The EPA issued a letter on August 8, 2008 to Computer Server Manufactures or Other Interested Stakeholder, asking for a review of draft 2 of the ENERGY STAR® specification for computer servers. The response is due by September 19, 2008. 

The purpose of this specification is two-fold: “(1) to identify the top performers currently available in the marketplace based on key energy efficiency metrics and (2) to provide end users with more detailed product and performance information so they can make informed purchasing decisions based on consistent manufacturer reporting.” 

The EPA intends to define ENERGY STAR® specifications for servers by the end of this year and then turn their attention to storage. SNIA has established a Green Storage Initiative (GSI) to manage the overall Green Storage program for SNIA including the industry relationships with EPA, The Green Grid, and SPC. There is also a Green Storage (GS) TWG that is primarily focused on the definition of power efficiency metrics to be added to the EPA’s ENERGY STAR® document for Enterprise storage. 

While this draft document is focused on servers, it will be important for storage vendors to review the methodology that is being applied to servers. One of the paragraphs in this letter caught my attention: 

“Power management and virtualization have been removed as potential specification requirements. EPA decided to exclude these criteria given the challenge of writing technology neutral requirements that are also dependent on end user choice and behavior to realize savings. EPA encourages virtualization where it can provide energy efficiency benefits as well as the deployment of power management techniques and features but considers these to be best practices as opposed to specification requirements. However, EPA does believe that full disclosure of these energy saving criteria is important. As such, manufacturers will be required to report power management features and virtualization capabilities as part of the standard information reporting requirements presented in Appendix A, (Required ENERGY STAR ® Qualified Product Information) of the Draft 2 specification.” 

If we follow the same thinking with storage, this may mean that features like MAID, spin down, thin provisioning, de-duplication, copy on write, single instance store, tiered storage, archive and virtualization may not be in the specification, and may only appear in Appendix A. Is this good or bad? A product with a better ENERGY STAR rating may actually provide less energy savings than a product that has a lower rating but can provide more efficient use of storage. That will be one of the issues that the SNIA GS TWG will be addressing. If you have thoughts on this please funnel them into the SNIA GS TWG.

WIKIBON, a storage industry group, has launched a service to help IT companies and their customers to qualify for rebates from utility companies like PG&E for installing energy efficient equipment. These rebates can be based on energy saving technologies like virtualization, thin provisioning, tiered storage, etc; technologies which may not have an ENERGY STAR rating.  

By the way, one of the Appendix A, “Required ENERGY STAR® Qualified Product Information” for this server draft is performance benchmarks. This makes sense since higher performance is often a trade off with higher power consumption. However, this may be a problem for a storage vendor who declines to publish benchmark results. 

The move to network unification recently took a step forward with an FCoE Test Drive Event in June 2008. FCoE can send FC protocols over an Ethernet network and makes it possible to have a common Ethernet network support FC storage and Ethernet message traffic. This would greatly reduce connectivity costs, especially for servers. Today servers require redundant connections for storage, LAN traffic, and clustering connectivity. A redundant pair of FC host Bus Adapters alone could cost as much as the server itself.

When one thinks of Storage and Ethernet, there is a concern about guaranteed delivery and quality of service. While LAN traffic can tolerate dropped packets and variable service, FC block storage expects packets to be delivered in order and within a limited window of time. Since disk storage is a rotating media, if a sector address is missed, the disk takes another spin before we have another shot at the sector, and eventually the I/O times out. In order for FCoE to support block storage it must overcome these deficiencies. Extensions are being made to Ethernet standards, called Convergence Enhanced Ethernet or CEE. CEE will provide for congestion management, elimination of dropped packets, ensure interoperability, priority processing, and packet scheduling. These standards are expected to be finalized in 2009.

Since the biggest advantage of convergence is for the servers, they are expected to be the first to move to FCoE with a Converged Network Adapter, CNA, combing a FC HBA with an Ethernet NIC card. Qlogic has a CNA that is ready to ship and CISCO has an Ethernet switch available now.  Brocade will also have an Ethernet switch shortly. On the storage side, adoption is expected to be slower due the investment that data centers have in FC and the fact that FC works. It is expected that Ethernet switches will initially provide a bridge to FC storage. Moving to FCoE for storage arrays would only require the replacement of the front end port modules. 

