine, the redundant path can enable operations to continue or mission-critical applications, increasing overall system availability. DEPLOYING TAPESTRY ARM IN EXISTING SANS
Although most x86 server deployments traditionally have used local direct-attached storage, the lower costs and complexity o SAN implementation should alter that approach. In turn, Tapestry ARM can leverage the fexibility o these SANs to streamline server deployment and management. Although Tapestry ARM can be deploy dep loyed ed in a variety o environments, the the greatest g reatest advantages advanta ges tend to be in enterprise data centers with high SAN attachment rates. Table 2 shows some typical deployment environments and the key requirements that Tapestry ARM addresses.
INDUSTRY
KEY REQUIREMENTS
Oil and gas exploration
• High perormance and bandwidth • Signifcant data analysis analysis • Periodic system upgrades to the latest technologies technologies
R&D/engineering development
• Signifcant prototyping prototyping • Extensive testing matrix, including including x testing
Table 2. Typical Tapestry ARM deployment environments
R E P A P E T I H W
• One-o-a-kind systems Financial services
• High availability • High perormance and bandwidth bandwidth • Extensive qualifcation cycle and data analysis
Electronic design automation
• High perormance and bandwidth bandwidth • Large compute and storage storage arms • Extensive design verifcation verifcation
FOR MORE INFORMATION
As the advantages o cost-eective storage networking continue to grow g row, Tapestry ARM provides unique value or server deployment and management. The ability to rapidly deploy servers across the SAN increases operational eciency, reduces complexity, and maximizes resource utilization in a wide range o data center environments. For more inormation, contact an authorized Brocade sales partner or visit www.brocade.com.
Corporate Headquarters San Jose, CA USA T: (408) 333-8000 in
[email protected]
European and Latin American Headquarters Geneva, Switzerland T: +41 22 799 56 40 emea-in
[email protected]
Asia Pacifc Headquarters Singapore T: +65-6538-4700 apac-in
[email protected]
© 2006 Brocade Communications Systems, Inc.All Rights Reserved. 02/06 GA-WP-777-00 Brocade, the Brocade B weave logo, Fabric OS, Secure Fabric OS, and SilkWorm SilkWorm are reg istered trademarks and Tapestry is a trademark o Brocade Communications Systems, Inc., in the United States and/or in other countries. All other brands, products, or service names are or may be trademarks or service marks o, and are used to identiy, products or services o their respective owners. Notice: This document is or inormational purposes only and does not set orth any warranty, expressed or implied, concerning any equipment, equipment eature, or service oered or to be oered by Brocade. Brocade reserves the right to make changes to this document at any time, without notice, and assumes no responsibility or its use. This inormational document describes eatures that may not be currently available. Contact a Brocade sales oce or inormation on eature and product availability. Export o technical data contained in this document may require an export license rom the United States Government.