HP snorage must wake up before biz bosses kick it out of bed At the HP analyst day this week, enterprise group executive vice-president Dave Donatelli laid it on the line for subsidiary 3PAR and the StoreOnce team: you gotta help get us out of this place. This place being the rut HP's growth figures are stuck in, and CEO Meg Whitman has bet on 3PAR and StoreOnce in her turnaround plan …
If performance takes a nose dive at 70%+, then why are all of 3PAR's SPC-1 results way above 70% utilization.
It doesn't scale well ? it scales from two controllers and 32 drives to 8 controllers and 1920 drives in a fully active / active cache coherent system.....and that is not scaling well?
It sounds like you're
A. a competing vendor with some FUD to sell
B. you've drunk some other vendors koolaide
or
C. you haven't go a scooby doo (clue)
No, the previous dude is right. Yes, it can scale, but you need to buy a whole additional frame to add controllers, not a modular design where you add a disk node and you get more proc, cache, etc. Therefore, when you get up to higher levels of utilization, it chokes the controller pair (like any dual controller array). Yes, you can expand it and add more perf... for several hundred thousand dollars/pounds.
Like VMAX, if you have tons of cash, it scales well, but not as cost effectively as alternative solutions. Difference between paper scaling and real world financial consideration scaling... adding new frames is costly.
"No, the previous dude is right...." No, he's just not.
"....Yes, it can scale, but you need to buy a whole additional frame...." Please supply details of the 1900+ disk solution you have that resides in a single frame yet provides the same levels of performance, reliability, redundancy and featureset. ALL top-end arrays scale out into additional frames. From a resilience viewpoint, you actually want to spread your controllers over the different cabinets. Otherwise, if you have all the controler nodes in one cab and that cab has power probelms, how many nodes do you have left running? Zero. TBH, anyone knocking the 3PAR for spreading nodes across cabs is firmly in the "doesn't have a scooby-doo" section of the audience.
I don't think we disagree. If you need 1900+ disk, 3PAR scales well. Most people don't need 1900+ disk or anything approaching that range. If you are doing a 80-100 TB, 3PAR doesn't scale as well as other systems because it is not modular and assumes you are going to be working across multiple frames. It is built to compete against VMAX and the like at the high end, but HP is trying to push it into their mid-market account segments where it doesn't make sense... because EVA is dead and they don't have another mid-market offering. HP has one hammer, one solid storage platform, so everything they see is a nail for that hammer.
"....Most people don't need 1900+ disk or anything approaching that range....." Yeah, which is why hp have more than just one 3PAR offering, plus EVA, MSA, VSA/P4x00 and the D-series JBODs. And the D2D products, and tape and VLS. And even XP for mainframe holdouts. They seem to have it pretty much covered. I suggest you improve your knowledge by reading here for a start: http://h71028.www7.hp.com/enterprise/cache/418227-0-0-224-121.html
".....HP has one hammer...." Apart from the other products listed above, the 3PAR range is more than one hammer, being the lower-mid range F200, through the F400 and T-series to the top-of-the-line V-series all modular and scaleable. TBH, to call that "one hammer" shows someone has just been reading too much of another vendor's FUD.
"One hammer, what you mean like Netapp ? Doesn't seem to have done them much harm."
They stuck to the niche, mid range NAS, pretty well and it went pretty well. Now that they are trying to push their boxes outside of their niche, into unified storage, it isn't going as well. NetApp is a one trick pony. Fortunately for NetApp, that one trick was what everyone needed... unstructured file data management. 3PAR may do well in their higher end but not highest end niche, but it isn't going to work out if they try to make 3PAR a 50-100 TB array for a $500 million grocer and the like.
"3PAR may do well in their higher end but not highest end niche"
I'm not sure they want that niche, it's mainframe dominated and declining
"but it isn't going to work out if they try to make 3PAR a 50-100 TB array for a $500 million grocer and the like."
Why's that then ?
"I'm not sure they want that niche, it's mainframe dominated and declining"
Mainframe would be highest end. DS8 and VMAX dominated, with some HDS here and there. 3PAR doesn't support a mainframe FICON connection.
"Why's that then ?"
Too many $$$. 50-100 TBs with decent performance. V7000, VNX and the like will make come in at half the cost.
Didn't I just say they wouldn't want that niche ? why break the simplicity of their solution by adding all the legacy mainframe stuff to address a market that takes years to gain credibility in and is declining.