Chris Mellor points out that, SAN virtualization products, like EMC Invista and IBM SVC which reside in the FC SAN will require significant effort to convert to FCoE. Since HDS’s storage virtualization is done in the USP V storage controller, and not in the SAN Fabric, it is expected to be much easier for HDS to convert to FCoE.

One more advantage for Storage Virtualization in the Storage Controller.

Congratulations to SNIA on their 10^th Anniversary and to Vincent Franceschini of HDS and Larry Krantz of EMC who have just been chosen to join SNIA’s Hall of Fame Contributors. Larry is a past Chairman of the Board for SNIA, and Vincent is the current Chairman of the Board. Both of them made major contributions towards the success of SNIA through their leadership during important growth periods in the storage networking industry. They join other Hall of Fame Contributors, Mike Dutch, Paul Massiglia, Roger Reich, John Tyrrell, and John Wilkes.  Above and beyond their day jobs, they contributed their time and talents to achieve the goals of SNIA which benefits us all as we work in an increasingly networked world.

The Storage Network Industry Association grew out of the early interactions between vendors who suddenly found themselves connected to each other in this new thing called a Storage Area Network. Up until then storage was direct attached and everything was proprietary. Now all of a sudden we were attached together in a network with other storage vendors and new vendors that built the switches and host bus adapters. Server vendors no longer had a lock on storage and now many new storage hardware and software vendors were able to able to bring new solutions to the market.

It became quickly apparent that when the network was down, all the vendors had to be involved in resolving the problem. Interoperability had to be worked out, interfaces had to be defined, and cooperative support agreements had to be in place. These were the early problems that SNIA had to address. Without that work in SNIA, we would not have the SAN’s we have today. It also makes it possible to move to the next step which is virtualization of heterogeneous storage systems.

The success of Storage Area Networks and Storage virtualization depends on the work of SNIA and SNIA depends on the support of volunteer, individuals, like the Hall of Fame Contributors.  You can read their profiles in the link above.

Rick Villars of IDC believes that the predominant storage system in the future will be modular, two controller, storage systems. While modular storage will continue to be the storage of choice for the midrange market, it will have an expanding presence in the enterprise market as “role based” storage systems. This is storage that meets a particular role, such as NAS, VTL, Archive, or as a back end tier to a storage virtualization controller. While modular storage systems do not have the high availability of large enterprise storage systems that have 32 to 128 controllers clustered around a global cache, that deficiency can be masked though virtualization behind an enterprise storage virtualization system.

Since an enterprise storage virtualization controller can provide thousands of virtual ports for connectivity, large front end cache capacity for enhanced performance, thin provisioning for increased utilization, and non-disruptive, copy, move, and replicate for high availability and business continuity, the back end storage can be designed for low cost, modular capacity. Rick also believes that enterprises will be refreshing this modular storage on an accelerated 18 month cycle to take advantage of the price erosion of capacity which will continue to be 30% per year. Virtualization will make it easy to swap out or refresh storage modules without disruption to the application and without the need to swap out the entire storage farm. The days when data centers replaced their storage farms on 3 years or 5 year cycles, buying future capacity at today’s price, are over.

With storage virtualization and dynamic provisioning (thin provisioning) it is now possible to provision smaller increments of storage dynamically. Storage virtualization alone or thin provisioning alone will not be able to provide this capability. Thin provisioning enables you to share a pool of storage across multiple applications as though it were a liquid pool of capacity. Storage virtualization enables the extension of this pool of storage across multiple heterogeneous storage systems, adding modular storage systems as required.

An intelligent storage virtualization controller can provide a platform for current and future, enterprise, storage services, which can be applied to lower cost modular storage systems. This separation of intelligence from commodity capacity will enable IT to provide the right mix of, cost effective, storage services to meet business needs.  

