During the last year or so I have had my eye on the HP Moonshot platform. I followed what other community members had to say about it, how HP marketed the product and of course the announced partnership between HP and Citrix, introducing the Workspace Pod series at Summit not that long ago. I was rather skeptic, and I guess I still am. On the other hand, I also acknowledge that times are changing and that, if the use case fits, HP Moonshot might be a valuable solution, it just depends.
I actually wrote this article about three weeks ago and hadn’t gotten around to review and/or post it since, with no specific reason really, other than being busy working on all kinds of other projects, professionally as well as personally. I wasn’t expecting any announcements on the Moonshot platform, since it has been (very) quite for some time now, so I wasn’t in any hurry. And while nothing new came from HP, or Citrix for that matter (WorkspacePod), there was this podcast not too long ago that got the whole VDI / GPU discussion going again, including Moonshot. To be clear, this is not a reaction to that, at all, it’s just something that recently came on my path, which got me thinking.
Hands-on experience…
I know that a lot of you out there feel that you (at least) have to have some kind of hands-on experience to be able to fully appreciate and ‘judge’ a product. I agree, that’s by far the best approach. However, that doesn’t mean you can’t have an opinion on something you have never build or worked with before. You just need to make sure to base your thoughts on facts rather than assumptions. In fact I think, no let me rephrase, I know, that a lot of blogposts are based on just that, plain facts I mean. And that’s fine. Unless you are overly negative and burn a product or solution down to the ground without having done proper research, you might want to re-evaluate your opinion.
Hands-on can be hard
I mean, judging a book by its cover, what does that mean? Really? If you have gone through the manuals, the support docs and forums, have seen the product in action, have read multiple blog posts of others who were in the lucky circumstances to have used the product, etc… Than you could argue that ‘the book’ has been read, at least that’s how I feel. Will you have all the details? No. Is that relevant? It might be, it all depends on what you have to say. Don’t get me wrong though, I’d still prefer hands-on any day. But I also feel that there are a lot of examples where hands-on experience with a certain product is hard to get, and Moonshot is probably a good example, but that shouldn’t stop us from writing about it. Do I have any personal hands-on experience with HP Moonshot you ask? No I have not. Anyway, back on topic.
The unpublished
A while back I wrote an extensive blogpost in where I compared HP Moonshot (VDI only), with some of the better-known hyper converged solutions out there. Basically virtual vs. physical. In the end I concluded, based on several examples, that virtual, for me personally, came out on top. And if you are wondering to which specific blogpost I am referring, I never published it. After writing it I felt it was to early and that I probably should wait it out just a while longer before jumping to any premature conclusions. And I’m glad that I did. While I still feel that in most cases virtual comes before physical (read HP Moonshot, VDI and XenApp), in some cases HP Moonshot does the trick, and does it well!
A few other things first
Before I jump to my ‘use case’ example, I’d like to first share a few thoughts on Moonshot. And no, it’s not my intention to burn it to the ground.
First of all, stating that with Moonshot you get rid of the Hypervisor complexity doesn’t cut it for me. Sure, you won’t need a hypervisor for your XenApp / XenDesktop workloads, saving you the trouble of sizing your virtual platform accordingly. This saves some time, potential frustration and money on licensing. However, this is something we have been doing for the last five to ten years or so, so we should be quite good at this. And while some experience and common-sense are important, there are some great sizing tools out there to assist us.
And even with Moonshot, take the HSD (XenApp) orientated 710 cartridge for example, some ‘sizing’ will still be needed to see how many users will be able to run a certain workload on a single blade while still offering an acceptable UX. Numbers based on standard workloads only tell you so much.
Secondly, Moonshot is all-physical, so you will still need some sort of a base platform in place running all of your infrastructural services like Active Directory, DNS, DHCP etc. but also your StoreFront servers, Delivery Controllers, Database server(s) and so on. I think it’s safe to say that most companies today will have these roles virtualized.