So we've established 3PAR is fast, scalable, very efficient with all that thin stuff, it's also simple to use with all the inbuilt virtualization, it 's feature rich and provides a common operating environment from entry level to high end, however it carries a premium vs it's competition in the mid range.
The assumption is that HP are not aware of this....................
"For the few mainframe cases hp will just sell an XP array."
No one is buying mainframe storage from HP, XP or otherwise, but 3PAR doesn't qualify for the highest end with VMAX and DS8 and VSP because they don't support FICON. You can't be highest end without mainframe. Mainframe is the mission critical platform for many large shops.
"Mainframe is the mission critical platform for many large shops"
But fewer and fewer large shops every year, why invest development $$$'s in a a declining market. ? This seems to be used as stick to beat 3PAR with and provide a convenient excuse to deny its tier one credentials. When in fact 3PAR never wanted to address that market, so it's a conscious decision not a feature gap.
"But fewer and fewer large shops every year, why invest development $$$'s in a a declining market."
Mainframes are going to be critically important to large shops for a very long time. If you want to play in the mission critical environments at any large bank/insurance company, nearly any large government, most of the major industrials, you need mainframe support. It doesn't mean that you can't make good storage that doesn't have mainframe support, but it isn't a toothpaste to tartar sauce, tier one array. With a tier one array, you don't even have to ask if it has x feature, it has it. EMC VMAX, IBM DS8, and HDS high-end are the tier ones.
Yes EMC VMAX, IBM DS8 and HDS VSP all have mainframe support, but actually that's just a function of their origin and age, it's no longer a growth area for any of them, if they could easily dump much of the mainframe baggage they have to carry, they would.
If you take a tier 1 array and try to use it for Opensytems and Mainframe you end up partitioning it anyway. Drives have to be formatted specifically for mainframe access, licensing is different and all of the intervening infrastructure also needs to be kept separate. You haven't simplified management or really saved anything commercially, you're now paying over the odds for all storage to be mission critical, just in order to support your legacy mainframe environment.which very likely is not a high growth area anyway.
Better to provide a small amount of legacy tier 1 to your mainframe and take you pick of the vendors for your growing open systems environment..
The toothpaste to tartar sauce analogy is just a function of the platforms underlying age and time in the market,.Of course they've collected every feature imaginable over the last 20 + years, but it's all bolted onto a legacy software architecture originally designed to serve mainframe only environments. Open system support was an afterthought when they realized mainframe wouldn't be able to sustain the market going forward.
They're stuck with lots of complex and arcane underlying structures that are really only applicable to mainframe. Whereas the newer platforms have the advantage of not needing to support such legacy constructs, allowing them to simplify integration and provide a faster time to market for new features. It's not going to be long before the new platforms are pretty much on feature parity for open systems if they aren't there already.
"......EMC VMAX, IBM DS8, and HDS high-end are the tier ones." The XP is the badged Hitachi array. By still offering XP, hp can still meet the requirements for those few mainframe holdouts at minimal cost as Hitachi are paying for the development. Meanwhile, 3PAR lets them mop up the more lucrative areas of the market.
Um, NetApp has been doing unified storage NFS, CIFs, iSCSI and FC for TEN years, the have been operating outside of their niche and taking share for some time now. If their day in the sun is over then why are they growing at 20% per year, are the #1 storage vendor (by a longshot) to the US Federal government, are the #2 storage vendor by revenue and the #1 storage vendor by usable capacity shipped?
What does the US federal government have to do with anything ? all these public sector contracts are based on frameworks, not the best or most appropriate technology.
Netapp do a decent NAS box but a poor SAN, if you do go the unified approach all you end up with is NAS on one head and SAN on the other and a bunch of mutually exclusive features you were initially led to believe would work equally well in either environment but don't.
Netapp's success has been playing to both audiences and anyone who sits in the middle Want NAS ? here you go. Want SAN ? here you go, not sure what you want ? have some of this. Oh the Netapp SAN isn't working out for you (who'd a thunk) ? why not put all your eggs in out NAS basket, it's jolly nice but you'll never get off it .
Their technology is looking dated and isn't really cutting it outside the low end of midrange, as for the scale out clustering it looks overly complex, like something cobbled together in a desperate first effort to enter a new market.
According to this article Netapp shares also fell 40% over the same period.
http://www.thestreet.com/story/11730938/1/netapp-needs-to-show-some-enterprise.html
The article infers they'll have trouble growing given the state of the US Government storage spending. Basically they've remained a "one trick pony" and the state of the economy has exposed that fact.