Last night the executives of our Hitachi development laboratories hosted a dinner for the Asia Pacific Media tour at a traditional Japanese Restaurant in Odawara. It was a fabulous meal which lasted almost three hours. I had forgotten what it feels like to sit on a tatami floor for three hours. As I struggled to stand up while maintaining my balance, I was reminded about modular storage.

The advantage of modular storage is its lower cost and modular architecture which makes it well suited for the midrange market. Modular storage can be expanded by adding drawers of disks into a standard 19 inch rack. It can be configured in the same rack with server, switch, and other peripherals that are also packaged in 19 inch drawers. Modular storage does not need a raised floor data center since each drawer has fans which draw the cooling air in from the front and exhaust out the back.

Up to now the limitation of modular storage is that it only has two controllers. Each controller has its own ports to the servers, manages its own cache, and provides RAID parity generation and access to the disks that are assigned to it.  The controllers are designed to fail over to the other in the event of a controller failure. In order to maintain cache coherency, the controllers are assigned separate volumes or LUNs for processing. Writes to one controller’s cache is mirrored into the redundant controller’s cache to avoid data loss in the event of a controller failure. While this avoids the loss of cached data for the failed controller, this does not provide availability since the remaining controller now must take over ownership of the other’s volumes and do the work of both controllers. With one controller we no longer have write protection for cached data and we must either switch off the cache to write directly to disk or take the risk of data loss if the remaining controller should fail. Sooner or later the system would have to be shut down to fix the failed controller. This also applies when the controllers have to be taken down for maintenance. While modular storage is more affordable than monolithic storage, the two controller, modular storage is not high availability. However, it is very cost effective if you have maintenance windows.

That is why I was reminded of modular storage after sitting on my legs for three hours. A two controller modular storage system is like a pair of legs. My legs are redundant but if I injure one leg, I usually stop and wait for it to heal it before I go back to work. 

For high availability modular storage, Hitachi provides the USP VM. It is packaged in a 19 inch rack and priced for the midrange market. But instead of two controllers it has 32 controllers which share a common cache with write protection. With a common cache for all the controllers, volumes do not have to be partitioned between the controllers for cache coherency. With this system you can lose one or more controllers with no data loss, and if you have alternate controller paths to the same cached data, there is no need to transfer LUN ownership and no disruption to the application. Since the controllers and cache are connected through a switch that can switch around hotspots and failures, you may not even see a performance degradation.  The availability of more than two controllers and the ability to share the same cache provides high availability with no disruption to the application.

The USP VM also has all the functions provided in the USP V. This includes the ability to virtualize existing storage assets behind the USP VM, and enhance their capability with services such as dynamic (thin) provisioning, replication, and tiered storage. If you attach a modular storage system behind the USP VM and need to do maintenance on it, the USP V or VM can move the data off during the maintenance and then move it back afterwards, all without disruption to the application, making your existing modular storage a high availability storage system. 

Eventually I was able to recover the use of my legs and hobble out of the restaurant. My legs are definitely not built for high availability or for sitting on them. 

I had the pleasure of hosting a panel of CIO’s for Camp IT’s  IT Leadership Strategy Conference in Rosemont, Illinois. The topic was “The Next IT Wave: What Might It Be and How Can You Prepare For It”. The panelists we had were excellent with a wide range of experiences. Kim Tracy, Executive Director, University Computing Services, Northeastern Illinois University, also teaches at the University and had worked at Bell Labs. John Phillips, Vice President and CIO for AllScripts, with twenty years of technology experience in business alignment and operations also had experience at companies like Accenture, RR Donnelley, and Thomson Corporation. Anuraag Bhargava, CIO for Electro-Motive Diesels, had a wealth of international experience, as well as experience as a Principal in the Strategy Practice at A.T. Kearney. Steve Carlberg has responsibility for the overall design, selection, and integration of various networking technologies at University Health System Consortium. Steve also had nine years experience in the US Navy Nuclear Power Program on board sumarines.

One of the IT Waves they discussed was the one identified by Forrester as “Technology Populism”. According to Forrester, this next wave of change will be fueled by the proliferation of consumer devices, social networking tools, and cloud-based collaboration services making their way into the enterprise.