And thirdly, and building on top of the previous one(s), hypervisors are a commodity nowadays and have been for some time, while they can be complex in some cases they have been around for years and years and knowledge is wide spread. For most of us (and here I mean the sys admins, consultants etc.) managing a hypervisor is a daily task and business as usual.
Moonshot however has not been around that long, and I have still to find the first consultant or article telling me they have set up and configured Moonshot without any issues or the need for HP support. If you look at some of the steps involved, the physical aspect, cabling, PXE, firmware (for the chassis and cartridges), networking, the base image, PVS, WDS (at least for the first node in the chassis), HP chassis management etc. some skills are needed, still manageable of course, but I wouldn’t call it straightforward.
What else?
I’m sure (in fact I know) things have improved (and will improve further, some nice wizards have been introduced for example, to simplify process) from an initial setup perspective, including overall (ongoing) maintenance, which I guess, comes close to managing a ‘normal’ virtual XenApp / XenDesktop Site except for a firmware update here and there. But easier, or less complex than a hypervisor (which you probably still use anyway)? I have to say no. Especially when hyper convergence comes into play, the initial setup and overall management process will be simplified even further.
Though it might get slightly more complex when GPU’s are involved (but still very doable) which is one of the main pros when it comes to Moonshot. As mentioned earlier, you also won’t have to size (over-committing resources and what not) your platform to run a certain amount of VM’s, which is often the ‘hard’ part. Moonshot comes with pre-configured cartridges and scales linearly with regards to performance, storage etc. although most hyper converged appliances will offer a similar guarantee once properly sized per appliance.
It is also (often) argued that Moonshot can, or will, safe you money when it comes to power consumption, but when compared to some of the better know hyper-converged appliances I doubt if there is much difference (I only did a quick check) if at all.
Virtual = more flexible?
I also feel that a virtual environment offers more flexibility. You need a few more VM’s? Sure, I’ll spin those babies right up, go and have some coffee. With Moonshot however, you are limited to the amount of cartridges you have in your chassis. If you are out of cartridges you will have to buy a few new ones. And in most cases you probably won’t have your chassis fully loaded with cartridges if you only need 15 or 20 to start with.
In fact, they sell per 15 at a time if I’m not mistaken, although I’m sure you will be able to cut some kind of a deal with HP on that. What if your chassis (it can hold up to 45 cartridges in total) is full? You will have to buy a new one, and those aren’t cheap I can tell you, and this goes for the cartridges as well, especially when you have to purchase 10 or 15 at a time. And I know that with virtual you can run out of compute as well, but it just isn’t the same.
Due note that there is a BIG (list) price difference between Europe and the United States. My point is it’s less flexible, with Moonshot up front scaling becomes very important. The cartridges are one purpose only, you can’t use any of the ‘left over’ or ‘spare’ compute that isn’t used for XenApp or desktops in any other way as we can with virtual. Potentially making your XenApp servers or desktops more expensive than they need to be. For me these are some of its main drawbacks combined with some of the (list) prices I have seen. Why not spend the same (or a bit more perhaps) amount on virtual?
Again, Moonshot cartridges come pre-configured hardware wise (some more details on this in one of the next sections below), which has its advantages, but what if you need to scale beyond that? It has cons as well. I wouldn’t say that the ‘use case’ is extremely specific, but it’s something to be aware of nonetheless.
What do I like?
Ok, enough ‘negative’ talk, although it’s not meant that way. What are some of the (main) advantages that Moonshot has to offer? Well, predictability and an excellent user experience (isolated workloads and no Hypervisor overhead) to name two (or three). If you look at VDI for example (although in the case of Moonshot it is actually referred to as HDI, Hosted Desktop Infrastructure, since it’s all physical, no hypervisor involved), each user will get a physical CPU/GPU (AMD Opteron™ X2150 APU, 1.5GHz, (4) x86 cores, and Integrated GPU with AMD Radeon HD 8000 Series Graphics) 8 GB of memory (DDR3 PC3-12800 SDRAM (1600 MHz) and flash like storage (32 or 64 GB of integrated Solid State Storage (iSSD) per server) best thing is, it’s on a one to one basis. Now if that doesn’t get things moving I don’t know what will.