Until they bring the 3Par downmarket and refresh the F class HP have no answer to Storwize v7000, EMC VNX or Dell EQL.
EVA is dead, the P2000 is over priced and under featured and LeftHand is a joke at the moment with no sub-LUN tiering making the P4900 largely useless because peer motion is just a joke.
P9500 is on the books for the SMALL number of mainframe shops that run HP and I would think most of them will move to VMAX, HDS or DS8870.
So their only really competitive array is the 3Par P10000 which is a monster entry point.
If cost is your main driver or you absolutlely must have the very latest hardware iteration. But if you want performance, availability and features then the current F400 beats v7000, VNX and EQL including Compellent hands down. Don't believe me ? take a look at SPC-1 results, 3PAR persistent cache, non duplicative snapshots, secure and resource based multi-tenancy etc etc etc, Same software and functionality from entry through midrange to enterprise unlike all of the above you mentioned.
P2000 "over priced", where do you buy your storage PC-World ?,
"if you want performance, availability and features then the current F400 beats v7000, VNX and EQL including Compellent hands down."
Exactly, it is mid-range, we are talking about cost. If we were talking about "performance, availability and features", then we would be talking about DS8/XIV and VMAX vs. 3PAR... not V7000 or VNX. EQL is low end iSCSI junk, not comparable with any of the others mentioned.
So these features are not important to mid range customers, only the very high end ? 3PAR offers these in mid range or high end, but unfortunately they're not for free. I do think they have some pricing challenges, but the biggest one is customers and other vendors trying to compare apples to oranges.Being able to scale beyond two controllers in the mid range is a huge benefit, so you have to look at the bigger picture, which I appreciate is often difficult in mid range land.
"So these features are not important to mid range customers, only the very high end ?"
Sure, people will take them at a mid-range price point, but HP needs to offer them at a mid-range price point. IBM did this well by taking their high end, high cost SVC, Easy Tier, etc and bundling it with V7000 at a price people can afford. HP is inviting the comparison of apples and oranges. They see a mid-market customer with maybe 80 TBs and modest functional requirements. IBM brings in V7000, EMC bring in VNX, HDS brings in HUS... HP brings in a stripped down 3PAR at a higher cost than any of the previous three... and then says that people are comparing apples and oranges. Great, bring in your HP apple so we can compare it with the other apples. IBM, EMC, and HDS didn't bring in DS8, VMAX, or VSP, why are we talking about 3PAR? Traditionally they would have brought in EVA, but no one is buying that these days.
"....HP brings in a stripped down 3PAR at a higher cost than any of the previous three... and then says that people are comparing apples and oranges....." It's called a feature sell. In the old days it was a staple tactic of Sun salesgrunts - get into the conversation the idea that feature X was to-die-for, wait for vendor Y to thrash about trying to match feature X, often having to throw away their first choice product and go up-range to their next level and more expensive offering to match the featureset, then put in a price below that more expensive product from vendor Y. The vendor pushing the feature sell gets to control the conversation, making them more likely to get the win even if their product was actually the WORST choice originally. I have seen CIOs be brainwashed by such tactics into buying servers that really were dogs compared to IBM or hp equivalents of the day.
My guess is hp are selling 3PARs into lower-end deals using stories around thin provisioning, "autonomic" operation, Remote Copy, etc, etc. Having been a customer on the receiving end of such from Sun my advice is be sure of what you want, pin down your list of requirements, and then don't let any vendor push you into their choice of talking points. In such cases, if 3PAR really is not the right choice then you will realise that.
Feature sell - sounds just like what Netapp have been doing the last 5 years, although I do think they're now beginning to struggle and they've never really had any high end credibility. EMC have added loads of bolt on features specifically to combat the Netapp feature and unification story. The problem for both is that few of their features are tightly integrated, in many cases you have to sacrifice one feature for another. Just goes to show you can refresh with all the latest and greatest hardware available, but if your underlying software architecture is stuck back in 90's then in reality you're pretty much standing still. I think this is why the newer architectures from HP 3PAR, IBM v7000 & Dell Compellent will start to make a much bigger impact, they aren't going to have complete parity day one due to their historically smaller market, but t's only a matter of time.
