first of all, congrats on this launch. Working as a Storage Architect for a Service Provider myself, I'm very excited about this launch, and am looking forward learning a lot more about the product.
There are a couple of things I would like to re-confirm and make comments on.
-GA is march 7th, correct?
-This version does not support replication outside the array? Only local?
We currently have lots of customers that require replication from their production environment hosted in our public cloud to other public clouds hosted in one of our other data centers (RecoverPoint or SRDF). Not having this feature on the VMAX Cloud Edition is a problem for adoption of this product.
This version comes preconfigured with 14 service levels, which cannot be altered. If these service levels do not work for a SP then you should concider a core VMAX offering.(according to the live blog).
So if a SP decide that they do not want to make the investment in one or more of these levels because the use case is not there, the SP can or just eat the cost of that service level and loose money on it, or go with a core VMAX?
I understand that there have to be some standards and limits around this, but a little bit more flexability would be great.
Looking at the spec sheet, it looks like it is based on a VMAX 10K, and preconfigured it comes with ~237 TB, but it does not list over how many spindles this is. In other words, with this model, how far can this array scale, and do you scale all tiers at the same time, or can you increase each tier as they require.
Keep up the great work and I look forward seeing a lot more on this product.
Hi Fred! Thanks for the feedback and your congrats ... it's always nice to know that what you're working on has relevance and a positive impact on others.
Now, on to your queries;
1. GA is 7 March 2013, correct; we are, however, in DA [Directed Availability] with customers purchasing VMAX Cloud Edition today.
2. Local replication is supported with the remote replication you speak of not far behind. I wish I could give you an exact date, but rest assured RP/VPLEX and related technologies are certainly being reviewed for inclusion in the next quarter or two and the roadmap is as robust as you would expect from EMC, but I take your point and respect your view regarding this being a barrier to entry for some customers. That said, the team wanted to ensure that everything shipped with VMAX Cloud Edition was fit for purpose and there's a tremendous amount of work occuring behind the scenes, as you might imagine.
Whilst remote replication is certainly coming, speaking as a former CTO of a service provider, my view would be there's nothing inherently wrong with a structured and deliberate approach to releasing what is, at the end of the day, a new product. Were we to be having this conversation a year from now, I would say we would have missed a very big boat ... but I'm prepared to bet we won't be havinit and we'll see the replication you and others are asking for.
3. With regards to the service bands, you're nearly there; what is being said is that there are 14 service bands available from Bronze through to Diamond, however it isn't required that a customer must buy all service bands but, rather, the service bands are designed to give a service provider [whether an external public SP or an internal corporate SP] a predicted outcome with a linearised price.
What this means in practice is, when we're using an SP as a specific use case, where the architect knows the [performance] + [desired capacity] we have a service band which will deliver predicted results [read IOPS, IO density] at at a known and linearised cost. In other words, Bronze costs [x] today, tomorrow, and next week as an 'atomic unit', with the complexities of media selection, engine upgrades, ordering more DAEs/chassis, harmonising engines against drive types abstracted altogether.
Hence an SP can then order chunks of Bronze, Silver, Gold, et al and need not worry about balancing engines, finding the dosh to include engine upgrades they weren't sure they would need when ordering more capacity, calulating RAID overhead ... you get the point ... you order more capacity and the VMAX Cloud Edition does the rest.
And that's got to be worth something, hasn't it?
How a customer, in your use case an SP, consumes the service bands and constructs products/services out of them will ultimately be down to the SP and you as an architect and service designer.
4. I'm not sure what you are referring to re the spec sheet, that it would be preconfigured with c237TB is new on me so I'll need to review and come back to you but I can tell you that the unit was designed to be; a) responsive to user need, hence increase of independent service bands is indeed possible [ie start with [x] Bronze, [y] Silver, and [z] Diamond but you want to consume Bronze in a 3:1 ratio over Diamond], b) the VMAX Cloud Edition platform [read portal and/or REST API] can manage multiple VMAX Cloud Edition units with exact number depending on what you're doing across what geographic locale, however, broadly speaking, if you said more than 9 complete units and less than 11 just now you probably wouldn't be far wrong.