The same applies to the XenApp cartridges; each (710) cartridge functions as a XenApp server (one cartridge equals one node) and also has a physical CPU/GPU (Intel® E3-1284Lv3, 1.8GHz (3.2Ghz Turbo), (1) x86 cores, and integrated graphics with Intel® Iris™ Pro Graphics P5200), 32 GB of memory (DDR3 PC3L-12800 (1600 MHz) SODIMM Low Voltage) and flash storage (120 GB or 480 GB of M.2 solid state storage per cartridge)
If you want more exact details on the 700 and 710 cartridges, the Moonshot chassis etc. I suggest having a look here. Due note that the 710 cartridges can also be used to host other application delivery and/or video transcoding workloads, and of course Server VDI is optional as well.
With HDI one (700) cartridge holds 4 nodes, each node being a single separate Hosted Desktop and thus consists out of the specs mentioned above. So with 45 cartridges in a full chassis you are able to have a total 45 x 4 = 180 Hosted Desktops. Or 45 XenApp cartridges, here the user workload (types of applications) will determine the total amount of users per XenApp server (we are currently testing with 30 on average, including HD graphics).
With each user having its own physical CPU/GPU, enough memory etc. (when dealing with HDI) the overall user experience will be top notch, as to be expected. And because it scales linearly it’s very predictable, you also won’t have to deal with boot storms for example, which is another big plus.
And again, you won’t have to scale / guess / test how many VM’s your virtual environment can handle, it’s 45 XenApp servers (here you still have to scale the number of users per server depending on workload) or 45 times 4 regarding one to one Desktops (HDI).
As a side-note, with virtual, and there are many examples out there, we can also provide our users with a more than average overall user experience. You might not end up with the exact same compute configuration per VM or the exact same numbers from a performance perspective (this would probably make it too expensive) but in many, or most, cases that isn’t needed anyway, especially with all the hyper-convergence platforms available today, which (because of their architecture) are performance boosters on their own. And as for GPU’s, whether it’s HSD or VDI, they are more than a nice to have for sure, but mandatory? I think not.
Citrix Service Provider
If we take it one step further and look at it from a service provider perspective we can now offer our customers a real Windows 7 or 8 / 8.1 cloud based desktop (though the 700 cartridges are only officially certified for Windows 7 today) without to much hassle. No sizing is required (which can also be a limitation if you need ‘heavier’ servers / desktops) and performance is guaranteed, sort of. When using XenApp (Moonshot supports Windows Server 2008 as well as 2012) customers can also bring in their own licenses for Office 365 for example, another bonus. There are other ways in achieving similar results, but this certainly sounds appealing.
Multi tenancy
Offering XenApp hosted shared desktops from a cloud perspective isn’t new, it has been done for years, and very successfully I might add. Offering a Windows Desktop OS however, isn’t that common since you will need to use separate dedicated hardware per customer / tenant. Microsoft licensing doesn’t allow a desktop OS to be offered from a multi tenant / shared hardware platform.
When we use Moonshot for this, all you have to do is dedicate a full cartridge to a customer / tenant. So even if a customer (only) needs two HDI desktops for example, you will have to dedicate a whole cartridge (which consists out of four nodes) to that tenant exclusively. Microsoft has verified this, but… you will have to get an official (written) statement from Microsoft stating that they allow you to offer this configuration to your customers. Of course this still isn’t perfect, but gives us a bit more to play with.
Enter lease…
Yes, I still prefer virtual, but what if…
We could lease a HP Moonshot cartridge, a chassis (or multiple), all the cabling and other components needed to build a full HP Moonshot architecture next to our existing virtual or physical platform, for a standard (low) monthly fee? I know HP has several financial offerings on their website, which I didn’t bother to have a look at (though I should have).
