In a followup to my post Monday about VMware and VSS integration issues as pointed out by W. Curtis Preston, Symantec emailed over the following statement from Peter Elliman, senior manager, product marketing for the Information Management Group.
Yes, we agree, that VMware should update their [sic] VSS writer code and we believe this is the best place for this issue to be resolved. We’re probably not the only one who believes this which is why only two vendors have created a work around here. VMware tools is constantly updated [sic], so when updates are provided by VMware, it lowers administration effort which is not trivial when you have 100s of VMs. With third party VSS writer code in a VM you run the risk that an upgrade from VMware tools will cause a conflict with the VSS writer there and you have separate code that has to be updated. This is why we focus on integration with VMware, rather than work-around efforts. We believe that VMware will address this issue in the future. Finally, we recommend agents when protecting mission-critical applications within VMs because not only does it assure consistency and proper log management, it also offers many more recovery options tailored to that application, e.g. Oracle tablespace recovery, or SQL Server filegroup recovery.
So there you have it. Until VMware’s VSS integration changes, we will probably still see users deploying backup software agents on guest machines, as recommended by Symantec.
Among the updates Symantec announced today to its NetBackup 7.0 and Backup Exec 2010 backup applications are enhancements to the granular backup of applications running in virtual servers, in part through integration with VMware’s new vStorage APIs for data protection.
These APIs are among the more widely hailed updates in vSphere 4 for storage pros. They promise to eliminate the cumbersome VMware Consolidated Backup (VCB) from the infrastructure and allow existing enterprise backup software tools to make backups directly from virtual machines, the same way they’ve been doing for physical servers.
Backup Expert W. Curtis Preston has been among those claiming the vStorage APIs for data protection will be a boon for improving virtual server backups, but points out in light of Symantec’s announcement that its approach of integrating NetBackup 7 with VMware’s VSS implementation leaves something to be desired.
According to Preston’s research, VMware’s VSS support will perform consistent backups of a data volume, which he calls “table stakes” in the snapshot backup market. But Preston says the VMware VSS integration can’t perform application consistent backups with Windows 2008 hosts, and in the case of either Windows 2003 or Windows 2008 hosts, it won’t notify the application when a recent backup has been made or refresh where it starts tracking incrementally changed data. (Think of it as resetting the trip speedometer in a car after a trip is finished).
“What this means is that anyone wishing to get proper backups of applications in Windows must run an agent of some kind in their guests in order to make this happen,” Preston wrote in a Jan. 11 blog post. He goes on to warn, “This means that any backup tools that are using only VMware’s infrastructure are going to have the same limitations.”
Symantec declined comment on the limitations cited by Preston. VMware officials confirmed that for Windows 2008, vSphere supports backups at the operating system level (as opposed to the application or transactional level). They also confirmed that vSphere’s integration with VSS doesn’t make the application aware it’s backed up (see the trip speedometer analogy above), but said through a spokesperson, “Back in the old file-level days there used to be an archive bit that was changed, and hence the application was aware of the backup. [But] the question is, does it really matter for the image level backup?”
“This has nothing to do with the archive bit,” Preston responded in an interview with Storage Soup. “Applications need to know when to truncate their transaction logs.” If transaction logs aren’t truncated, in the case of a database application, “they’ll eventually fill up and crash the database.”
VMware plans to continue innovating around VSS backups, the spokesperson added. Preston’s blog post also mentions VMware is working on more granular VSS support.
In the meantime, I’m wondering if anyone out there reading this in blogland has personally encountered these problems, or better yet, any workarounds they would like to share.
Coraid, which has stayed mostly under the radar with its ATA over Ethernet (AoE) storage systems, is looking to make some noise this year.
Coraid today named a new CEO and said it closed a $10 million A funding round. An “A” round means it’s the vendor’s first VC funding, even though it has been around since 2000.
The CEO is Kevin Brown, who was CEO of desktop virtualization startup Kidaro until Microsoft acquired it last year. He also ran marketing for storage encryption appliance vendor Decru before NetApp bought it in 2005, and then spent time as VP of NetApp’s security business unit.
