NetApp Product and Technology Insights

Getting Your Head Around the Clouds - Open Management

- blog post by VS Joshi – Sr. Manager for NetApp Manageability software.

This month I had the opportunity to attend IDC Directions. I've borrowed a slide from Matt Eastwood’s presentation as it caught my attention relative to IT spending and the potential for savings.As you will notice from the IDC chart that Matt presented - Power & Cooling, Management and Administration, and Server Infrastructure Spending form the bulk of IT operations expenses. In the last few years, companies have been focused on attaining , and they have indeed achieved,  efficiencies when it comes to reducing the hardware and power and cooling costs. As virtualization has become more main stream, it has been a major disruptor to traditional datacenter operations and helped curb physical server sprawl, but not the management costs - at least, not yet. This provides a significant opportunity for IT operations for consolidating management and achieving significant savings by focusing on operational efficiencies.

And then there are Cloud Computing models. As Cloud models gain IT attention and traction, management of cloud infrastructure and cloud resources become more critical for sustaining and exceeding the service level objectives and the operational efficiencies they demand. Service providers worldwide are dealing with the issue by consolidating the management of Cloud resource. The ability to integrate management across your entire infrastructure to realize greater efficiencies—servers, networks and storage—is becoming a measurable success factor for IT organizations. The notion of cloud management "platforms" is starting to emerge with respect to external and internal cloud-computing environments. Vendors offering cloud management technologies span the spectrum of technology providers. They are either the Big 4 IT Service management platforms like BMC, CA, HP and IBM or management platforms provided by virtualization vendors like VMware, Microsoft or emerging cloud management players. Ideally, you need to monitor and manage everything - applications, servers, network and storage from a single interface where efficiency measures need to start. Storage vendors, however, have focused more on developing feature sets of their own management tools than they have on integration with other higher level management tools. As a result, Service providers often have to use separate proprietary tools to take full advantage of advanced storage features. This increases management complexity, reduces efficiency, and may require orchestration between the different domains and additional, expensive staff training.

Throughout all of this, NetApp’s approach continues to advance our management tools including the significant efforts to integrate them with the overall IT ecosystem making them continuously vital and relevant. NetApp’s focus and objective of the integration strategy is to give existing orchestration and management platforms the same level of access and control. NetApp’s open management interfaces (API’s, SDK’s and plug-ins and adapters) offer fast integration and greater storage abstraction that leverages our policy-based automation features to better integrate with customers’ existing management platform including virtualization management tools,  homegrown management tools, IT service management and orchestration tools. See diagram below for graphical representation.

To achieve significant operational efficiencies, IT enterprises and service providers should pay more attention to the integration efforts undertaken and delivered successfully by their IT vendors. Having a good integration strategy with demonstrable proof points will increasingly become a pre-requisite to be a serious player in this new rapidly evolving cloud environment. As I end this blog post I see that BMC has just released their Atrium Orchestrator Adapter for NetApp storage. The enabling technology- you already guessed it - NetApp’s Open management interfaces.

VS Joshi
Sr. Manager, NetApp Managability

Storage Multi-Tenancy: Is it safe?

Many movie buffs (and reluctant dental patients) remember Laurence Olivier’s infamous quote from the 1976 production of “Marathon Man”-Is it safe?  I was reminded of this by a recent skeptical comment in the storage blogoshere which questioned the security provisions of our MultiStore software product. This comment was made in response to our best of breed Secure Multi-Tenancy (SMT) reference architecture which we announced jointly with Cisco and VMware on January 26. And although the inability to answer that question resulted in one heck of a toothache for Dustin Hoffman, we are very confident in saying-Yes, it’s pretty darn safe.

 

Since its inception (in 2002), MultiStore has incorporated the ability to not only partition a single storage system into numerous virtual storage containers but also to logically isolate each tenant such that its data is inaccessible to other tenants. Having said that, we know that there are some skeptics out there-so, in 2008, we commissioned an independent security audit to see if we could stand up to a series of attacks on MultiStore’s security. The company that conducted this audit, Matasano Security, put together a threat model which included attacks on a high security environment in which customer protected information was stored. Matasano noted that an attack on MultiStore’s security would be considered successful if attackers could:

 

• Access storage resources provisioned to a high security storage partition from any other storage partition
• Run commands in a high security storage partition
• Run configuration commands or access the configuration registry of the core storage system
• Execute and run code in the Data ONTAP kernel, across all storage partitions
• Use the network stack in Data ONTAP to route traffic across storage partitions

At the conclusion of the testing, Matasano confirmed that none of the attacks on MultiStore’s security were successful.  And we have lots of enterprise accounts that have been experiencing the same result, so we are confident that we picked the right technology to align with the Cisco and VMware components of our joint SMT design.

 

If you’d like to read the full Matasano report, you can find it here. And in case you’ve not yet seen “Marathon Man”, you can find a good review of it here.