This post has been deleted by its author
IMO, NetApp's day in the sun is over. They are really expensive for a file server and they don't have the functional or performance chops to play at the higher-end. IBM, EMC, HDS, etc have finally figured out that if they just staple an NFS head on to their mid-range gear, it is difficult for NetApp to win. NetApp basically had a free reign for a number of years in the NAS space.
Um, NetApp has been doing unified storage NFS, CIFs, iSCSI and FC for TEN years, it's been a long time since they only did NAS. If their day in the sun is over then why are they growing at 20% per year, are the #1 storage vendor (by a longshot) to the US Federal government, are the #2 storage vendor by revenue and the #1 storage vendor by usable capacity shipped?
EMC also has persistent cache these days :)
SPC-1 means crap TBH, if you want real world configurations that customers actually use (and I am building these across EMC, HP, IBM and HDS all day) in the mid-market space then you would know that v7000 and VNX currently have HP beaten up and down the block all day long. The only way HP can be price competitive is to pitch EVA or discount the ass out of 3Par.
To match a v7000 config with a 2 node F400 you will be paying nearly double for a box that is limited to 4Gbps FC and 1Gbps iSCSI (Yes I know it has a lot of ports compared to the v7000).
As for P2000 being expensive, it is when compared to IBM's DS3524 and EMC's VNXe 3150 which are it's natural competition.
No EMC have a me too feature on VNX/CX/Clariion that keeps writeback cache enabled when a controller fails. That is not persistent cache since you are sacrificing data integrity to maintain performance in the event of failure. EMC had to do this due to high profile service provider outages. 3PAR provides this functionality with four or more controllers only, so that there is always a mirror copy of the cache contents. That way data integrity and performance can be maintained safely unlike EMC's half arsed implementation.
It took us 3 months to get a single 2TB smartcell for our customer, who urgently needed it, just because of all the HP business processes and moronic middle-managers getting in the way of our request.
And thats just the latest in a long line of pointless delays getting additional HP storage.
Sort that shit out Meg, and maybe you'll be getting somewhere.
" just because of all the HP business processes and moronic middle-managers getting in the way of our request.
And thats just the latest in a long line of pointless delays getting additional HP storage.
Sort that shit out Meg, and maybe you'll be getting somewhere."
THIS THIS THIS THIS THIS!!!
As a previous 3PAR customer sucked into the sinking void of micro-mismanagement and ever expanding microcells of non-interacting business units who can't find each other in a well lighted gymnasium, I would throw out my 3PAR F400 if it was not such a perfect storage for my SMB. And I get the pleasure of paying for all this parasitic overhead.
Meg, when refresh time comes, the pure pain of the HP heirocracy and the waste of my time will weigh heavily in the equation. I would much rather spend my time addressing technical hurdles than mind numbing HP organizational issues.
"why are all of 3PAR's SPC-1 results way above 70% utilization?"
Download and read HP's SPC-1 disclosures sometime. The most enlightening bit is that all volumes are defined as being limited to a node, bypassing the cluster interconnect and effectively running the nodes as individual systems. Load generators drive the node they are attached to, which in turn uses its direct-connected disks. No one uses a 3PAR system like that in the real world.
What's next, filling the trunk of Meg's armoured Range Rover with SSD and doing SPC-1 runs on it?
SPC-1 is a benchmark, as such it's a speed test open to abuse by any Vendor. The key is that every vendor on there is attempting to show their product in it's best possible light, they aren't spending lots of $$$ to provide a poor result. Totally agree the configurations are unrealistic for the majority of the market place, but knowing that all the vendors are squeezing as much out of the benchmark as possible provides a good method for comparison.
The good thing about SPC is there are a few other metrics recorded as well as the IOps, such as cost per IOp, meaning anyone putting together a crazy configuration will be spotted almost immediately because the $ per IOp will be very poor. They also record the capacity utilization, which means you can easily spot the Vendors who short stroke their systems to get the most performance and finally if you look at the full disclosure, you get to see how complex the setup was from a command line perspective.
If you only look at the I/Ops number on SPC-1 then it helps with the benchmarks are meaningless argument that EMC are fond of trotting out. Looking only at the IOp number means you're playing into their hands and only fooling yourself, SPC provides much more relevant information under the covers. If you look you;ll see all of the 3PAR systems were fast, had relatively good $ per IOp figures, non were short stroked all were fully configured meaning they perform best when fully loaded.and all were simple to configure.