The new backers are Azure Capital Partners and Allegis Capital, and Coraid has added former Cisco honcho Charlie Giancarlo and Veritas founder Mark Leslie to its advisory board and they have also invested in the company.
That’s an impressive haul of money and talent for a company that has been so quiet that its new CEO says, “Despite having been in storage industry for some time, I hadn’t come across Coraid until recently.” Apparently, Brown doesn’t read SearchStorage enough or he would know at least this or this about his new employer.
What he does know about Coraid is its Etherdrive AoE systems don’t use the TCP/IP Ethernet protocol, which he claims gives it a performance advantage over iSCSI. He says Coraid has more than 1,100 customers and its systems cost about $500 per terabyte running with Gigabit Ethernet and $900 per terabyte with 10-Gigabit Ethernet.
Brown said when he did take a good look, “I saw Coraid from a disruptive perspective, and was impressed what they did with organic growth.”
Coraid lists NASA, Dunkin’ Donuts, the U.S. Navy, National Institutes of Health, and several large universities as customers. Not bad, for a vendor that has had one sales person until now. With the funding, Coraid is looking to build its sales, support, and development teams.
Brown positions Coraid as a low-cost alternative to iSCSI, which has made inroads as the low-cost alternative to Fibre Channel. He says Coraid is cheaper than iSCSI because AoE doesn’t require TCP/IP offload engine (TOE) cards. The problem with that claim, though, is that most iSCSI systems work with software initiators that also remove the need for TOE cards.
iSCSI systems also have more mature data protection and management software, while Coraid relies mainly on virtual servers and third-party applications for those features. Brown says higher-end Coraid units are coming, and points out Etherdrive already supports SAS and solid state drives (SSDs). While Coraid isn’t looking to taking over the enterprise, its new CEO says there are plenty of organizations seeking low-cost quality storage.
“We’re not going to attack every environment,” he said. “We’re not going to do a rip and replace on an investment banking company using SRDF. But there are a lot of new opportunities out there for us.”
IBM wants everybody to know it is still advancing its tape storage systems, even as newer technologies such as 6 Gbps SAS and solid state drives (SSD) are poised to make their mark on enterprise storage.
IBM today said its Zurich IBM Research lab in tandem with FujiFilm has demonstrated technology advances that increase the density of tape 44 times from what is possible in today’s LTO-4 cartridges.
IBM scientists say they have recorded areal data at a density of 29.5 billion bits per square inch on an advanced prototype tape. That density would enable cartridges to hold up to 35 TB of non-compressed data.
IBM Fellow Evangelos Eleftheriou says the demonstration involves new dual-coated particulate magnetic tape from FujiFilm that uses an ultra-fine, perpendicularly-oriented barium-ferrite magnetic medium. The new tape enables high-density data recording without using expensive metal sputtering or evaporation coating methods. It also involves advanced servo control technologies for more accurate head positioning, allowing a 25-fold increase in the number of data tracks that can fit onto a half-inch wide tape. Other advances include new signal processing algorithms for the data channel and new head technologies for low-friction giant magnetoresistive (GMR) read/write head assemblies.
Eleftheriou said his team surpassed its target of achieving 20 billion bits per square on tape by 2009 by hitting 29.5 billion. The previous record set in 2006 was 6.7 billion. The team’s next goal is 100 billion bits per square inch.
“We’re now in position to set the bar higher,” Eleftheriou said. “This demonstrates that tape has a lot of life left.”
So when will we see the results of this higher density in shipping products?
“I don’t know,” Eleftheriou said. “I’m a scientist, not a product guy.”
Disk drive maker Seagate Technologies last night reported its earnings for its fiscal second quarter of 2010 (ending on Jan. 10), and the results follow on IBM’s storage sales upswing with continued reports of an enterprise storage spending rebound.
“Over the course of calendar year 2009 the technology industry has improved faster than the broader economy and the storage sector has outperformed almost every other sector in technology,” CFO Stephen Luczo said on the company’s earnings call. “As a result, the demand for storage continued to accelerate throughout the calendar year.”
