Storage Consolidation Expert Rajeev Bhardwaj

Behind labor, energy costs are emerging as the second highest operating cost in most data centers worldwide. Operating in an economy still in the early stages of recovery, companies must explore all options available for reducing data center energy costs.

Today’s entry in our Data Center Leaders Interview Series sets out to help companies make the critical IT decisions that will lower costs, including those associated with storage consolidation, without hindering performance.

For help, we turn to Rajeev Bhardwaj, Senior Director, Data Center Switching Technology Group, Cisco:

Evolving Solutions:
What do you recommend as a first step in the development of a cost efficient storage consolidation or networking plan?

Rajeev Bhardwaj:
End-to-end virtualization will consolidate the datacenter and will deliver operational efficiencies from the perspective of being cost-effective and applications being easier to manage, provision, and deploy. End-to-end virtualization in the datacenter includes server virtualization, server I/O virtualization, Storage Area Network (SAN) virtualization, and Storage virtualization.

End-to-end virtualization should begin with SAN virtualization as the SAN network interconnects all the storage and server elements in the datacenter and a consolidated SAN can serve as a platform for further virtualization of servers, server I/Os, and storage.

Customers typically deploy silo’ed infrastructure with multiple SAN islands based on departmental, application, and performance needs. The consolidated SAN must maintain the logical separation between the SAN Islands from both data path and control plane perspectives while consolidating them on a single, physical infrastructure: Virtual SAN (VSAN) allows to easily achieve that virtualization abstraction.

Consolidating multiple SAN islands in a virtualized SAN allows to significantly reducing the number of infrastructure assets required, being now shared and dynamically allocated based on actual needs.

This approach optimizes assets’ utilization and dramatically reduces deployment, management, as well as power, space and cooling costs. A consolidated SAN has the additional advantage that intelligence can now be deployed in the SAN to enable both server virtualization and storage virtualization.

Evolving Solutions:
What are the inherent dangers in trying to make your storage consolidation plan too cost efficient?

Rajeev Bhardwaj:
It is very important to consider the underlying architecture of each device required to deliver the consolidated SAN. If the SAN switch architecture is not robust and delivers unpredictable performance and latency, the consolidated SAN becomes a central bottleneck affecting all applications and all departments.

One should consider questions such as – how does the switch architecture handle congestion, does it exhibit consistent latency and performance, does it provide enough buffer-to-buffer credits per port to handle different applications and distances, does it provide high availability and in particular for links that carry a lot of bandwidth between switches?

Evolving Solutions:
Green IT is seen by many as a solution to upgrading existing data centers to be more cost efficient.  Do you feel green IT practices can significantly impact the cost of storage consolidation?

Rajeev Bhardwaj:
A piecemeal approach of deciding each IT component based on its power consumption can work fine for marginally reducing power as a first step, but can actually be counter-productive overall if a holistic view of the data center power is not considered.

Solutions that at the chassis level seem to provide slightly better power consumption, have proven to increase the overall power cost of the datacenter: poor network support of server virtualization constraints the consolidation of servers, consuming over 75% of the overall datacenter power budget; limited networking capabilities and unpredictable performance require to increase network size as soon as new applications are deployed in the datacenter, transforming the initial minimal saving in a major cost.

Green IT can reduce cost and power usage while increasing operational efficiency in the data center if achieving end-to-end virtualization is maintained as a goal. Deploying the right networking infrastructure can provide security, performance, Quality of Service (QOS), and ease of management required to enable reductions in the size of the server farm, the number of LAN and SAN switches, and the number of storage arrays.

Evolving Solutions:
In a recent interview, author and StorageIO found Greg Schulz helped dispel the myth that storage networking and consolidation are business expenses that only apply to large enterprises.  In your experience, are smaller businesses aware of the need for storage consolidation and networking plans?

Rajeev Bhardwaj:
In these tough economic times, smaller businesses are looking to reduce both capital and operational expenditure. Smaller businesses understand that times are tough but are also looking to position themselves for the turn-around in the economy and so are not willing to sacrifice on scalability.

IT is increasingly considered as a means to increase the company efficiency and competitiveness by providing real-time data to any sales person, partner, or executive anywhere at any time while maintaining security and keeping costs low.

Smaller businesses are very much interested in buying and maintaining fewer servers, fewer switches, and getting more mileage out of their existing storage arrays. Thus we see smaller companies collapsing their multiple fabric switches into a larger director to gain scalability and high availability while reducing management expenses.
Similarly, we have also seen keen interest in our Fibre Channel over Ethernet (FCoE) portfolio from smaller business looking to consolidate their LAN and SAN expertise and infrastructures.

