EMC aims to define software-defined storage with ViPR

LAS VEGAS -- EMC Corp. is dedicating EMC World 2013 mainly to defining and setting a high bar for its software-defined storage vision.

In 2012, EMC announced more than 40 new or upgraded products. There have been only eight product releases this year, and only one can be considered major: the ViPR software-defined storage platform, which the firm launched Monday.

EMC is pitching ViPR as an enabler of Google-like storage that allows customers to use commodity hardware to reduce cost and complexity, things EMC is not exactly known for.

EMC executives admitted the term software-defined storage can be a tough sell because it means different things to different people. Other storage vendors have pushed the concept without much success or even agreement on what it is. At times, it seems little more than a new take on storage virtualization.

During a session with media and analysts Monday at EMC World, Amitabh Srivastava, president of EMC's Advanced Software Division, actually drew applause when he said, "The term software-defined storage is overused. Nobody knows what it means. People are using work they've done in the past and recasting it as software-defined storage. Some people think that using software to manage commodity storage is software-defined storage."

To sell ViPR, EMC will have to convince customers it can make good on its low-cost, low-complexity promises rather than selling them on the software-defined storage term. Although, some customers don't have to be sold because they have been waiting for a successful form of storage virtualization for years.

Michael Leeper, director of global technical infrastructure at Portland, Ore.-based Columbia Sportswear, said the concept of software-defined storage makes sense to him. He also said it is the obvious next step after virtualized servers and software-defined networking.

"It's the third one in the game," he said. "If you can't figure out what it means by now, you're in trouble. We had software-defined compute with VMware, last summer the talk was all software-defined networking, and now the obvious progression is to software-defined storage. I get it. We're going to separate the intelligence in the storage from the hardware."

Leeper said Columbia has more than 1.5 PB of EMC storage under management at four data centers. He uses VMAX, VNX, Isilon and VPLEX Geo for primary storage and Data Domain for backup. He said Columbia committed to VMware approximately three years ago, and its servers are 98% virtualized.

Columbia doesn't have much non-EMC storage to virtualize, and Leeper said he has no real use for the object services that will be a big piece of the first ViPR release, but he's interested in separating the hardware layer from software in storage the way he does with servers. Columbia will be an early beta customer for ViPR, he said.

"We are used to being able to move our workloads in a compute stack back and forth, up and down, left and right, in and out; all sorts of things," Leeper said. "But we found our flexibility is hindered by two things. One is the network; networking configurations don't move with my VMware nodes. Obviously, there is a reason VMware is moving into that space [with the Nicira acquisition]."

He continued: "Then we started looking around and said, 'Wait a minute, what are we doing for storage?' EMC finally came to us with the [ViPR] stuff and said, 'Here's our option; here's where we're going.'"

Leeper said the main benefit will be, "as we decide we need to move workloads, the identities and configurations of each workload travel with them; that could be within my data centers, between my data centers, or ultimately between my data center and someone else's data center."

Columbia has a large private cloud configuration today, Leeper said. By moving to "someone else's data center," he's referring to a public cloud. He said his company doesn't use a public cloud today, but he would consider it for specific applications.

He said another benefit of ViPR would be, "We start getting out of these vendor games between what makes my hardware better than your hardware. We try to figure out what makes my application perform better. As we look at the storage array, we start seeing a lot of the same physical components in these arrays, but at the core, what's differentiating one array from another is that software layer."

But EMC executives insist that the differences in arrays won't go away. They say each storage hardware platform can maintain competitive advantages even if ViPR allows a common set of software services across them.

"If the VNX is twice as fast as competitors' arrays, we feel people will still see value in [it]," said Eric Herzog, EMC's senior vice president for product management for unified storage. "ViPR provides high-level management orchestration, but there are all kinds of things we can touch in Unisphere [VNX management software] that ViPR can't see. Even with ViPR, integration with applications like Oracle and SQL will still be controlled by the array."

EMC's Srivastava added: "One size just doesn't fit all. There are reasons for different storage systems with different properties."

EMC President Dave Goulden called ViPR "revolutionary" during a Monday EMC World keynote.

"A lot of people have a piece of software that runs on hardware, that's not what this is," Goulden said. "We believe this is the way everybody should do software-defined storage."