At the forefront of this wave is the new generation of IT consumers, the millennial generation who grew up with social networking and always on, always available, everywhere access. These are the IT consumers that Kim Tracy serves everyday in the University. While cloud computing is not on the near horizon, he will have to be able to serve his users on whatever device they use to check their grades and course materials. He noted that while this new generation is an easy user of new technology they are not necessarily knowledgeable about the technology. This means that IT must be more knowledgeable about technology and its implication on operations, service, and security. This is changing the role of the CIO into that of a CTO.

John Phillips is beginning to use the cloud to make his business processes more agile. He cited the use of SalesForce.com as a way to make his sales more productive. He is also a strong proponent of virtualization. In the past when marketing wanted to initiate a new marketing campaign, they would have to outsource it in order to meet the targeted timelines. Now he can provision new virtual servers and storage and keep the spend in house with better control.

Steve Carlberg, also sees the value of cloud computing and the ability to easily provision services to meet changing business needs. However, in his business where he provides University Hospitals with proprietary analysis, he must keep his core applications in house. Steve is also a strong advocate for virtualization both on the server and storage sides. He is using a USP V with Dynamic Provisioning (thin Provisioning) on external storage.

While Anuraag Bhargava, might be considered to be working in a structured, manufacturing, supply chain, business, he is looking for ways to exploit these new technologies. He sees the use of mobile devices on trains to monitor the electro-motive engines and upload real time data. While he is not ready to use third party cloud providers, he sees the need to develop his own cloud services. His future data center will be a services oriented data center, responsive to changing business requirements, where the end user will not be concerned about what is behind the cloud. EM Diesel also has USP V’s with Dynamic Provisioning to provide Services Oriented Storage.  

Not only is it clear that IT technology is changing, but the CIO role and prerequisites are changing as demonstrated in the background and views of these four individuals. For a bio on the distinguished speakers at this IT Leadership Strategy Conference please click here.

IDC published a new report “The real costs to Power and Cool All the Worlds External Storage”.  Here are some of the excerpts from this report as reported in their press release of 17 June 2008

“With an aggregate worldwide electricity cost of $0.07 per kilowatt-hour last year, IDC estimates that the total amount spent on powering and cooling these drives exceeded $1 billion in 2007.”

“As companies continue to add storage capacity at an aggregate rate of over 50% per year, the number of spinning disks continues to be a larger part of the overall power and cooling costs within a datacenter,” said David Reinsel, group vice president for IDC Storage and Semiconductors research. “Vendors must do more to promote and enable well-rounded green storage strategies that include datacenter redesign, data consolidation, and data reduction.”

Here are some of the things that Hitachi is doing to enable a well rounded green storage strategy.

Data Consolidation
Hitachi is enabling data consolidation, in a well rounded way, to tiers of storage. Recognizing that disk drives consume roughly the same amount of power and cooling whether they contain 75GB or 1000GB, Hitachi provides the ability to combine different capacity/performance disks to be utilized in a common pool of tiered storage, with non disruptive movement or copy of data across tiers of storage based on performance requirements. High Speed FC disks are usually considered tier 1, and slower speed SATA disks are considered to be tier 2 or 3 storage depending on their capacity. Instead of 10 high performance FC disk, the pool of disks could be reduced to 2 FC disks and 4 SATA disk for the same total capacity and equivalent performance with 40% less power and cooling. Both the Hitachi Enterprise USP V/M and the modular AMS storage can provide non disruptive move and copy of data across tiers of internal storage. The USP V/VM can also move or copy data non disruptively to external storage from different vendors.

Storage utilization
Besides consolidation of data, storage can be more effectively utilized, especially on open systems where the over allocation of storage is prevalent. The primary volume may be more than 100% over allocated to prevent out of capacity conditions, and every copy of that volume replicates that over allocation for backup cycles, data mining, development test, data extraction, business continuance, etc. The best way to address this over allocation is through Dynamic Provisioning. Hitachi’s Dynamic Provisioning not only provides thin provisioning to greatly reduce the over allocation of the primary volume but also in all the replicated copies as well. By wide striping the volumes across all the disks in the provisioning pool, Dynamic Provisioning also enhances I/O performance through   the parallel performance of all these disks working together. When Dynamic Provisioning is combined with the virtualization of the USP V/VM, it can extend the benefits of thin provisioning and wide striping to externally attached storage without the need to upgrade or replace the external storage.