 

-Paul Feresten

Evaluation of NetApp Ethernet Storage

NetApp sponsored an evaluation of our Ethernet Storage, conducted by Demartek LLC in our Research Park, NC lab and their Arvada, Colorado lab. The results were published in January 2010. The full 14-page report can be found in NetApp’s external AR Library. http://www.netapp.com/us/library/research-papers/rp-unified-networking-evaluation.html 

The report concludes:

 

NetApp is the first storage vendor to support traditional Fibre Channel, FCoE, iSCSI, NFS and CIFS in a single storage controller and high-availability system. Now storage administrators, datacenter managers and CIOs can deploy a storage product that supports all of today’s needs, is compatible with today’s technology, and will support the needs of future “converged” networks that support shared infrastructure architectures.

 

The evaluation took place on the NetApp FAS3170 handling multiple file and block protocols simultaneously on the same 10GbE converged networking connection. FCoE connections were driven at near maximum rated speed. The report includes an overview of the FCoE technology, along with details on the evaluation environment, test process, and results of IOmeter tests performing random and sequential reads and writes to LUNs.

Demartek concludes that the FAS3170 system proved to be a versatile storage system, handling multiple file and block protocols simultaneously including FCoE, SCSI, NFS, and CIFS on the same 10GbE converged networking connection. Managing the FCoE connection was exactly equivalent to managing a traditional Fibre Channel connection, making adoption of the FCoE technology seamless.

Mike McNamara, NetApp Product Marketing

 

SNIA Webinar: DAS to SAN: iSCSI Offers a Compelling Solution

The recently formed iSCSI Special Interest Group (SIG) from SNIA is hosting a webinar on the value of moving to iSCSI for your first storage area network (SAN).  Speakers from NetApp and Dell will discuss the benefits of migrating from a direct attached storage (DAS) environment to an iSCSI SAN.

This webinar is perfect for customers considering deploying a SAN for the first time, or for those considering a a migration to iSCSI from another storage networking environment or as an enhancement to your existing infrastructure.

The event will be held on February 4th, at 2pm ET.  For free registration, follow this link http://www.brighttalk.com/webcasts/7533/attend.

- Jason Blosil

Total Cost Comparison Study for FC SAN Environments

Oliver Wyman, a global strategy consultancy, was engaged by NetApp to conduct primary research with IT decision makers to determine the total cost of acquiring, deploying, operating, and managing storage environments from various vendors to support Microsoft Exchange and SharePoint. The principal focus of the study was a comparison of the total cost of enterprise-level NetApp, HP, and EMC storage products used for Microsoft storage environments, as well as potential advantages associated with virtualization of the application servers.  

The following are three reasons identified by the study participants who included storage administrators and IT managers from more than a dozen mid-sized and large enterprises across various industries and geographies on why NetApp solutions provide a total cost advantage over comparable competitive deployments.

· NetApp environments have lower up-front costs because NetApp systems provide greater storage efficiency.

· NetApp environments experience lower internal operational costs resulting from easier management of the storage environment and lower power, cooling and space costs.

· NetApp software provides 2x faster recovery in the event of application errors.

The detailed report is located here: http://media.netapp.com/documents/ar1048.pdf

Mike McNamara, NetApp Product Marketing

No Trouble with Tribbles ...

What to Do When VM Growth is Consuming Your Storage

Fans of the original Star Trek television show might recall an episode called "The Trouble with Tribbles".  Captain Kirk and the Enterprise crew are carrying bins of grain intended for a planet in a state of famine. Along the way, they stop at a space station where a trader gives Ohura an irresistible pet called a tribble and she makes the mistake of taking it back to the Enterprise. It turns out that her new pet is a pregnant member of a species that reproduces rapidly when given access to a plentiful food supply. Soon the Enterprise is overrun with tribbles and the grain has disappeared.

Virtual machines (VMs) can seem like tribbles when you are managing storage. It certainly felt that way to one IT team whose VMware environment grew faster than expected and free space on their storage system fell to only 8%. Their problem was not just the rapid growth in VMs; it was also in the way they stored and backed up their data.

This IT team was doing things in a business-as-usual manner. They had a midrange FC-SAN disk array from a major storage vendor. LUNs were allocated to virtual servers and a backup copy of each LUN was kept on the same disk array. It seemed like a good approach when they started but running out of storage capacity caused them to reconsider. The IT team needed a more efficient way of storing VMware boot LUNs and data as well as a better approach to backup. So, they invited NetApp to do a proof of concept test with a V-Series open storage controller.

The first step was to configure an alternate path from the virtual servers to the FC-SAN disk array.  (See Figure 1). LUNs were allocated from RAID groups in the disk array to the V-Series controller in the same way they would be to hosts. Multiple LUNs from the disk array were then striped into a NetApp aggregate across which NetApp LUNs were defined and assigned to the VMs. Several aggregates were created.