Greg Schulz, a senior analyst at Storage IO, said he thought EMC execs showed constraint by not overdoing the use of the software-defined storage term while explaining the underlying technologies.

"They realize they're still conditioning the market," Schulz said. But he said EMC's success with ViPR will depend on factors such as how well it drives adoption of its open application programming interfaces, how it prices the technology, and how well it explains the value of ViPR compared to other EMC technologies, such as the VPLEX federated storage platform.

View the original article here

Read More

Denver Broncos score with NetApp filers for primary storage

When the Denver Broncos play at home, more than 76,000 fans can watch Peyton Manning and his teammates from their seats in Sports Authority Field at Mile High. What they might not realize is that their National Football League experience is more virtual, digitized and cutting-edge than ever.

Entertainment video for the scoreboard, playbooks

accessed field-side by iPads and video for post-game film study are among the IT requests that need to be available at a moment's notice.

"We're a small shop and almost everything we do is being digitized -- whether it's a server becoming virtualized, entertainment video for the scoreboard, player video or pictures. We have a lot of pictures from the history of the franchise, [and] they're all getting digitized," said Russ Trainor, vice president of IT for the Broncos. "So from a storage standpoint, it's really exploded for us."

The Broncos' IT department is responsible for two data centers: one at the stadium and the other at the team's practice site in Dove Valley, Colo. The department has 320 people responsible for handling 100 TB of data across the two sites. Two entry-level NetApp Inc. FAS2240 filers, one for each site, were purchased in July 2012. Chris Newman, an IT architect for the team, explained that having identical storage was essential to making everything easier to manage. The two arrays back up data to each other.

The Broncos began using NetApp filers in 2004 when they bought four FAS270s to store 1.4 TB of data. But because of budget constraints, the two environments evolved to have different NetApp filers: a FAS2040 at Sports Authority Field at Mile High, and a midrange FAS3140 at Dove Valley.

"We used to have a 3140 at the valley site, which actually was kind of an overkill for us," Newman said. "[The] 2040 we had here at the stadium was working fine, but we really wanted to get into that 64-bit platform and have it at both sites so we could eventually take advantage of clustering."

The Broncos' servers are 95% virtualized with VMware vSphere, and connect the two FAS2240s via Cisco Fibre Channel switches. They run Microsoft SQL Server 2008 and one Exchange 2010 server (with plans to add another in a couple months), and handle 46 virtual machines. The environment is managed through NetApp's OnCommand Unified Manager.

Trainor said the team sought a new system that could be clustered and support 10 Gigabit Ethernet to keep up with expanding virtualization. He said the Broncos bought the NetApp filers without testing any other products. "It's been steady, reliable and the 2240 gives us that performance bump as well," he said. "The 10 Gigabit backbone provides us with a nice growth path."

The Broncos also use the NetApp filers for disaster recovery. Backups are performed nightly using NetApp's Snapshot and SnapRestore technology, so there are full data copies of everything at both sites. Additionally, the team is saving as much as 25% of its capacity using NetApp's primary data deduplication feature.

Trainor and Newman both said they are looking forward to getting another FAS2240 for the 2013 season, which will be provided by a recent independent deal between the NFL and NetApp. It will give them an additional 120 TB of capacity that Trainor said will be used at the Dove Valley site primarily to store coaching video.

View the original article here

Read More

Storage considerations for a Hadoop implementation

Can SAN and NAS storage systems be used as the primary storage layer with a Hadoop implementation?

The short answer is yes. The longer answer is that, depending on the size of your Hadoop implementation and the number of nodes, it may not make sense from a cost perspective to use SAN and network-attached storage (NAS) systems as the primary storage layer.

A good place to start is by understanding what Hadoop is. Simply put, it's an Apache open source project that allows users to perform highly intensive data analytics on structured and unstructured data across hundreds or thousands of nodes that are local or geographically dispersed. Hadoop was designed to ingest mammoth amounts of unstructured data and, through its global file system (Hadoop Distributed File System), distribute workloads across a vast network of independent compute nodes to rapidly map, sort and categorize data to facilitate big data analytical queries.