Data Reduction
Data reduction involves the elimination of redundant copies of volumes, files, and bit strings.  This involves de-duplication, copy on write, single instance store, compression, and elimination of intermediate volumes for business continuance. This also involves software management tools to discover volumes and files that are orphaned or in active. See Hitachi Data Systems web page for software management tools for de-duplication, data discover and capacity reporting.

Reduce the working Set
This is a category that has to do with data that is over 60 days old and is clogging up your production systems, or your working set of data, and is being backed up and replicated and managed on a daily basis. Sol Squire of Data Islandia calls this type of data “Toxic Waste” since it sits on your production systems taking up power and cooling without providing any production value. Hitachi’s solution for this is HCAP, the Hitachi Content Archive Platform.  With this solution you can archive a “gold copy” of an object and store it in a compliant active archive, where it can be managed and protected and technology refreshed through policies. This platform works with many of the major data base, email, file, and document management archive systems, providing a common search capability and scalability to 32 billion objects and 20 PB of capacity. For an Enterprise Strategy Group report on this product please click here.

Data Center Redesign
Hitachi ltd our parent company has set a goal to reduce power in the data center 50% in 5 years starting in 2007. The benchmark for this program which we call “Cool Center 50”, are the two main data centers in Yokohama and Okayama. The Yokohama data center is targeted to reach a Power Usage Efficiency rating of 1.6 by late summer next year, which is expected to be the lowest for a tier IV data center. While Hitachi ltd has an arsenal of technologies to address this with power and cooling systems, virtualization servers, and communications equipment, a significant part of the saving will come from the green storage strategies mentioned above.

Last week in Orlando, Andy Kyte, Gartner Research Fellow was quoted as saying: ‘”IT modernization is the escape route from the accidental architecture of 20th century IT. Every company is dealing with the history of IT. Some fundamental change is needed. I don’t think we can carry on the bad habits and siloed decisions of the past. I’m looking for infrastructure and operations to begin the process of transforming IT.” IT must be overhauled from the siloed decisions of the past and transformed into an agile service provider. For a report on Andy Kytes keynote at Gartners IT Infrastructure, Operations and Management Summit see Denis Dubie’s article in NetworkWorld, “Stop buying more technology, start maximize current investments; Gartner” 

Andy has been leading Gartner’s initiative to make IT modernization a key strategy for 2008. He has predicted that many of the applications, infrastructure, and processes that were developed during the last 20 years will be obsolete in the next 5 to 7 years as new technologies, new processes and new business demands evolve. IT must begin the modernization of out dated IT processes and technologies, and that modernization begins with the way that IT acquires and maintains technology. Instead of measuring success in terms of ROI, the return on investment of acquiring new technology, he recommends a new metric, ROA, return on assets. ROA is a measure of success on the returns that IT gains on assets that are already deployed and a metric for planning for future assets. 

ROA is a great measure for Hitachi Data Systems’ Services Oriented Storage, SOS. With control unit based storage virtualization, SOS can leverage and enhance existing storage assets with the latest services of an enterprise class control unit. An example of a service is Dynamic Provisioning, which includes thin provisioning for improved utilization of capacity, and wide striping for improved performance.  

Andy also observes that IT can not wait for a change in business requirements before initiating a change in IT process or infrastructure. Business can not be competitive if it has to wait 6 months to a year or more for support from IT. 

Services Oriented Storage addresses this by providing a platform for storage and data services technologies that can be quickly adapted to meet changing business needs. SOS is similar to Services Oriented Architecture, SOA, in providing business agility through the use and reuse of common services. SOS provides services and not solutions. While a solution addresses a particular need like backup, SOS provides services to make that solution more efficient. An example of a service is snap copy that enables the backup to be done off of a point in time copy without the need to interupt the production volume. It can combine this with Dynamic Provisioning services so that the snap copy only contains data and not allocated unused space, reducing the elapsed time for backup. It can also use VTL and de-duplication services for faster recovery. SOS provides services that can be combined to support different solutions and make them more efficient, and cost effective. 