A central design principle for VMAX Cloud Edition has been 'predictive not prescriptive' and we really do hope that we got this part right, but great feedback such as yours is hugely informative and a helps us to iterate so please do keep your views coming!
All the best,
VMAX Cloud Edition | EMEA Lead
Great question! And I'm delighted to say that the answer is ... it depends.
I'm jesting, but only partially; the VMAX Cloud Edition is designed to allow promotion/demotion of workloads between service bands ... or, in the language of VMAX Cloud Edition, 'dial up, dial down, dial in'.
Now, the only real caveat would be to ensure that Gold has been acquired prior to the dial up, but if it has been then a tenant in Bronze or Silver can dial up to Gold no problem.
Where we see customers making the biggest differentiation, and I'm thinking about both traditional service providers AND enterprise customers, is in how they will meter and bill tenants for dial up/down/in capability.
One could argue [I certainly would!] that a unique value prop would be in allowing a tenant who requires Bronze 361 days each year but Gold for the four billing days per annum to dial up without charge ... but I might digress into discussion of enterprise grade cloud v 'public' cloud so I'd best stop whilst (hopefully!) still ahead.
Thank you Matthew, your answers are very clear and understood.
There is a big difference between working for an Enterprise and working for a SP when it comes to cost. For an Enterprise you buy 2,3, or 5 years worth of capacity and add later if you need it. In the SP world you are talking about months of capacity. Upfront cost and capacity add cost are very hard to predict and model, and it becomes exponential worse when you add multiple storage tiers.
Seeing that EMC is making most of that go away is a great time saver, and I will be glad to have that no longer on my plate.
Regarding the ~237TB in the Spec Sheet. Most likely I misunderstand what the table is trying to explain. Here is the URL to the pdf.
The table is on page 3: http://www.emc.com/collateral/spec-sheet/h11495-emc-vmax-cloud-edition-ss.pdf
Great feedback and I couldn't agree more; the 'perfect storm' of working within the SP industry means that one is constantly trying to harmonise what was, before VMAX Cloud Edition, nigh on impossible ... pricing products and services for future consumption with 'just in time' procurement [where possible] and allocation is incredibly difficult in the absence of the volume which would be required to achieve the price per unit required for an SP to be competitive. A maddening process, to be sure, and I can't tell you how delighted I am to hear someone from the industry I know so well hold such a positive opinion of what we're trying to do to help.
I have reviewed the Spec Sheet and, whilst I would love to take credit for solving the conundrum, it was my entrepreneur wife who solved the mystery for me ... she identified that you were adding all of the service bands together.
Now, whilst I am not technically a sales person [I just sell ideas. ], I'm certain that I know more than a few sales folks who would be delighted if you bought each service band!
However, this is not the case and the service bands are designed to meet your needs in designing products and services. The figures given on page 3 are the capacity increments by which the service bands are acquired.
For instance, if you required; 300TB of Bronze 1, 100TB of Gold 2, and 20TB of Diamond, your BOM would look something like this:
1. QTY 30 Bronze 1 // Calculated by dividing 300TB by the 10.3TB increment chunk gives us 29.13 units required, so we must round up to 30 units.
2. QTY 4 Gold 2 // Calculated by dividing 100TB by the 26.1TB increment chunk gives us 3.83 units required, so we must round up to 4 units.
3. QTY 12 Diamond // Calculated by dividing 20TB by the 1.8TB increment chunk gives us 11.11 units required, so we must round up to 12 units.
I've just now realised that in trying to illustrate how we use the increment chunks, I inadvertantly selected example increments which required me to 'round up' each time. This is not by design! In fact, I'm certain that a good architect would be able to organise the acquisition much better than I to avoif rounding errors, but hopefully this gives you an idea.