Hadoop was also designed to natively work with the internal disk resources in each of the independent compute nodes within its clustered framework for cost efficiency and to ensure that data is always available locally for the processing node. Jobs are managed and delegated across the cluster farms, whereby data is parsed, classified and stored on local disk. One block of data is written to the local disk, and two are replicated for redundancy. Data copies are readable so they can be used for processing tasks.

In determining the answer to the question, we have to consider that Hadoop is disk-agnostic; that means SAN and NAS resources can be used as the primary storage layer to service Hadoop workloads. But a natural follow-up question is this: Will SAN and NAS storage be the most cost-effective way to deploy Hadoop? If a Hadoop implementation will be confined to one or two locations, the benefits of managing a centralized storage resource should make sense, especially if an existing array has already been depreciated.

But if your firm's Hadoop implementation will consist of many hundreds or thousands of compute nodes across multiple data centers, the costs for deploying multiple SAN or NAS systems may prove to be too costly. The bottom line is that Hadoop's native support for internal disk resources gives data center managers a lot of options for leveraging either existing SAN or NAS assets, or low-cost, commodity disks (or a combination thereof) to support big data analytic workloads.

About the expert
Colm Keegan is an analyst at Texas-based firm Storage Switzerland LLC and has been in the IT industry for 22 years. His focus is on enterprise storage, backup and disaster recovery.

This was first published in May 2013

View the original article here

Read More

Fusion-io buys NexGen Storage, adds hybrid arrays

Server-side flash vendor Fusion-io is moving into the storage array business with its $119 million acquisition of NexGen Storage, which uses Fusion-io PCI Express cards along with hard disk drives in its n5 hybrid systems.

Fusion-io revealed the acquisition Wednesday evening during its quarterly earnings report. It is the third -- and largest -- acquisition Fusion-io has made since becoming a public company in 2011. It paid $95 million for caching software startup IO Turbine in August 2011 and an undisclosed price last month for ID7, which developed and licenses the open source SCSI Target Subsystem (SCST) that enables companies to turn a Linux box into shared storage.

NexGen came out of stealth in 2011, started by iSCSI vendor LeftHand Networks founders John Spiers and Kelly Long. NexGen storage systems use flash, memory and hard drives for storage, and have a performance quality of service feature that allows IT administrators to guarantee performance to each application and isolate workloads on the SAN.

Fusion-io CEO David Flynn said NexGen brings his company into the small to medium enterprise market as a standout from Fusion-io's current enterprise-focused products. But although NexGen sells systems, Flynn said its software is a key piece of the acquisition and fits into Fusion-io's software-defined storage strategy.

"NexGen was a clear acquisition choice for us," Flynn said on Fusion-io's earnings call with analysts. "To be clear, we are not getting into the box business. As always, we focus on adding value in software. NexGen Storage will be taken to market as an open platform, as opposed to a closed system."

Flynn has also pledged to keep ID7 technology open, allowing other vendors to use it to build flash storage systems.

"Customers will decide whether they want to continue their dependence on closed systems or move towards an open platform that gives them choice," Flynn said. "We believe open platforms will win over closed."

Flynn characterized NexGen's software, which allows customers to run multiple workloads with different performance requirements on the same system, as "horizontal application acceleration." That is different than Fusion-io's ION Data Accelerator virtual appliance, which he calls "vertical application acceleration" because it deals with applications large enough for dedicated infrastructure.

Flynn said Fusion-io will make NexGen technology available to its OEM partners who sell its PCI Express (PCIe) cards, but no deals are in place yet. Having a storage product also makes Fusion-io competitive with its partners. Fusion-io listed Hewlett-Packard, Dell and IBM as 10% customers last quarter, meaning those vendors sold a significant amount of Fusion-io products on their servers. Cisco and storage vendor NetApp also have OEM deals to sell Fusion-io cards.

All of those partners, except for Cisco, sell storage systems that compete with NexGen. But the flash market is rife with competition as vendors move to cover as many bases as possible. For example, ION Data Accelerator can be seen as competitive with IBM's FlashSystem (formerly Texas Memory Systems) all-flash arrays, and IBM's pledge to spend $1 billion to develop flash products will likely roll out other products competitive with Fusion-io. Server vendors can also sell other PCIe cards that compete with Fusion-io.

