HP may replace 3PAR hardware El Reg is hearing that HP will replace its flagship 3PAR T- and F-Class storage arrays with new, more powerful hardware next quarter. HP 3PAR T-Class array HP 3PAR T-Class array The current 3PAR arrays use X86 processors plus 3PAR's third-generation ASIC hardware to speed up array operations and increase the array's …
3par customers have been aware of these upgrades for about a year now(assuming they have talked with their reps in the past year), while they haven't announced anything in any formal manor, they are pretty open about it if you talk to them directly (well maybe not their marketing folks) It's never been a secret, what has been kept secret is beyond the basic design what new abilities the system will deliver (I don't know myself). One would assume the "storage federation" stuff that you reported on last year I believe, but I don't know.
They must be working on something big though they haven't had any major software releases in a year and a half, and of course no new hardware platforms in more than two years now. But still they are king of the hill when it comes to spinning rust performance within a single system on SPC-1, which is really impressive.
Well, I was considering 3par for a project but looks like this is one of those times where the vendor makes significant changes to their systems and it is prudent to wait for a while before the new range comes out and gets stabilized.
I would rather wait for a while rather than be the fool who buys about-to-be-obsolete technology in a clearance sale ! Hang on till the end of this year. Not worth it right now.
Yeah but my HP rep never told me that a refresh was due. I accidentally stumbled upon this article and found out that many existing customers already know about an impending refresh. Whereas my HP rep was still pushing the about-to-be-obsolete 3par to me. I consider that an ethical issue . Also, who knows the new refresh may have 8Gbps FC or 10GbE iscsi support. Hence I think its prudent to wait and watch till the last quarter of the year for the new generation.
Did you ask? Did he deny it? Chances are he will know for definite no more than the rumours and speculation floating around on this site. If you ask for a session under NDA, I am sure you will get it then you can stop speculating about what is or isn't included, when it will arrive and whether it is worth waiting for.
The development of ASICs is very expensive and time consuming, 3PAR needs approx. six years to develop a new one. The current "third generation" ASIC in the T- and F-Class is PCI-X 133/100MHz based and can not support higher bandwidth protocols like 8/16 Gbps FC or 10Gbps iSCSI and FCoE. The new fourth? or fifth? 3PAR ASIC generation ( they count failed ASIC designs - nobody has seen the "second generation" ASIC, the old S-, SX- and E-Class models are equipped with the "first generation" ASIC ) should cover these protocols with a PCI Express 1.0 design, but we have now PCI Express 3.0 . Old and slow hardware is the price for ASIC implementation. A much better approach is the pure use of Intel ( or AMD ) newest commodity server chips, a "brute force" method which exploits the newest hardware performance gains. This can compensate the ASIC advantages easily. A good example is the Intel "Jasper Forest" Xeon processor which is used in the IBM Storwize V7000 controllers. This processor has some storage array - related functions like PCI Express Non-Transparent Bridging NTB for interprocessor communication, RAID 5/6 parity generation and Asynchronous RAM Refresh ADR with an external power source. Therefore the windy ASIC road is death-ended when Intel extends this functionality. Hitachi has accepted this situation meanwhile and is switching to commodity designs, see the excellent article " Hitachi goes XYZ over next-gen storage" from the same author.
Firstly - the 3PAR array is still the fastest array on the market, with an industry-leading design and feature set. So I would hardly say that 'old and slow hardware' is an issue. Yes the arrays don't support 8Gbps FC but you can have plenty enough 4Gbps FC ports to more than compensate. The host connectivity of the 3PAR arrays is better than any typical comparable solution.
Secondly, there are huge benefits in building functions into an ASIC rather than layering them in software on top of each other on a commodity chip.
3PAR is the only array that can thin provision data on the way into the array rather than once it has landed on the disk.
3PAR is a genuine active-active cluster (up to 8 way) not active / passive or some sort of software cluster.
3PAR is built upon thin provisioning, not thin provisioning added as a software layer on top of all the other layers and therefore it doesn't come with the variety of considerations and complexities that competitive solutions come with.
Reliability and predictability - ASICS behave in a predictable fashion and very rarely fail.
Finally, Performance - the ASIC is doing exactly what it was designed to do. It is not a general purpose chip with 'some storage array-related functions' built in. It's a specialist vs a generalist. It's the classic "Spartans, what is your profession?" scene ... You might have thousands of potters, blacksmiths, etc but it looks like we brought more soldiers than you!!