Still, Seagate was not planning for an economic rebound during the quarter and officials said it came as something of a pleasant surprise to see stronger than forecasted demand. The company’s top line revenue for the quarter was $3 billion, a 33% increase year-over-year and a 14% increase sequentially. Net income was $533 million, and the company was able to generate $753 billion in operating cash flow and repaying $246 million in debt, which Stifel Nicolaus Equity Research analyst Aaron Rakers wrote in a note to clients this morning was the biggest quarterly cash flow generation the company has achieved since his firm began tracking it in 2001.
Wall Street analysts, many of which were also pleasantly surprised by earnings from Seagate above their estimates, are now looking to see this boon for Seagate ripple out to its enterprise storage OEMs, especially EMC, which reports calendar first-quarter earnings Tuesday.
During its conference call to report fourth-quarter earnings last night, IBM reported its storage sales were up 1% year-over-year, in line with the cautious optimism about storage sales that has been growing over the last few months.
Specific product highlights in storage included Tivoli software and the XIV disk array. “Tivoli storage continued its robust growth as customers manage their rapidly growing storage data,” said CFO Mark Loughridge in prepared remarks on the earnings call. “Data Protection as well as Storage Management grew double-digits with broad- based geography and sector growth.”
Loughridge added later, “we added more than 130 new customers to our XIV platform in the fourth quarter and 400 since the acquisition.” IBM also claimed to have taken share in both the disk and tape markets, even though tape revenue declined 10% for the quarter.
Stifel Nicolaus Equity Research Aaron Rakers pointed out that the 1% year-on-year gain doesn’t tell the whole story. “This implies a sequential increase of [about] 55%, which compares to an average sequential increase of [about] 36% in [the fourth calendar quarter] over the prior 7 years,” Rakers wrote. “IBM estimates that it had gained [one percentage point] of share in the storage market during [the fourth calendar quarter of 2009].” Rakers compared this with his estimate that competitor EMC Corp.’s Information Infrastructure business grew 19% sequentially last quarter. EMC reports earnings results next Tuesday.
EMC launched a news midrange Clariion model today, the High Density Clariion CX-4.
The new storage frame can hold up to 390 disks in three rack units and supports 2 TB 7200 RPM and 5400 RPM SATA drives, or a combination of SATA drives and up to 60 solid-state drives (SSD).
The new Clariions are 2U higher and five inches deeper than the standard Clariions, and let customers build larger capacity configurations in a smaller footprint. For instance, a standard Clariion CX 4-960 with support for 1 TB SATA drives will take up six racks to store 945 usable TBs. The High Density units can store the same amount in three racks.
The Clariion announcement itself wasn’t anything all that earth-shattering, but it led to an interesting discussion with Enterprise Strategy Group analyst Mark Peters that I think bears summarizing here.
Thoughts on “high end” vs. “midrange”
I spoke Monday with a Pillar Data Systems reseller about Pillar’s new Axiom 600 controller. He said Pillar had recently come up against Symmetrix V-Max in a deal. Pillar lost, but the reseller claimed that was due to EMC’s brand recognition rather than a technical disadvantage for Pillar in this environment.
At first, the comparison between EMC’s high-end disk array and Pillar’s Axiom, which has lived squarely in the midrange, might seem a jarring one. While a fully configured Axiom 600 can scale up to over a petabyte and a half, a single V-Max engine pair can address that much capacity, and scale up from there. V-Max also offers more bandwidth, at 8 Gbps FC while Axiom uses 4 Gbps FC, and mainframe connections through FICON, which most midrange arrays, including Pillar, don’t offer.
But this kind of comparison isn’t nearly as much of a stretch as it might’ve been even four or five years ago, when “midrange” or “low-end” meant “stripped of certain features and functionality.” “It used to be when you wanted a feature that was really complex or clever, you would get a high-end system,” said ESG’s Peters.
These days, the features advertised for Axiom and V-Max have quite a bit of overlap, from thin provisioning to quality of service to tiered storage data migration and solid-state drive support. Since EMC came out with the V-Max Symmetrix model, a successor to the monolithic DMX series, the two have also, had a broadly similar architecture, a matrix made up of separately scalable performance (“engines” for EMC, “slammers” for Pillar) and capacity nodes (“capacity” for EMC, “bricks” for Pillar).