Evolving Solutions:
Omar Sultan, your colleague from Cisco, envisioned matured virtualization, greater use of automation and the evolution of cloud computing as solutions that would lower data center costs five years from now.  What would you add to this list?

Rajeev Bhardwaj:
Most organizations have separate LAN and SAN departments managing separate networks today. Five years from now I see this distinction being blurred and administrators who are aware of both LAN and SAN environments running a common network based on common switching components.

I see this common, cost-efficient network providing the under-pinning for higher data center virtualization, automation, and ultimately evolution of highly scalable, secure, and highly available private and public cloud infrastructures.

Evolving Solutions:
Wild Card:  Anything else you would like to add?

Rajeev Bhardwaj:
Energy costs are emerging as the second highest operating cost (behind labor) in most of the data centers worldwide. While networks consume a very small percentage of the overall data center power, deployment of intelligent networks can drastically reduce end-to-end power consumption and management costs while reducing provisioning time for new applications and achieving greater operational efficiencies in the data center.

Storage Consolidation & Networking Expert Greg Schulz

As businesses in any industry grow, they often find themselves with both more data to mange and more costs to contain. Storage consolidation, networking and Green IT can all make for effective solutions if implemented in line with overall business needs and objectives.

Our Data Center Leaders Interview Series offers detailed advice on how companies can best keep synch needs and objectives with data center solutions, courtesy of Greg Schulz.  Schulz is founder of the independent IT analyst consultancy firm StorageIO Group and author of the books The Green and Virtual Data Center(1) and Resilient Storage Networks(2).

Below, we discuss with Greg how to begin development of a storage consolidation or networking plan, detail the benefits of Green IT and debunk the myth that business continuity and disaster recovery are only for the rich and famous:

Evolving Solutions:
In your book, Resilient Storage Networks, you speak of the growing need to store any piece of data and access it at any time.  At what point do businesses put themselves in danger of not being able to achieve this goal without a storage networking solution?

Greg Schulz:
The reality is that a networked storage, or, storage networking solution if you prefer, enables scaled connectivity either from a performance, accessibility, available or all of the above perspective beyond the limits of a dedicated direct attached connection.

Note that a direct attached connection could be a Serial Attached SCSI (SAS), Fibre Channel or iSCSI without a switch topology. Thus as more servers need to access shared storage than what can be provided by the native ports on a storage system, or, more storage is needed than what can be attached to the native ports on a server, a storage network becomes beneficial.

On the flip side to this, a storage network or networked storage solution can also exist in the form of a switch-less configuration, granted some of the storage networking police who prefer switches might not agree with the definition. For example, a large storage system with 8, 16, 32, 64, 128 or more Fibre Channel ports could be considered and are often marketed as a storage network solution.

The point is this, a benefit of a networked, or, storage network storage solution is the ability to scale performance, availability, capacity and connectivity beyond the limits of a smaller dedicated internal based storage solution.

Thus the emphasis should be on getting organizations to deploy shared storage that can grow with the business including attaching to larger switch based storage networks when and where needed.

Resilient Storage Networks by Greg Schulz

Evolving Solutions:
What do you recommend as a first step in the development of a storage consolidation or networking plan?

Greg Schulz:
Take a step back from the various technologies and tools. Have a clear understanding of business objectives and requirements to know how to make a linkage to the benefit of the solution and how the business is sustained or enhanced.

This helps to build a business case and then measure the ROI impact for future project enhancements particularly during tough economic times. Where this comes into play is being able to map to the business the technology that is needed to enable sustainability thus requires funding compared to those technologies or initiatives that would be nice to have or seen as discretionary and thus miss out on funding.

Also, keep a balanced approach to avoid tunnel vision on space capacity utilization. While there is a lot of rhetoric about the need to drive up storage or space utilization, don’t do it at the expense of degraded performance or introducing application instability or degraded quality of serve. Some systems have a low utilization to meet quality of service or other performance or service level requirements in which case the solution can be to move those applications onto faster performance storage in order to achieve affiance.

Keep in perspective what it is that you are looking to move or the problem being fixed. Thus, management insight tools are needed that can shed light not just on storage space capacity utilization, that also show performance and activity usage information. Find a balance between performance, availability, capacity and energy (PACE) o meet a given service level requirement and cost points.

These and others themes are discussed in more detail in my new book “The Green and Virtual Data Center” (CRC) available at Amazon and other venues.

Evolving Solutions:
How do green IT strategies synch with storage networking?

Greg Schulz:
Green IT strategies which are really about addressing and enabling business sustainability and growth by optimizing IT and data infrastructures have a strong synergy with efficiency enabling technologies including networked storage or storage networking as well as virtualization among others.