"Chances are, Fusion-io read the tea leaves and saw some of their OEMs doing more with their own NAND flash SSD offerings or with other suppliers," said StorageIO Group analyst Greg Schulz. "So if you can't beat 'em, join 'em.

"Can Fusion-io turn NexGen into an OEM model? Perhaps in some sectors," Schulz noted. "However, it also could help transition it into a more direct model or give their VARs something else to work with. Maybe we are seeing the formation of Fusin-io 2.0. The question is no longer if flash is in your future, it's how much, where, when, with what and whom?"

Fusion-io lost $20 million last quarter as its revenue of $87.7 million declined 7% from last year. The vendor forecasted the loss last quarter, saying its largest customers, Facebook and Apple, would decrease spending. Even with the decline, Facebook was Fusion-io's largest customer last quarter, accounting for 19% of sales.

Fusion-io CFO Dennis Wolf said revenue outside of Facebook and Apple came to $68 million, up 61% from last year.

View the original article here

Read More

Software-defined storage: Answering frequently asked questions

A new concept and a niche product just a couple of years ago, software-defined storage has found its way into the mainstream data storage market as most storage vendors are now offering some type of software-based storage networking product.

Software-defined storage (SDS) can be difficult to understand. It's often lumped with other products under the umbrella term

storage hypervisor, and there is also a hardware aspect to consider.

We asked Russ Fellows, a senior partner at Boulder, Colo.-based Evaluator Group, to step us through some of the basic concepts related to software-defined storage.

What is software-defined storage, and is it the same as a storage hypervisor?

Russ Fellows: SDS is similar in concept to other software-defined elements of the data center, such as software-defined networking (SDN).

In many respects, this transformation is more about packaging and how IT users think about and design their data centers. The reality is that storage has been largely software-defined for more than a decade, with the vast majority of features designed and delivered as software components within a specific storage-optimized operating environment.

SDS is sometimes referred to as a storage hypervisor, although the two concepts are somewhat different. Both terms are evolving, with multiple vendors using these terms for different aspects of their storage systems.

In many ways, a storage hypervisor is part of software-defined storage, with a storage hypervisor the core element of an entire storage software stack. Again, vendors apply the term differently, so its meaning is not yet standard by any means.

What benefits does software-defined storage provide?

Fellows: The primary benefit of software-defined storage is separating the hardware from the software. This provides flexibility for configuration and the ability to drive down hardware costs by substituting equivalent hardware platforms. The issue with this model is that support can become more problematic, which is why existing software stacks are often only qualified and supported on a limited set of hardware.

How much of a DIY project is it to implement SDS? Do you need special expertise?

Fellows: While separating the hardware and software provides more flexibility, and creates unique configurations, IT may find they will need to pay special attention to support. Another option is to purchase a preconfigured system through a reseller. The ability to customize a product and offer support is appealing both to resellers and many businesses and IT staffs. While it's possible for adept IT professionals to configure a system, the support issues for an in-house system may not make it a worthwhile endeavor. This is the reason we see vendors deliver SDS on pretested hardware for which they can provide support services for end users.

Who are the top software-defined storage vendors?

Fellows: As I mentioned, the technical reality is that software-defined storage has been prevalent for more than a decade, with vendors such as NetApp being among the first to offer it. Today, many of the top storage vendors, along with a number of startups, are all leveraging SDS.

Currently, the ZFS software stack is one of the most readily available SDS options, along with companies that have commercialized software stacks such as Nexenta and others. A new wave of proprietary SDS stacks is also emerging, such as those from GreenBytes and Virsto, which was acquired by VMware. Traditional vendors have also been selling software-defined storage for many years, such as IBM with SVC, NetApp and HP's SDS version of its LeftHand product, which it now calls StoreVirtual VSA.

Will the SDS vendors provide adequate support even though they don't provide the actual hardware being used?

Fellows: That's hard to say. IT must carefully examine the support plan provided by the vendor and understand what is in their control and out of their control. We expect most software-defined storage products to provide approved configurations for installation, which would mean they have tested the configurations in their lab and will support those configurations. You'll also find system vendors, like Dell, HP, IBM and Oracle releasing software-defined storage with their hardware. Again, in those cases you can expect that the hardware and software has been tested and is thus supported.