If you are considering making storage investments this year, consider moving to a Services Oriented Storage approach, instead of investing in siloed storage solutions that can not be adapted to new requirements or in capacity solutions that only provide a short term fix.       

IDC published a report this month entitled, “The real Costs to Power and Cool All the World’s External Storage”.  It’s a good report which I would encourage you to get for an understanding of the relationships between, capacity, performance, power, cooling, and cost. IDC has been able to extract their findings from their comprehensive collection of data on shipments of external disk storage since 1998. By the end of this year, 2008, IDC is predicting that there will be:

An install base of 49 Billion HDDs

Consuming 22 BkWh for power and cooling

At a total cost of $1.76B

Producing 34.1 B pounds of CO₂

While they see storage capacity growing by 50 to 55% per year, they see power consumption growing around 10% per year due to increasing HDD capacity and other strategies to reduce power costs associated with external storage.

While they enumerate these strategies they caution that customers must be careful in deciding which technologies best fit their needs. IDC points out that green technologies like spin down or larger capacity disks may impact performance and run counter to their other business goals. They also encourage vendors to offer complimentary strategies to help customers reap the full benefit of “turning green”. Rather than list their green strategies here, I will leave it to you to obtain this paper from IDC.

However, one of the strategies that is not on their list is Dynamic Provisioning. Hitachi Dynamic Provisioning has the ability to dramatically increase the utilization of storage by eliminating the waste of allocated but unused capacity and increase the performance of larger capacity, power efficient, disks.  This is done by striping pages across the width of a pool of disk RAID array groups. This wide striping enlists more disks in servicing an I/O request which increases performance. These pages are allocated dynamically as a volume is written to, rather than allocating the entire volume on one RAID array group at the beginning. This strategy has many complimentary benefits: increased utilization of storage capacity; increased performance through the use of wide striping, and reduced operational costs with simpler provisioning and performance tuning.

For example, let’s say that you have your volumes are distributed to 10 RAID array groups and have only 20% utilization. With Dynamic Provisioning we could combine 4 of these RAID array groups into a common pool of storage where the volume’s capacity is allocated from pages that are striped across all the disks of the 4 RAID array groups. Since we no longer have silos of RAID array groups we can now get better utilization of the capacity and we only allocate the capacity as it is actually used. Also since we now have volumes that span 4 RAID array groups, we have 4 times as many arms processing the I/O. We can release the other 6 RAID Array Groups and more than halve our power and cooling requirement just by going from 20% to 50% utilization.  

This strategy for dynamic provisioning is considered a common reusable service on the Hitachi Universal Storage Platform. V/VM. It can be combined with other services that are available on that platform, such as copy, move, and replication services. It can also be combined with virtualization of external storage so that dynamic provisioning can be extended to enhance the utilization and performance of lower level, external, storage.

The combination of Hitachi Dynamic Provisioning with virtualization of external storage can increase utilization of storage from around 20-30% to 60-80% and dramatically reduce the total power per disk which IDC estimates today to be 48 w (12 w per disk, 12 w for storage infrastructure, and 24 w for cooling).      

I will be speaking at the Virtualization/SOA World Conference and Expo in New York City on June 23-24. Virtualization and SOA are two of the hottest technologies that are combined at this one conference. My session will be Storage Virtualization, the Key to Unlocking Tomorrows Data center. You can checkout the link above to see the list of other speakers.

As a speaker I have been given the opportunity to invite you all to attend with a VIP Gold Pass. A VIP Gold Pass allows access to all of the technical breakout sessions, Keynotes and Power Panels, and vendor presentations in General Session, as well as the Expo Floor - for the entire event.

This is available entirely free of any charge to you as my  VIP guest!

Click Here to Register for their VIP Golden Pass:
https://www3.sys-con.com/virt0608/registernew.cfm?a1=gold **Select Golden Pass Registration package and enter the coupon code: speakerguestVIP (case sensitive) to redeem your pass**See you there!See you there!

See you there!