Whilst I generally agree that an ASIC can be more tailor-made to a task than a generic CPU design, there is a point int he "brute-force" brings quicker solutions argument. I'm guessing it is also cheaper to manufacture. I'd also question the reliability claim - all hardware can fail, just make it double- or triple-redundant, please.
One of the main reasons that arrays need the 'brute force' of bigger / more CPU's is because of the extra functionality that is added in layer up interdependent layer placing big processing requirements on the arrays. The ASIC is damned efficient at what it does. Take thin provisioning for example - every other array has a list of considerations and complications when thin priovisioning, with 3PAR over 90% of customers thinly provision EVERYTHING because they are no performance or functionality issues with doing so. 3PAR has more thinly provisioned storage in customer environments than everyone else put together!
Yeah, a commodity chip can probably be manufactured cheaper than a custom ASIC, that's pretty obvious really. But the processor is one small part of the cost of an array so by the time customers have spec'd a solution and negotiated with their vendor of choice, I'd argue this is irrelevant. The processor is however one huge part of the effectiveness of the array so it is worth getting it right!
And in terms of ability to tolerate failures, 3PAR stands out from the crowd. Up to 8 way fully active controller architecture, no write through mode for a controller if you have 4 or more controllers, etc, etc.
Yes, I am in love with this box!
OK, I come from an engineering background, so I too like the idea of tailor-made tech. But I work for a business that says "Fudge the tech, if cheap'n'cheerful does the job for less than buy that". Businesses have business requirements, not requirements for fancy tech. If the "brute-force" generic CPU solution allows a product to be developed faster, at less cost to the end user, do you really think CIOs care how many levels of software need to be on top? If it works and does the job, all they will look at is the bottomline cost. Just look back a few years and see how surprised EMC were when NatApp filers started replacing EMC arrays in a lot of businesses, because a cheap'n'cheerful NAS actually started to offer many Windows users more for less than a traditional SAN array. I get that you think the 3PAR is great tech, I'm not arguing that point, but great tech is often not a winning selling point (think classic argument of VHS vs Betamax). Instead of telling me how great the thin-provisioning tech is, try telling me what business advantage it gives me (which will be hard as I think thin-provisioning is largely a waste of time, TBH, and far down the list of features I want in a storage device).
Matt, over 90% of 3PAR customers thin provision everything, with the average customer presenting over 2.5TB to their server estate for every 1TB of disk in the array. We have customers who are still in their first 3PAR array but would be in their 3rd or 4th EMC array based on how they operated prior to 3PAR. Using thin provisioning, thin conversion and thin persistence HP will often guarantee that you will need an array 50% of the size of your existing array if you move to 3PAR. Can you not see a business benefit in having half the amount of storage bought and spinning? Yeah, there is a performance consideration in terms of number of spindles but with wide striping and multiple active controllers, 3PAR is the fastest performing array of its' class. Don't get bogged down in other manufacturer's clumsy and intricate implementations of thin provisioning which dilute the benefits and make it more effort than it is worth - 3PAR was built on thin provisioning and you genuinely would be a fool to have a 3PAR array and not be using it! Business benefits in terms of the absolute ease of management, zero reclaim (never even written, not even reclaimed) giving VMWare functionality and performance improvements, wide striping maximising performance, multiple controllers with persistent cache meaning you don't go into write through mode if you lose a controller, etc, etc. There is a reason why 3PAR were winning huge huge deals with massive names that as a tiny little company they probably had no right to win!!
This is just the 3PAR refresh that was on the cards pre-acquisition anyway - ie. just delivering against 3PARs original roadmap. Assuming that is the case then it's only the T that gets refreshed as the F is fairly new anyway. This is the same approach 3PAR have been doing for years. Plus new SAS backend should be compatible with existing systems. Existing 3PAR customers will already know about this and new ones should rest assured that the platform will continue to be the best on the market.
Understatement of the year: "Our understanding is the the new arrays won't impinge on P6000/EVA sales."
You gotta be kidding. The current iteration of the EVA is probably the last ever. Any EVA customer due for a refresh is being recommended 3PAR. The guy in charge of storage globally at HP is David Scott - the ex-3PAR CEO - I don't see him thinking it would be a good idea to continue to develop EVA when that R&D cash could go into 3PAR.
HP said "EVA was the architecture for the last 10 years, 3PAR is the architecture for the next 10 years."
3PAR is awesome - and comes with the price tag to match!!
3PAR street pricing is comparable to NetApp, EMC, EVA in my experience - better if you take into account that you need twice as much physical storage from those other vendors. Perhaps you just need to brush up on your negotiation skills if your experience is different.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