Most organizations do not buy and deploy networked storage, regardless of if SAN, NAS, Fibre Channel, FCoE or iSCSI just for the sake of buying it, that is unless they have a really good sales rep.  Instead, most organizations deploy storage networks as a means of improving efficiency and optimizing how IT resources are managed and used. Likewise, most organizations don’t or can’t afford to go green just for the sake of meeting PR or other initiatives and many organizations see green as being all about carbon footprints.

The reality is that most organizations have a need to address their power, cooling, floor-space or footprint as well as other environmental issues to support and sustain business growth, these are the other aspect of being green.

Ironically it is these issues that most organizations have been tasked with addressing and in many cases not making the connection that these are in fact green it issues that by addressing them, they in fact are being green. This paradigm is known as the green gap, thus tools, technologies and techniques including storage networks can help to improve on affiance and optimization enabling enhanced IT infrastructure resource management (IRM).

The Green and Virtual Data Center by Greg Schulz

Evolving Solutions:
In your experience, are small businesses or major enterprises more aware of the need for storage networking solution?

Greg Schulz:
There is a common myth that business continuity/disaster recovery (BC/DR) are issues and concerns only for the rich and famous meaning only for large enterprises or high profile financial organizations. The reality is that BC/DR along with general HA and data protection is affordable and the issues are applicable to organizations of all size.

Given the different size, scope and scale of these various sized organizations, there may or may not be the awareness of the benefits of specific technology including networked storage, thus there is an opportunity to show how data protection including BC/DR can be enabled and enhanced leveraging storage networks and other related technologies.

Likewise, there is an opportunity to leverage networked and shared storage for organizations of all sizes. At the high-end that can mean Fibre Channel and moving forward Fibre Channel over Ethernet (FCoE) or iSCSI in mid-sized environments along with NAS, while smaller environments can leverage low cost Fibre Channel, iSCSI, NAS or even shared, multi-host SAS based storage systems for small Exchange, SQL and VMware based clusters.

Thus tiered access for storage networks enables the most applicable technology to be mapped to the business needs of different environments. For example, a small environment can attach and share a dual controller RAID array such as the equipped with SAS/SATA disk drives attached via shared SAS, or larger environments share the same array across more servers with native iSCSI attachment, or even larger environments use the same array configured with Fibre Channel for either high performance scenarios.

Those are great example of tiered access to a tiered storage device enabling a networked or storage networked device for organizations of different sizes.

Evolving Solutions:
Leaving the company anonymous, what is the largest data disaster you’ve witnessed that could have been avoided with resilient storage networking?

Greg Schulz:
Not all disasters make the headline news, in fact, every day there are countless mini-disasters that due to configuration, best practices and how technology is used combine to contain these from expanding and rolling into major disasters.

In one situation it was a scenario where resilient components were interconnected thereby negating the benefits of no single point of failure (SPOF) resulting in downtime. In another scenario, lack of change control and configuration management, in another, it was a focus on capturing data in as small time slice interval as possible, only to not have captured all of the data that was still in memory thus resulting in an inconsistent recovery.

Another scenario was an outage where the IT systems were available; however no one could access the data as the network links were down including those from a secondary carrier who just happened to use the same common back-bone carrier.

Let’s not forget about scenarios where RAID and replication without point in time based snapshots or backups were confused with being a replacement for backup.

In those scenarios, data was in one case replicated to another site however when deleted at the primary, it was deleted at the secondary. In a related scenario, RAID was assumed to be the backup, that is until the entire RAID storage system became un-available including the snapshots that were stored on the device.

Don’t confuse or assume RAID and replication alone are replacements for backup and data protection as they need to be combined with some type of time recovery point based technique for complete or comprehensive data protection.

The common theme in all of these is that it’s not how much hardware or software, how many 9’s availability, it’s the people, processes and procedures including how they are configured to avoid, isolate or eliminate faults from spreading into disasters.

Thus, look for single points of failure, avoid being a penny wise and pound foolish. For example in the quest to save a few hundred or perhaps thousand dollars by skipping a second HBA or adapter on a server that is going to be used for virtualization and consolidation only to expose it as a single point of failure where multiple servers no exist (e.g. multiple eggs in one basket).

The bottom line is this, if something can happen it will, all technology will or can fail regardless of vendor marketing spin, however what differentiates various vendors is how they have learned from these faults or issues to configure around them leveraging best practices and their experience.

(1)The Green and Virtual Data Center is published by Auerbach-CRC.
(2) Resilient Storage Networks is published by Elsevier.

Learn more at www.thegreenandvirtualdatacenter.com or on twitter @storageio.