This was first published in May 2013

View the original article here

Read More

Dell DX Object Storage Platform dumped in favor of software-only model

Dell is discontinuing its DX Object Storage Platform appliance, which includes software from Caringo integrated with select PowerEdge servers. Dell will instead sell the Caringo CAStor

object-based software as a digital download that customers can run on any of the vendor's servers.

Sales of the Dell DX Object Storage Platform began in 2010 through an OEM deal with Caringo. The DX consisted of Caringo CAStor software bundled with Dell PowerEdge r500 and r700 servers, and was used mainly for archiving and compliance.

The news of Dell's DX object storage replacement first came in a press release from Caringo announcing the software-only model.

In an email response, Dell spokesperson Lon Levitan confirmed the DX integrated appliance will be discontinued. He said customers have been asking for more flexible hardware options to deploy object storage.

Brandon Canaday, Caringo's vice president of strategic accounts, said customers often wanted different hardware configurations than the PowerEdge r500 and r700. Canaday said the integrated DX software and hardware storage nodes were limited to a 2U form factor, supporting up to 12 drives and 36 TB of capacity. However, Dell has other chassis with higher density.

The Caringo CAStor object storage software is designed to work on standard x86 architectures.

"Many customers have a broader set of configurations that are not supported by the r500 and r700," Canaday said. "The appliance model is built on the assumption that one size fits all and in the market we serve, that is not the case. We need to be more flexible to meet our customers' needs."

Dell and Caringo currently are working on a timeline for the transition, which is expected to be completed on July 31.

"In the future, our software will be configured based on the use case," Canaday said.

Dell will continue to support DX systems sold, which apparently is not a high number.

"If you're interested in how many DX6000 customers we have, I'd say it's a fraction of one percent of our Compellent and EqualLogic customer base," Dell's Levitan wrote.

Dell's storage division has gone through recent management changes. Former Storage Chief Darren Thomas left the company in December. Alan Atkinson, vice president and general manager of Compellent, and Pete Korce, vice president and general manager of EqualLogic and NAS Storage Solutions, are leading Dell's storage business. Michael Dell and Dell's directors are also attempting to take the company private.

View the original article here

Read More

HP StoreVirtual expands into FC, new storage blades

Hewlett-Packard's LeftHand iSCSI platform, rebranded as StoreVirtual in November 2012, is now classified by HP as its software-defined storage

line. HP expanded that platform today by adding a StoreVirtual 4630 and the D2220sb Storage Blade while adding Fibre Channel connectivity to StoreVirtual arrays released in early 2013.

The HP StoreVirtual 4630 is a scale-out block storage system based on the LeftHand operating system. One blade is connected to one external 2U JBOD, and it scales to 11.24 TB per rack unit and 180 TB within an HP BladeSystem chassis. It supports 6 Gb Serial Attached SCSI (SAS) drives and HP ProLiant Gen8 server technology.

Dale Degen, worldwide senior marketing manager for HP Storage, said the 4630 blade device offers a 50% smaller footprint and double the density compared to when it was known as the P4800 G2 SAN BladeSystem.

"The previous P4800 footprint is two blades in a chassis and a 5U JBOD," Degen said. "This is a smaller footprint and more density. This is for customers looking for more capacity and performance in a blade for shared storage in virtualized environments."

The 4630 supports iSCSI and is VMware-certified for multi-site disaster recovery. It also supports LeftHand Remote Copy for asynchronous replication, thin provisioning and snapshots between primary and remote locations, along with LeftHand Peer Motion to move volumes across all HP StoreVirtual nodes.

The D2220sb Storage Blade is bundled with HP's StoreVirtual VSA virtual storage appliance. It uses the HP Smart Array controller, and is designed for customers who don't want to purchase external storage hardware. The blade supports 12 drives, which can be SAS or SATA hard drives or solid-state drives. Eight D2220sb storage devices can be supported in a single BladeSystem c7000 enclosure to support up to 115.2 TB of capacity.

"The D2220sb brings storage inside the enclosure with StoreVirtual VSA," Degen said.

HP added FC connectivity to the StoreVirtual 4330 and 4730. It announced FC support for the 4330 upon launch but is now shipping.