NetApp aggregates made it possible to spread data across a larger number of disks than contained a single RAID group and to use spare capacity more efficiently. The NetApp LUNs were created using our thin provisioning feature (FlexVol) which lets you expand or decrease the size of a LUN or volume as needed.

The second step in this proof of concept was to migrate the data from LUNs native to the disk array into NetApp LUNs. Storage VMotion, a VMware utility, was used for this task. It was accomplished in a stepwise fashion because there was so little spare storage available to use. Using a NetApp disk shelf as temporary "swing storage" behind the V-Series controller would have made the job easier.

As a side note, it is worth pointing out that data migration is not an all-or-nothing proposition with the V-Series. You can leave the data for some applications in LUNs that are native to the disk array and continue accessing them via the original FC-SAN path. You can also choose to migrate no data at all and simply deploy new applications using capacity managed by the V-Series controller.

The third step in this exercise was to shrink the storage capacity used by the VMs. Thin provisioning was a good start but the big savings came next. Using NetApp deduplication, the team eliminated redundant blocks within each aggregate for a savings of 86% on the boot LUNs. Deduplication savings are particularly high with virtual servers and desktops because the blocks of boot LUNs are largely the same. Deduplication savings for VM data LUNs varies depending on the data type.

The final step in this proof of concept was backup. A full copy of each LUN had been stored before the V-Series was introduced to the configuration. Now the IT team could use NetApp Snapshot "copies", which are space efficient, time efficient, and performance efficient.

The combined savings from NetApp thin provisioning, deduplication, and snapshots increased free space in the disk array from 8% to 67%, as shown in Figure 2.

Improving storage efficiency was just the start for this IT team, which purchased the V-Series controller and evaluated more ways to improve their storage infrastructure.

Later, they implemented a simple, cost-effective scheme for disaster recovery by mirroring data between the V-Series controller and a NetApp FAS2050 in another location with SnapMirror software. The process was fully automated and the IT team saved on WAN bandwidth costs because they were transferring deduplicated data from the primary storage system to the DR site.

Another possibility for this IT team is to use their DR site for the development and testing of VMware applications. NetApp FlexClone software will enable them to instantly create space-efficient clones of production data for every developer, so they can speed project completion without compromising quality.

Perhaps the best part of this story is that the IT team was able to protect their existing storage investment while adding new capabilities. By reclaiming inefficiently used storage and reducing the ongoing rate of consumption, they were able to cancel disk drive purchases for many months. This yielded savings that offset the cost of buying the V-Series controller. 

Getting back to the Star Trek theme, I'd like to summarize by amending a few lines from opening of the original TV show:

Storage: the final virtualization frontier. This was the voyage of one IT team. Its mission: to use existing storage more efficiently, to seek better ways of protecting data, and to confidently travel into a new storage future.

- Mark Woods

Coping with an Avalanche (e.g. Boot Storm)

One of the hazards of a virtualized computing environment is the dreaded boot storm, where so many virtual machines are trying to boot at one time that the storage system containing their files becomes saturated. Tens of minutes can pass before the last virtual machine is up and running.

The delays associated with these storms can be greatly reduced by using deduplication in combination with solid state caching (e.g. Performance Acceleration Modules) in a NetApp FAS or V-Series storage system.

Here's why the combination is so effective. NetApp deduplication typically reduces the storage footprint of VMware, for example, by 60% to 80%. This is because most of the blocks in one boot file are identical to those in the next boot file of the same operating system type. NetApp deduplication changes the WAFL metadata pointers to redundant blocks so they all point to one master block. (The extra blocks are freed for reuse.) The first time a master block is read it will be placed in the cache of the storage controller. Then every virtual machine with this block in its OS file can get quick access.

Corporate Technologies, Inc. (CTI), a NetApp partner, ran a series of tests in a production environment to determine how much deduplication and solid state caching can reduce the duration of a boot storm. Their environment had 50 virtual machines (or "guests") for each of four operating system types, totaling 200 virtual machines.

The CTI tests showed a 71% reduction, from 15:09 (min:sec) to 4:27, for all 200 machines to boot when the combination of deduplication and caching was used. I suspect the results would be equally dramatic in a virtual desktop infrastructure (VDI) with thousands of clients booting as employees stream into the workplace each morning.

You can read more about the tests run by Corporate Technologies in Jesse St. Laurent's blog post titled VMware Boot Storm on NetApp.

- Mark Woods

Hypervisor Agnostic Data Mobility

In my recent blog about data mobility and cloud infrastructure, I mentioned a new NetApp technology called Data Motion and how it provides a “large grain” companion to VMware’s Storage VMotion. Since then, there have been a few questions regarding Data Motion’s fit with other virtual server environments such as Citrix XenServer and Microsoft Hyper-V. So-let me clear that one up. NetApp Data Motion operates at the storage layer, so it’s basically Hypervisor agnostic.