Take a look at many of the other configurations on the SPC-1 website and you'll be hard pressed to find systems that meet all of the above.
1. It's not sh*t. It's merely not the floor wax and dessert topping that HP portrays it as.
1.5. There are really good ideas in the product; it's sad how the implementation lags and is managed by HP.
2. Customer decisions are a product of multiple factors, product superiority often being secondary.
It is gaining market share because its market share was very small when HP bought it, so naturally it is going to be growing by double digits when the entire HP channel is going into EVA and XP accounts and trying to flip them to 3PAR. It is like Oracle talking about how Exa is growing by 200%. When you start with $50 million in rev, it isn't difficult to do.
No your're just repeating the same old FUD put about by HP's competition.....Gaining market share is different to double digit product growth, if 3PAR were just replacing existing EVA and XP customers and flipping them to 3PAR as you say, then HP's high end market share would be flat or more likely declining. Instead it's growing which doesn't really tally with your assessment now does it.
Anyone else noticed how HP's channel is constantly being poached for anyone with 3PAR skills, not that the competitions worried though :-)
This post has been deleted by its author
This post has been deleted by its author
"if 3PAR were just replacing existing EVA and XP customers and flipping them to 3PAR as you say, then HP's high end market share would be flat or more likely declining. Instead it's growing which doesn't really tally with your assessment now does it."
HP's overall storage market share is not growing. HP storage was down 6% last quarter in revenues. EMC is gaining storage market share, unfortunately, everyone else is just holding steady at best.
I do have to question why every poster is "Anonymous Coward" vendors??
As a customer and a vendor there is good reason why we went with VNX heck even the largest super computer in the world runs on VNX's.
Did we look at HP yea we used to buy and ship to customers allot of HP kit now its nothing why well HP the problem is there has not been substantial improvement in technology for some time. Followed by the fact every time we price in HP vs other vendors in all things HP they come out 3 times more expensive and take allot longer to get anything done.
VNX is a good example of why elegance is immaterial these days. It's a total hodgepodge, but it runs on state of the art hardware and delivers the promised results. HP could learn something from that, but no, it's old hardware for 3PAR and a scale-out NAS platform suitable only for frightening small children with.
VNX is a mess from an engineering perspective, well past it's replacement date, but EMC market the hell out of it and they keep refreshing the hardware, which in turn creates the illusion VNX is a relatively new platform. I'm not sure that it delivers on promised results vs other solutions out there, but there's a certain comfort in buying from the biggest vendor on the block.
"As a customer and a vendor there is good reason why we went with VNX heck even the largest super computer in the world runs on VNX's."
According to EMC.... The world largest supercomputer is IBM's Sequoia at LHL Labs. A ton of IBM Power nodes running Linux with Lustre as a back end loader... I believe the back end is NetApp, but not certain.
http://www.techradar.com/us/news/computing/inside-ibms-sequoia-16-thousand-trillion-calculations-per-second-1099285
Um the reason "Anonymous Coward Vendors" post here is because much of the information here posted by misinformed individuals is absolutely incorrect like your statement "even the most powerful supercomputer in the world is connected to VNX" Absolutely not true, the most powerful supercomputer in the world is Sequoia at Lawrance Livermore National labs and is backed by,.... wait for it,.....NetApp.
You never bought 3PAR because it was a barn burner, you bought 3PAR because it was THE innovator in things like wide striping, thin provisioning, and tons of other technologies that made it the industry innovator. That magic quadrant before HP bought 3PAR would have had HP closer to Oracle. It fit in a fantastic niche, better than average midrange with high end features.
Suggesting it's amazing because of it's SPC-1 scores is frankly terrible. While they didn't shortstroke nearly the amount Oracle or NetApp does (15 percent versus the others 50 percent or more), the read response time numbers as the load ramps up are terrible, the write's suffering less only because the entire box is configured as RAID-1.
But the fact is, once upon a time if you wanted to know what features the other vendors where putting into their arrays, you looked at 3PAR, but you don't anymore. And while everyone likes to point at HP and say, LOOK, they're screwing something else up, remember that they haven't innovated at all, which means the issues with moving forward predate HP.
I think 3PAR still are innovating, but the other vendors have cottoned on, 3PAR announces a feature such as multi-tenancy, wide striping, federation, cache persistence,etc and all the other vendors suddenly also have that feature. It's not that they can do the same as 3PAR, but it gives them some credibility when trying to fend off 3PAR from their Customer base.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