The 4330 FC system has a starting price of $28,000, while the 4730 FC costs $63,000 and the 4630 FC, $64,000. The D2220sb costs $5,999.

View the original article here

Read More

EMC World 2013: Google storage coming to your data center

LAS VEGAS -- When planning your next data center, EMC Corp. wants you to think along the lines of what Google, Amazon and Facebook have done with theirs. That was a key message from the vendor at EMC World 2013, where EMC offered its ViPR software-defined storage platform as the first major step in the direction of a Google storage model.

"You can build a Web-scale data center without hiring 1,000 Ph.D.s or rocket scientists," said Jeremy Burton, EMC's executive vice president of product operations and marketing. "The Googles, Facebooks and Amazons have built a data center for one -- just for their environment. We're building a data center for everyone."

"We built it for people who want to run their data center like Google, but [who] don't want to write and maintain their own custom environment," EMC president Dave Goulden said of ViPR.

Building Google storage requires Hadoop or similar frameworks that Google, Facebook and Amazon use. EMC promises ViPR will manage data as a pool across any type of storage and handle Hadoop workloads. It has also given its Isilon clustered network-attached storage platform tight support for the Hadoop Distributed File System.

But EMC's message about a new data center goes beyond storage. EMC-owned VMware and Pivotal played big roles at EMC World this year, trumpeting their roles in the software-defined everything world.

EMC CEO Joe Tucci spoke of a transition from a PC and server-centric data center to one of mobile devices, the cloud, big data and social networks. He said the new data center will require software-defined compute, networking and storage. "The new buzzwords are abstract, pool and automate," he said. "This is what the software-defined data center is about at the highest level."

Pivotal CEO Paul Maritz also used the Google storage model in his example of what the new data center should look like. "The consumer Internet giants have much larger information stores," he said. "When Google set out to index all the information on the Internet, it had to innovate and build a new architecture." And the key to that architecture -- the Google File System -- is a form of object storage, he said.

Juergen Urbanski, vice president of big data and cloud architectures for T-Systems, the $13 billion IT services division of Deutsche Telekom, is building that type of data center. Urbanski said his data center has more than 20 PB of data and "right now we see a very dramatic disruption happening in storage, primarily driven by Hadoop."

He said the difference between what he is doing and what the Internet giants do is that his workloads are far more diverse because his clients come from a wide range of industries.

Urbanski said his data center relies heavily on Hadoop because: "We don't like vendor lock-in and Hadoop is open source. That makes it able to run on industry-standard hardware and cheaper storage."

Urbanski said T-Systems is a large VMware customer, but doesn't use much EMC storage. He said the telecom has traditionally been a NetApp shop, although it has "storage from all kinds of vendors." However, he does use Isilon to run some of his customers' workloads because of its Hadoop capabilities.

"Our firm belief is that by 2015 … 80% of new data that comes in will land first in Hadoop," he said. "What we don't know is what the physical infrastructure underneath will be -- is it DAS, is it Isilon? We don't know. But at the data management layer and the distributed file system, it's Hadoop."

One thing he is trying to determine now is whether enterprise storage adds enough value to Hadoop clusters to make it worthwhile to run on more expensive storage rather than commodity white boxes.

For example, Isilon brings better data protection and performance while adding multi-tenancy that is missing with Hadoop on commodity storage.

"Multi-tenancy is important for us," he said. "Does Daimler want to have its data on the same cluster with BMW? Probably not. But they're both customers. So multi-tenancy is a core design point for us to offer. Today, Hadoop by itself does not support that. So what's the right tradeoff between open source -- commodity if you will -- and something that arguably has higher Capex and vendor lock-in, but brings other benefits?"

He said those tradeoffs will also be an issue with ViPR, which EMC claims will fully support commodity and third-party storage as well as its own arrays.

"With ViPR we can have one control plane and that helps us manage storage across pools," Urbanski said. "That part I buy. The $64,000 question is, 'How much of that pool will be white box stuff?' For EMC, that boils down to a fairly big business model issue. If you're a sales guy with a $20 million quota, how do you make up your quota based on what's probably a few hundred thousand dollar deals just for software? That's a big disruption in the industry, and that's the thing staring people in the face."

View the original article here

Read More