But it was only a few weeks ago that I found out (while talking to a few of our own engineers closely working with HP on a Moonshot PoC) about the various options we (as a company) have when it comes to leasing HP equipment. I’m sure these offering will differ per company and will depend on the contract and/or relationship you have with HP as a vendor, but it’s definitely worth checking out.
Using this model, if I run the numbers again the amount per desktop (shared or dedicated) per user drops immensely, equal or cheaper then virtual even (and this includes several ‘spare’ cartridges as well). I won’t go into any specifics as I don’t know the official statement HP has in this, and again, it will probably be different depending on the company you work for, but I can tell you that it’s way more flexible with regards to the number of cartridges you need to lease compared to the numbers you need to buy. Also, you won’t have to worry about exact numbers. Is your chassis full? No problem, here are your next set of cartridges (and you won’t have to order 15 at once) including a fully armed and wired chassis!
HP engineers will mount the chassis and any cartridges that come with, take care of all cabling, network setup etc. and will help you configure everything so you will be up and running in no time (or so they tell us). Know that these services are also available when you would like to purchase Moonshot. Or, if you prefer, you can do all this yourself with HP as a back-up resource, so to speak.
All this, in our case anyway, comes with a four-hour service level agreement including onsite support, replacing parts etc. when needed. For now, although we haven’t got it all figured out yet, we are primarily focusing on the 710, XenApp, cartridges. Note that the lease agreement also includes upgrades to any new cartridges that may de developed over time, the same applies to the chassis.
Bottom line
If you, as a company, are able to get your hands on a HP Moonshot lease agreement, at a reasonable price of course, then this might be something to consider. Your users might or might not necessarily need (but will certainly benefit) all these physical resources, including GPU’s, Flash storage etc. (in the case of HDI) but if it comes at a price equal or cheaper then virtual, who cares? It goes without saying that this applies to your internal business users as well as any cloud customers you might have.
Of course there are still some things you need to consider. Virtual will still be more flexible when you need to scale beyond the physical limitations of a Moonshot XenApp or HSD node with regards to compute and storage, the overall maintenance efficiency with virtual will probably be a bit higher as well, you can use MCS and don’t have to worry about WDS combined with PVS and with virtual we can run all workloads, you won’t need a second (separate) platform, the initial setup and configuration will be a lot easier and ‘known’ to most etc.
Pick your poison.
13 responses to “HP Moonshot revisited. Well… by me anyway”
Moonshot seems very much like a solution looking for a problem, really. The future is software-defined-everything. Yes there are exceptions, but desktop visualization is not one of them. As for the leasing model, at the end end of the day you pay it, no matter what. This is HP, and what they do is hardware. Good for them, but I predict this is going to fail big time. 3 years from now it will be forgotten and buried. Or they come up with a better use case, who knows.
Fully agreed on the first two. The lease contract is pretty tight and not long term as in years and years. It also includes all future developments on the cartridges, chassis and everything that surrounds it. But yes, they won’t give it away for free of course, it is way more flexible though and thus doable to take in a product like Moonshot, at least for now. Don’t know if it will be forgotten and buried, it might still be there just in another form. We might consider Moonshot to be a 1.0 product in that respect. It needs to become more flexible.
Fair enough. If they price it aggressively trough low leasing rates and ‘free upgrades’ they might get some traction on the market, foregoing any profit during the first few years. Which is ok if it helps establishing the product on the market. But I’m sure the bean counters at HP have a close eye on sales figures, and if it doesn’t take off during the first 2 to 3 years they pull the plug.
As Chris mentions in the comments, there are use cases outside of VDI that could make sense. Time will tell.
They will have an eye on the sales figures all-right, no doubt :) And in the end, like you mentioned earlier, you will pay for the product one way or the other. Leasing just makes it a lot more flexible, in multiple ways.
“have read multiple blog posts of others who were in the lucky circumstances to have used the product”
There’s plenty of marketing message out there, but very little information from somebody who isn’t being paid by HP or Citrix. I wish there was more.
