Download - Insider's Guide- Building a Virtualized Storage Service

Transcript

1

STORAGE VIRTUALIZATION: AN INSIDER’S GUIDE

Jon William Toigo CEO Toigo Partners International

Chairman Data Management Institute

Copyright © 2013 by the Data Management Institute LLC. All Rights Reserved. Trademarks and tradenames for products

discussed in this document are the property of their respective owners. Opinions expressed here are those of the author.

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 2

STORAGE VIRTUALIZATION:

AN INSIDER’S GUIDE

Part 6: Building a Virtualized Storage Service

The dream of every rock star is to create timeless music that garners a

broad audience spanning many generations of fans and evokes wild-eyed

appeals for endless encores. That is the definition of star quality.

Even the most modest IT administrators and planners would like for their

data storage strategies to achieve a kind of timelessness. They want

storage to be more manageable and predictable in terms of cost and

performance – and even power consumption – while delivering

measurable efficiency improvements in every category that matters to

the business they serve. Ultimately, they want to be technology leaders

who deliver improved service levels while cutting costs of operation.

Storage virtualization is a technology that can enable this kind of

timelessness.

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 3

STORAGE VIRTUALIZATION: AN INSIDER’S GUIDE

Building a Virtualized Storage Service

STORAGE AS A SERVICE

In previous sections, we have talked about storage virtualization as a key enabler of storage as a

dependable service to the organization. Storage virtualization delivers improved capacity

management, performance management, data protection management -- and ultimately data

management – by establishing a software layer – sometimes called a storage hypervisor –

above the hardware infrastructure itself. In so doing, storage virtualization translates complex

storage hardware and plumbing into a centrally managed resource that can be allocated

dynamically to business applications and processes.

Implementing a storage hypervisor can have a significant impact in bending the cost curve of

storage hardware infrastructure and the large cadre of staff required to administer and

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 4

maintain it. This is important, especially given the capital and operating expense of storage

today.

Storage virtualization directly addresses the problem of storage infrastructure management

inefficiency by enabling fewer administrators to manage more capacity. A storage hypervisor

addresses high storage product costs by enabling value-add functions to be shared across all

storage rather than being isolated to specific arrays, where they tend to increase the price tag

on each product.

Bottom line: the storage hypervisor makes a simple adjustment to storage infrastructure that

makes it more manageable in operation and less costly to own. Just implementing a storage

hypervisor helps to rectify years of inefficient storage infrastructure design – when we bought

storage arrays with isolated on-array functionality that created massive difficulties in coherent

management and drove up operational costs – and moves us a lot closer to the holy grail of

an efficient and cost-effective storage infrastructure. Moreover, this strategy enables us –

To keep gear in service longer.

To purchase less expensive gear with no sacrifice in terms of either performance or

functionality: services like thin provisioning are extensible to all gear regardless of the

vendor name on the front of the box.

To manage and to administer more capacity with fewer folks.

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 5

To map resources to applications and machines, whether virtual or physical,

effortlessly and to balance workload across gear with minimal intervention.

And to perform machine maintenance with no downtime at all.

In addition to these important operational and financial efficiencies, it should be emphasized

that adding a good storage hypervisor contributes greatly to the safety and protection of

irreplaceable data assets. It simplifies storage and provides flexibility in assigning the right data

protection services to the right data. It reduces the risk of outages due to value-add software

failures on individual array controllers. It enables mirroring and replication without the

constraint of hardware lock-ins that allow replication only between two rigs from the same

vendor. Finally, using a storage hypervisor lets us protect data simply and efficiently without

buying and integrating multiple hardware and software data protection products. That goes

directly to the quality of service guaranty that is only possible with virtualized storage.

Perhaps the greatest contribution of an enterprise class storage hypervisor to Storage as a

Service is its ability to provide predictability in the face of changing technology. Three

contemporary examples help to illustrate this idea:

Server Virtualization

Desktop Virtualization

Cloud Computing

Support for Server Virtualization

Server virtualization, which began to take hold in enterprise computing environments around

2005, imposed new requirements on storage infrastructure that, arguably, were not foreseen

by many early adopters. The ability to transition workloads between different hosts carried

with it a requirement for storage volumes to migrate with applications themselves. Given the

limited flexibility of “hard-wired” storage fabrics (sometimes called SANs) and their inability to

accommodate such ad hoc data migrations, planners confronted a difficult problem. Did the

adoption of server virtualization necessitate the abandonment of investments in Fibre Channel

fabrics and a return to direct-attached storage?