It also used to be that a high end disk array could be identified by the amount of cache available to boost performance, as well as intelligent algorithms to manage placement of data on that disk. With the advent of Flash for enterprise consumption, vendors like Pillar are able to offer the kind of cache capacities that used to be available only in the highest-end arrays — 192 GB with the Series 2 Axiom 600 controller.
The overlap between midrange and high end is also increasing within EMC’s own product line — the Clariion, which it calls a midrange array, can now achieve capacities well into the Symmetrix range, especially with the high-density model. Both products also offer Flash support, QoS, thin provisioning, drive spin down, etc.
Of course this doesn’t mean that people will begin speaking of the two categories interchangeably, at least not tomorrow or in the near future. But Peters noted that users may emphasize new types of purchasing criteria now that the old lines of sheer capacity and horsepower are beginning to blur. This is where, as the Pillar reseller mentioned, brand recognition and vendor cachet will become more important than ever. “People are buying more than a product,” Peters said. “They’re buying interoperability, a sheer number of service engineers, money going into a research lab, and global support.”
The size of the environment will also continue to play a role, but this will find more of an emphasis on risk management depending on the size and profile level of the business involved. “Different size companies still have different bases on which they make purchasing decisions, and one of those is risk.”
It’s also important to recognize that different companies will use the same technical term to describe features that might still be different under the covers. “All the vendors will now say they have thin provisioning and remote replication,” Peters said. “But you always have to be careful in this game to look out for semantic similarities that may not mean the features are exactly the same.”
So the story as told by the Pillar reseller will probably continue to be retold in various forms.
As an example, Peters turned to a car analogy. “Hyundai and Mercedes might both say they have four-wheel drive,” he said. “And it may be that Hyundai’s really is just as good as the Mercedes, but I don’t necessarily trust them. That’s similar to a storage buyer [today].”
HBA vendors QLogic and Emulex say sales last quarter exceeded their expectations, and Wall Street analysts are predicting good numbers for storage systems vendors. But not every storage vendor finished 2009 strong.
FalconStor Software said it finished 2009 below its forecast due to a poor fourth quarter. FalconStor’s revised guidance is for annual revenue of $88.5 to $89 million compared to previous guidance of $96 million.
FalconStor CEO ReiJane Huai blamed the shortage on problems with his company’s OEM business. FalconStor says OEM partner H3C generated more than $2 million less in licensing revenue than expected following Hewlett-Packard’s acquisition of H3C parent 3Com. Oracle’s delayed closing on the acquistion of FalconStor OEM partner Sun Microsystems also resulted in less revenue than expected.
Perhaps the biggest problem for FalconStor was revenue from its largest OEM partner, EMC, also came in below expectations. That’s despite analyst predictions that EMC beat its overall revenue forecast for the quarter with strong sales of its Symmetrix V-Max enterprise systems and Data Domain deduplication appliances.
Along with EMC, Wall Street analysts expect good quarters from other storage system vendors large and small. RBC Capital Market analyst Amit Daryanani predicts NAS vendor Isilon not only beat expectations last quarter but broke even for the first time. 3PAR also exceeded expectations, according to Wedbush Securities analyst Kaushik Roy.
“Industry contacts indicate that all storage systems vendors appear to have met or beat internal expectations,” Roy wrote in a note to clients issued today.
After enjoying the last couple of hours of 2009 with my family, I thought how fitting it would be to end the year with a post!
I’ve been incredibly busy this year and my lack of posts really shows it, one would think I forgot my login or something. In that time, however, there has been no lack of great topics to talk about, and here are a couple that lit my candle in 2009:
Consumer computing is fast approaching levels of enterprise computing, making corporate citizens more computer savvy, and making IT management work harder to keep things humming along. Mark my words, you are going to see quite a bit of work-slash-home networking products come to the market in 2010, specifically around data protection and storage that are going to tout “office integration” or “workplace integration”.