Yes, I share that thought. But there’s probably a reasons for that. Although I don’t know of any numbers, I just can imagine it being sold much. Perhaps an assumption that I shouldn’t make. I know of a few CSP’s using Moonshot, but I have yet to come across the first company / customer who are interested in, or already using HP Moonshot. The social media channels are fairly quite as well. While it most certainly has potential with regards to performance, predictability etc. in its current form I don’t see a bright future ahead. But of course this can al change within a few months time. There is a lot coming (and going on) with regards to CPU/GPU integration. All for the best if you ask me.
I guess those CSPs aren’t big bloggers, then?
The environment I’m supporting is a small business. They’ve got 3 chassis. Moonshot is simultaneously refreshing and disappointing in its quirkiness. The administrative interface to the chassis is amazingly un-ambitious. On the one hand, there’s not much in the way. On the other hand, there’s basically no reason to log into the admin interface, because it has so few levers and buttons. I’m torn between hoping they leave it alone, and longing for more features.
The power situation is a mess. There are deployment modes in which A+B redundant power isn’t enough to keep the chassis happy. HP’s fanatical adherence to their “common slot” power supply is so strange. I’d rather have an PSU that does the job than one that interoperates with DL360s.
The switching OS in the non-“c” switch models has improved in recent months, but only to the point of “okay, I can use this.” It has not yet reached “this works correctly.”
Thanks for your input Chris. And no they are not :) This doesn’t sound very positive from an administration / setup point of view. How about your users? Any complaints, does it provide the UX you guys were hoping for? How about HP support, had to deal with them as well? Thanks again, regards, Bas.
We’re not hosting desktops on the Moonshot, so there are no users to worry about. It’s hosting application backend stuff, so nobody can see it.
My experience with HP has been good, but not because the front-line support is up to speed. Rather, it’s been good because the problems have been real product issues, and the product folks responsible for it have stepped up to make things right.
I see, I’m still a bit ‘stuck’ in the HDI / HSD mindset :) Glad to hear that support has been good. Thank you for sharing this.
You’re not the only one who’s stuck with that mindset.
In addition to being disappointed at the lack of end user feedback, I’m bummed that the only messaging from HP is around Moonshot used as virtual desktops. It’s like they forgot that Linux runs here too. Actually, if you look at the RHEL version (6.5) officially supported on some of the cartridges, maybe they really DID forget.
Though I’m not that familiar with the rest of the Moonshot portfolio, you would expect HP to be :) But indeed, most marketing related content that goes around is aimed at desktop use, while lately it has been rather quite around Moonshot. Must be something new coming our way soon.
Hi Bas
We are setting up our new environments with the moonshot 1500 at the moment.
And we think it’s a cool solution. We bought 4 chassis and
want to share it for different costumers, all in banking or finance. First farm
will be a XA 7.9 only for browsing.
One of the reasons we chosen moonshot is the GPU on each
cartridge, browsing, Banking apps, Java, Office and Adobe require more and more
GPU-Power.
Bank is rolling out 4K monitors, and want to watch movies
in HD quality.
In our own environment XA 7.9 Terminalserver, byod
environment we have 4 K2 cards vgpu and the admins need it!! i can tell you,
but buy Cards for 1000 CC user would be much more expensive, we think.
There are some cons, and we had to learn a view things
new, but thats what we do every day too? ILO is not so cool, only the new cards
getting better, Admin Console is – i hope – not finished yet, we don’t use WDS. Our network team does not like HP switches, but we handelt it by our self. No
powermanagment over Citrix Studio and integration with PVS – bad! But HP and Citrix
are very kind and try to help where ever they can!
i know two other companys here in Switzerland, they
have costumers running on moonshot, and i thought
Citrix themselves is using moonshot too
Setup for XD 7.9 is planned, can tell you at the end of the
year about that.
Maybe
i could make a little session about it, next year at the E2EVC in Prague?
greets
Thorsten