Thanks also for highlighting this issue, we'll of course revisit our collateral to ensure that we are more descriptive and less confusing in how the chunks and increments operate.
Hope this helps and keep the feedback coming!
I'm just getting my head around this thing, as I only had the time to start investigating it this morning. Let's see if I have my understanding right.
The array and management servers are essentially "free". You (EMC) put them in the customer's (my) datacentre so that I can lease space of you. I then buy licenses of you to use certain amounts of storage each with a guaranteed performance level.
Lets take a hypothetical project in my company where I need to sell Silver/Gold level performance to customers who can choose 1, 2, 3 and 4 TB of space.The capacities available for the Gold and Solver service bands do not map on to that at all. How would VMAX Cloud edition help me there. Would I just have to assign each customer a Silver-1 or a Gold-1 and rely on thin provisioning? Lets say I am also limited to VMFS3 (thus a little under 2TB LUN size). Can the service bands be cut up internally?
Lets also say my project takes off and I reach the limits of what a single array can take. Does a new cabinet auto-magically get shipped to me in good time for me to install it before the space runs out on the first one?
Hope you answer these questions in a way I like - for a grumpy FLOSS fan I am un-nervingly excited about this product
Great questions! Well, there are a couple of questions in there so let's answer them one at a time .
VMAX Cloud Edition is still sold through a "CAPEX" model, so you can still buy the array, though you can now do so at much more granular increments with the Service Catalog. Now, that's not to say that EMC doesn't offer other financing models - because we do - but it does mean that you can still purchase and outright own the storage array!
In either case, now you've got some gold and some silver on the floor and you have customers who need capacity, they can absolutely request storage at whatever capacity points that they want - so yes, while the service band capacity increments dictate the granularity for configurations, they can absolutely be cut up much more granularly for individual use by tenants.
And once one array is at its limit, you can order another increment of your service levels and they'll just arrive in a new array. It all plugs into the same view and gets managed as if it's a big pool of resources, so neither you nor your customers have to deal with selecting which array you want to provision from, etc. You just select gold or silver, how much, and off you go - Cloud Edition does the rest of the work.
Hope this answers your questions!
So I could by a VMAX CE with (for example) 50TB of Gold, and 100TB of silver. I only pay for that, but the array arrives with some standard EMC config that is "hidden" unless I count up the drives calculator in hand (which I would because I am nerd)?
6 months down the line, I have had more heavy duty clients than expected, I can order another 50TB of Gold and it just appears in on my array (assuming the above standard config has the capacity)? Kind of similar to VMware, if I get a higher license level, I put the new key in and the extra features appear.
If my understanding is correct then this really is awesome, it puts many bad (re. fortress building) storage admins out of a job (and out of my way).
Not to take anything away from what my Right Honourable Colleague, Sonny Zhao, has written, but I thought I would take a few minutes and record a short video for you (and anyone else interested!) re your queries specifically and how VMAX Cloud Edition works more generally.
Please do let me know if you've any further queries, I'm always happy to help.
nice movie with a good explanation.
One thing I would like to comment on, and would like to get your feedback on,
When I look at the IOPS per tier and in specific at the diamond tier. The max IOPS is only still at 4 IOPS/GB.
Amazon for instance is selling their high end band at 10 IOPS per GB, up to 2000 IOPS per volume.
What is your/EMCs view on that? What is the reason that the highest tier was set at 4 IOPS/GB, is it because of the cost and the capabilities of the array, or is it because when you studied all the VMAX arrays out there, you found that 4 IOPS/GB would cover the majority of the use cases.
I am asking this question because I get asked on a daily bases by our sales team, why we cannot (or want to) sell 10 IOPS/GB and directly compete with Amazon.
I have some answers and ideas on this, but would like to hear your opinion on this as well.