The mobile computing and storage space and the rate at which consumer mobile devices are making inroads into the datacenter is something that I’m paying close attention to. Specifically, the Android OS and the Nexxus One and Droid hardware–these devices are significant to enterprise computing because they take the whole idea of a netbook to another level!!
If you remember the Toshiba Libretto, these new devices are what the Libretto could have been. The mobile phones are both fast and offer the ability to the savvy user to essentially replace their office with a hand-held device. And for those with super security conscious IT departments, there are companies like Good Technologies “Good for Enterprise” that allows an administrator to remotely wipe Exchange data from a Droid in a fully encrypted container so “security” can’t be used as a reason not to support the platform.
Take this a step further, I’m sure you’ve been asked at least once already to store backups of a user’s phone to tape, or better still seen a backup of a user’s phone on their shared drive. If you haven’t yet, you’d better get ready for it!
Virtualization has been rampant, and I predict it will be in my toaster within the year, allowing me to virtually toast multiple slices of bread simultaneously and store the trend info on how many times I’ve burned my Eggo’s on SSD. While I’m being flippant, we may actually see a hypervisor capable toaster or fridge or washer, and apparently I’m not the only one that thinks so–in an article on a New York times blog Sehat Sutardja has been quoted as saying: “[Virtualization] will become pervasive…It will be used in everything from TVs to IP phones to digital picture frames to washing machines.”
If Android is in a washing machine, then I have Linux and everything that is available to Linux in that washing machine … just think of the Folding at Home scores you can rack up if we linked the neighborhood washing machines up!! Think about all the data that will need to be stored when they start tracking wash cycles of a particular garment via RFID!!!
On a more serious note, the age of operating systems for small to midsized branch office network attached storage devices, as well as smarter switches and other infrastructure devices, is upon us. Microsoft is not standing still — Windows 7 is small and much faster than its predecessors (why do I feel like that is a paraphrase of the architect from the Matrix?) and is definitely a viable OS for these devices, so now we have raw Linux; Moblin is making its way onto the stage with Android, among others. And remember, all these things have one thing in common: they need somewhere to store the data they produce.
Speaking of virtualization, the march of development in the virtualization management software space is going to pick up steam in 2010, and there are going to be some casualties. The winner will be the one that allows truly heterogeneous management of my virtual data center from storage up, and after taking a look at Cisco’s offerings I’m going to be paying very close attention to what they do. I’ve been digging really deep into vSphere, and it’s jam packed with goodies. Orchestrator is a little gem — properly executed, it can add a good bit of speed and agility to any rapid provisioning initiative you may have, BUT be careful, with a poorly orchestrated (you knew that pun was coming didn’t you?) workflow that shiny new NAS with 400TB of storage will be gone in a day.
Enterprise Storage has continued to move forward at a blistering pace, with drives breaking the 2 TB mark, and some serious performance increases in the form of SSDs, Sata III and Fusion IO putting Flash directly on the bus. I look at price in this space. The price of SSDs will get lower and lower and the performance will continue to go up. We will see the proliferation of end-to-end solutions mixing the two, a la Exadata and the Sun 7000 line. Take a look at what Fusion IO is doing in the high end gaming market! It’s funny but the consumer machines of today are looking more and more like the specialized workstations and servers of yesterday.
I see some things that we really missed the ball on last year, too. Convergence really isn’t here yet. The drive to make a device the “media hub” and then backing all that stuff up is getting there, but hasn’t quite caught on yet. I think once it gets closer it could drive an entire wave of datacenter build outs to handle it. I can also see telcos getting into the act a little more aggressively, offering storage services at their major POPs to enable some of the consumer products to work properly. This has some unintended but positive side effects for the small to medium business because they will have ready access to fast, reliable online storage. Well, at least in theory. I’m still waiting for it to happen!
Cloud storage also hasn’t really shaped up to be the game changer I thought it was going to be. I like the idea of not owning infrastructure and I’m a really big fan of the rapid provisioning/de-provisioning model, but I just don’t see the bandwidth needed for that to work here in the US the way it really should. In Korea and various places who’ve deployed infrastructure recently I see cloud as a viable model, but not here.
With that, folks, I’m back and rarin’ to post!!!