In the last blog, I pointed to the Data Motion demo that we used at VMworld. For a more general overview of how we address the needs of shared storage environments for both physical and virtual infrastructure, check us out our new Data Motion video on YouTube.

Paul Feresten

Our CTO’s View on Cloud

Cloud computing represents a massive
shift in IT Service delivery. New or
emerging companies may never again
employ full-time IT staff. As they grow
into mid-sized and large enterprises they will not have
the same ‘fixed’ IT burden as their more mature counterparts
and this could represent a competitive advantage.
Cloud computing will continue to evolve. It may even
get a name change somewhere along the line, too.
We actively review our portfolio and ask — what’s
core to our business? What do we value? For example,
architecture, business analysis and project management
are considered core, but not Level 1 or 2 systems
administration or network operations.
Now is an exciting time to be in the business. As cloud
evolves, organizations will have to look at IT and service
providers differently. Shifting from fixed to variable
cost is one of the key attributes of software as a service.
IT has to recognize this as an opportunity, because
business will buy this whether they like it or not.
Part of an IT organization’s skill is how effectively it can
select applications, source and procure services and
handle contract/SLA management, as well as work with
open authentication standards.
There is reluctance among CIOs to put the entire enterprise
on the cloud, which is as it should be. The challenge
is to think differently, focus on open standards,
and change the mix of what we buy to deliver value
more quickly to the business while not committing to
large fixed budgets to support and maintain services.
As to security and data protection issues, there are
things you need to think about but no real showstoppers.
Businesses will need to engineer the right level of
security, e-discovery and compliance into their agreements
and hold providers to SLAs. In the US, The
concern over fourth amendment rights is a bit trickier,
but I believe it’s overblown.

 

Dave Robbins
CTO, Information Technology at NetApp, Inc.

NetApp demos converged Ethernet storage traffic with FCoE at SC09

NetApp will be participating in a multi-vendor demo sponsored by the Ethernet Alliance at SC09 the week of November 16, to demonstrate the performance and efficiency of a unified and converged data center fabric using 10Gb Ethernet.  Specifically, NetApp will be showcasing file and “block” storage network protocols running on a single unified 10Gb Ethernet wire, including Fibre Channel over Ethernet (FCoE), iSCSI, and NFS. NetApp is the only storage vendor offering FCoE storage in the market today, and also supports NFS, CIFS, iSCSI and FC SAN protocols, including FCoE, on the same unified storage platform.

Enhancements to the Ethernet specification, defined as Data Center Bridging (DCB) by IEEE or Converged Enhanced Ethernet by other hardware vendors, include features that enable lossless network characteristics. These lossless characteristics are necessary to support Fibre Channel traffic over an Ethernet physical transport. DCB allows for the prioritization of data traffic as well as congestion management to avoid the dropping of data packs over the network. DCB will also offer benefits to IP traffic by reducing congestion conflicts and supporting prioritized data traffic. As a result, DCB will offer the ability to share network bandwidth with multiple data types and protocols.

The focus of the NetApp demo is to show bandwidth allocation across multiple data types and protocols as well as show how the pause mechanism in DCB functions. NetApp will demonstrate iSCSI and NFS sharing bandwidth with FCoE traffic on the same wire using NetApp’s recently announced Unified Target Adapter configured in a FAS3000 storage system.  iSCSI and NFS traffic will be set at priority 0 and FCoE traffic will be set at priority 3. Bandwidth at the target is set at 50/50 by default, but can be overridden by settings at the switch.  The demonstration will show how modifying priority and bandwidth settings at the switch are reflected in settings changes at the storage system. 

Moving to a unified 10Gb Ethernet fabric offers the ability to converge storage and data traffic on the same wire which reduces overall network infrastructure requirements, including cables and switch ports.  FCoE compliments iSCSI to offer additional choice for “block” storage traffic. And because it doesn’t require tunneling, FCoE allows you to easily connect into your existing FC network to extend the life of your existing FC storage until you are ready to upgrade.

As IT departments look to improve the efficiency of their data centers, DCB with support for FCoE will offer improved efficiency by reducing the size of the network infrastructure.  Moving traditional FC traffic off of a dedicated, parallel network to a shared 10Gb Ethernet network  and unified storage system removes the requirement for single purpose switches, storage and redundant cabling for FC traffic alone.  The reduction in switch infrastructure, storage and cabling not only lowers capital expenses, but reduces hardware foot print which results in lower power consumption and reduced airflow restrictions in order to improve overall data center efficiency.

For more information on the Ethernet Alliance demo, visit www.ethernetalliance.org.

For more information on FCoE and NetApp solutions, click here,or go to www.netapp.com.

-- Jason Blosil