The good news is that storage virtualization has helped many companies surmount this

problem. With virtualized storage, physical storage infrastructure does not need to change to

accommodate changing workload locations or access patterns. A “virtual volume” containing

the data associated with a virtual machine can “move” with the virtual machine while load

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 6

balancing connections occur behind the scenes. That’s one way that a storage hypervisor

brings flexibility in the face of changing technology.

Support for Desktop Virtualization

Potentially much more important to organizations than server virtualization technology is

burgeoning virtual desktop infrastructure (VDI) technology. Virtualization of desktops, contrary

to its popular interpretation, is not simply about replacing or reducing the cost of physical PCs

and workstations. The physical client device will likely remain in place (with less costly devices

such as terminals and mobile devices such as tablets and smart phones added where needed).

The real benefit of VDI is the centralized hosting and management of desktop software.

The challenge, from a storage perspective, has been how to provision storage resources to

meet such a dynamic and unpredictable workload. Hardware array providers have performed

tests and measurements and concluded that buying “brand X” array would support thousands

of desktops for a much lower price than current desktop equipment. However, this is of little

help to IT managers who seek to deploy VDI in stages – supporting only a few hundred desktops

at a time.

Ideally, planners want predictable cost and performance metrics for VDI so that it can be

deployed incrementally. Using a storage hypervisor, as DataCore Software demonstrated in

2011 in a groundbreaking research and development project, could deliver just such a practical

– and predictable – roll-out model.

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 7

Readers here can find DataCore’s studies on their website. Simplified and condensed, DataCore

offers a hub-and-spoke model called a STAR configuration, that can be used to build out

virtualized storage infrastructure in “atomic units” delivering predictable capacity, performance

and cost.

While focused on VDI, DataCore’s findings could be applied to the design and implementation

of most multi-user application workloads. This is another way that storage virtualization

provides predictability – a key strategic infrastructure planning value – adding to an already

compelling business and technology value case.

Support for Clouds

Taken individually, server and desktop virtualization may be strategic IT initiatives in many

companies. And, as previously stated, a storage hypervisor can support these strategies by

delivering manageable, flexible and predictable storage resources that can adapt to changing

workload requirements. For many planners, a storage hypervisor is already an essential

component of a software stack that will lead to a virtual data center, or “cloud.”

The bad news is there is competition within the server and desktop virtualization market

among vendors whose products take decidedly different approaches to their task. For example,

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 8

in the server virtualization space VMware competes with Microsoft Hyper-V and

Citrix XenServer for market share.

Competition, while a good thing generally, can (and has) created more complexity and

challenge for those seeking to build private clouds or to leverage public cloud services that

leverage different server or desktop virtualization software stacks. For those firms using

multiple server virtualization products (as well as some non-virtualized servers), the issue has

begun to take on nightmarish proportions as vendor after vendor seeks to add a proprietary

storage hypervisor to its server hypervisor stack. Doing so requires planners to permanently

assign some storage assets just to machines running brand "X" server hypervisors, other assets

to brand "Y" server hypervisor workload, and still others to non-virtualized hosts. Exchanging

information between these isolated stacks, or managing storage itself in the face of such

segmentation, is a daunting challenge.

Public clouds are also beginning to manifest this problem. Sourcing resources from multiple

public cloud service providers increasingly runs afoul of the proprietary virtualization software

stacks used by different providers, significantly impairing the promise of cloud technology to

enable firms to select cloud resources for integration with their own IT infrastructure on an as

needed basis and with assurance that the resources can be used and managed holistically.

This problem underscores another value of a well-groomed storage hypervisor: it should not

only be hardware agnostic, permitting the deployment of any storage gear, but also server

hypervisor agnostic – capable of delivering storage services to any guest machine, regardless of

the brand of server virtualization software it is operating (if any at all). With DataCore

Software’s SANsymphony-V, for example, virtual storage volumes can be delivered on demand

to any and all guest machines, regardless of the server hypervisor that is being used.

SANsymphony-V already offers plug-ins to the management consoles of VMware vSphere and

Microsoft Server and Hyper-V.

In the case of internal clouds, SANsymphony-V provides a “universal bridge” between the

different server virtualization products, empowering planners to select the best server or

desktop virtualization stack for the job at hand, eliminating any sort of lock-in to a particular

server hypervisor’s wares, all without sacrificing management or existing investments in

storage infrastructure.

This ideal of stack agnosticism is being extended to the realm of public clouds as well. The

simple solution to the problem of integrating public cloud storage services would be for every

cloud service provider to simply deploy the same storage hypervisor, such as DataCore’s

SANsymphony-V. That would enable the storage hypervisor to serve as a “Rosetta Stone”

capable of translating disparate resources into a common service.

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 9

However, the real world is a bit more complicated. So, another approach is to use an

intermediary appliance that translates storage at a cloud service provider into a resource that

can be pooled with storage locally virtualized and managed using SANsymphony-V.

DataCore recently partnered with TwinStrata to do just that. The TwinStrata CloudArray

software appliance seamlessly integrates with DataCore SANsymphony-V storage hypervisor.

This bridges SANsymphony-V virtual storage pools to external cloud resources, enabling

organizations to fully manage both private and publicly-sourced storage resources.

The three examples above illustrate how a best-of-breed storage hypervisor can make a

significant contribution to storage efficiency – even in the face of technology change.

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 10

GETTING STARTED

The previous sections of this document have sought to explain why you should consider

virtualizing your storage. All that remains is to provide some guidance for actually doing it.

This is based on the author’s experience in deploying DataCore Software’s SANsymphony-V in

his own company, and also on numerous interviews seeking best practices from other

SANsymphony-V users large and small.

Broadly speaking, getting started means determining which assets of your storage

infrastructure to virtualize first. We could offer guidance that reflects our view of the

limitations and inefficiencies of current storage hardware products, but the best guide may well

be something more simple and practical: payoffs.

Ask yourself, where do you hope to derive the quickest payoffs from virtualizing storage?

Where will storage virtualization success deliver value to you in the form of self assurance or

confidence in the strategy itself? Which parts of your storage infrastructure, when virtualized,

will deliver recognizable improvements in storage services – delivering quick wins and

engendering a better understanding of the broader potential of the technology? What storage

virtualization success will secure for you (and the strategy itself) the approval of those who

keep the budget for an additional investment in a storage hypervisor?

These questions may be difficult to answer without some careful consideration and reflection.

When you have decided, here is one approach you may find useful to start the project.

The good news is that deploying a storage hypervisor may cost you nothing in terms of

hardware. DataCore’s storage hypervisor, SANsymphony-V can be deployed readily using

repurposed servers and existing storage.

1. If you use server virtualization today, consider consolidating two or three physical

servers by consolidating them into virtual machines.

2. Repurpose two of those servers to use as DataCore storage virtualization nodes.

3. Disconnect your direct-attach disks from the apps servers when you have a convenient

planned downtime window and relocate them evenly behind the DataCore nodes.

4. Use the SANsymphony-V pass-through feature to bring those volumes under the control

of the storage hypervisor. No lengthy migration is necessary.

5. Set aside extra disk space to mirror those volumes to the companion node.

6. Mirror the pass-through volumes to the companion node on thin provisioned pools.

Copyright © 2013 by The Data Management Institute LLC. All Rights Reserved. 11

7. Redirect the workloads to the mirrored node.

8. Wipe the original volumes clean and reclaim the space as part of a thin provisioned pool.

9. Re-synchronize the mirrors, now fully virtualized, with any wasted space from fat

provisioning reclaimed.

10. Take the third server to a remote site as the SANsymphony-V disaster recovery node.

11. Connect the primary site to the DR site over an IP WAN.

12. Make a local clone of the virtual disks you want to replicate remotely using the

initialization feature.

13. Ship the clones to the remote site and connect them to the DR node (this is usually

faster than attempting to clone a large amount of data over a WAN link).

14. Turn on the remote replication to keep the DR copies reasonably up to date.

15. Set up Advanced Site Recovery so if you have to switch to the DR site, any updates that

transpire there can later be synched back to the primary site before restoring normal

operations.

These 15 steps will get you going on your first implementation of DataCore Software’s

SANsymphony-V. To get another view of what you should expect from this process, you may

want to consult the Lab Validation Report produced by Enterprise Strategies Group (ESG) and

posted on DataCore Software’s website. Their mixed workload benchmarking revealed that a

virtualized environment with the same equipment could handle five times more Microsoft

Exchange users than could a standalone implementation of the application without virtualized

storage.

That done, you have taken your first steps towrd a more efficient, more flexible and more

adaptive storage infrastructure. You may not achieve actual virtualization “Rock Star”

status, but you are sure to get requests for many encores in the future.

Copyright © 2012 by The Data Management Institute LLC. All Rights Reserved. 76