[UPDATE: Dec 16, 3:24pm ET – fixed some of the links for people using non-IE browsers – argh – IE]
There are surprisingly few good “big studies” of VMware and storage. Most tend to be very general (the big analysts) or very specific (and connected to a vendor in some way).
For fun, every year – I run a Virtual Geek study. I’m always amazed that ANYONE would respond, but respond they do (thank you!).
This year – participation was exceptional.
Here are the stats:
- 1935 respondents. Yikes. This is ~10x more than a lot of the “big official surveys”
- They ran the gamut when it comes to scale – ranging from 4 VMs to 200,000 VMs.
- the survey ran for ~1 month (Nov 10th – Dec 12th 2011)
- the sample covers the gamut when it comes to storage choices. It makes me happy (strangely?) that the readership doesn’t materially bias to EMC. It makes the feedback that much more useful.
- They came in from 65 unique countries, 941 unique cities – with a map shown below.
Questions asked included:
- If vSphere, what version?
- Do you use other hypervisors, and if so – which?
- Which storage vendors do you use?
- What protocols do you use?
- What do you like/not like?
- What advanced features (both at the VMware layer and the storage platform layer) do you use?
- What apps are you virtualizing?
- Where are you re: IaaS/PaaS?
- What cool stuff are you working on? (so much goodness, it’s a separate post)
- What cool stuff do you wish we would do (so much goodness, it’s a separate post)
- What is stopping you from doing more cool stuff (so much goodness, it’s a separate post)
… and MUCH MORE.
We gave away 10 Iomega SSDs, 10 Iomega IX2 Cloud Edition Units, 7 iPads, and a VNX3 3300 (winners announced here).
But – in the end – it’s about the DATA.
And as always, will always try to share…
- You can download the summary powerpoint here (if not using IE, here)
- You can download the raw data (I pulled out IPs and names/emails – but otherwise – all the raw data is there, so you can munge it for yourself if you want) here (if not using IE, here)
- You can download a PDF report here (if not using IE, here)
Here are some interesting highlights before the break (and my favorites after the break).
The first question (asked last year also) is about how the customers are adopting VMware’s core cloud infrastructure stack (ESX/vCenter). Its interesting in that the bulk of customers quickly tend to upgrade – in my experience, more quickly than other “core platform” type scenarioes. As you can see below, almost 60% of the respondents are already on vSphere 5. Remember that multiple answers were possible (as many customers run multiple versions).
This one was interesting to me. I’m less surprised to see the appearance of a lot of Hyper-V. Hyper-V is free of course, and included in a common guest OS (Windows Server 2008). But – still almost 50% – that’s impressive. What was more surprising to me was how many people said XenServer.
The percentage virtualized consistently moves up every year – BTW – I got the read-out from EMC IT – we’re at 86% right now, so ahead of the average, but not by much.
Here was the breakdown of “what array do you use”. I’m particularly happy about this – as it shows Virtual Geek has a pretty broad readership – and frankly, the survey would be less useful if it biased towards EMC. I’m glad to see EMC make a strong showing – with about 2x the next closest (btw, larger than our “natural” storage market share) People are always surprised that HP, Dell have about the same as NetApp – based on how active NetApp is the community relative to the server vendors, but I’ve seen this borne out by a lot of datasources. BTW – this is similar ratios as I’ve seen in the Goldman Sachs study and in other studies too (though they generally have smaller and more localized/vertical sample sets). The “Other” category is fascinating – I dig into this after the break – interesting to see who is in that….
This is the surface to whet your appetite. As there are ~2000 respondents, and some of the questions were more open ended – the detail past the link is… quite something. For those of you curious – curious re:
- how people are virtualizing mission critical apps?
- the perennial “protocol war”
- percentages of folks using IaaS and PaaS both internally and externally
- what cool things they are working on…
- what cool things they wish WE would work on (VMware and the storage community)
- what things are inhibiting them from cool things…
Here is the breakdown of how many ESX hosts the customers have:
Here you can see the number of VMs. The largest one in the survey had 200,000 VMs in total. WOW.
This chart shows the answers to “how many virtualized clients do you have”. To me it was interesting because there were clear groupings in the 10-50, 100-500 and 1000-5000 bands. It’s useful as we create reference architectures. It’s also interesting while we work on the really big VDI environments, and often it’s an area of chest-beating – they are very rare.
Before the break – I showed this table – of the array use amongst the survey respondents (remember, multiple choices were accepted – and in fact, many customer had more than one).
I noted that this is similar to other studies I’ve seen. Here’s the summary from the Goldman Sachs IT spending study which asked the same question (and interestingly did it in April/June of 2011). If you add up the columns in the “Official Unofficial” study – you’ll find that the ratios are actually pretty darn close (EMC had ~80, HP had ~40, Dell had ~30, NetApp had ~30, IBM ~15, HDS ~10)
Also fascinating was the “Others”. It’s a suprising 21% (to put it in perspective – that’s about the same as the VNX listing!) I “double clicked” on this to pull out whether there were standouts. Here are the most commonly mentioned folks…
- FreeNAS – 5
- OpenFiler – 24 (~1%)
- Iomega – 17 (~1%)
- QNAP – 36 (~1.5%)
- Datacore – 20 (~1%)
- Nexenta – 32(~1.5%)
- Nimble – 6
- Nutanix – 2
- Fusion-IO – 3
- IBM XIV – 6
- Sun – 5
- Whiptail – 4
First observation was NO there is no single large player (individually – the bigger ones are in the ~1% range) – but there are a lot of them. There seemed to be a grouping around “homebrew” (FreeNAS/Openfiler), the “SMB” options (QNAP, Iomega), and then some startups/smaller folks (Nexenta and Datacore being the most prevalent) – as well as some of the really new entrants.
Next up to bat – the perennial “protocol passionista” (who I hope people are starting to realize how silly they are as this is mostly irrelevant – use what works for you :-) For a “compare” – here is the data from last year:
Here’s the question from this year. Overall – very little movement. If you’re still excited about this – man, you have got issues :-) That said – on a personal level – interested in continued adoption of 10GbE. And if you want my vote – I’m a fan of using VMFS and NFS in conjunction (and leveraging their super-powers and avoiding their kryponites as the workloads warrant).
Was interesting to see how the respondents answered the question about what advanced data services they are using in their storage platforms. Thin is in (more than in past years). Snaps/Clones – yup. Surprised “flash as a cache” and SSD as a tier/autotiering weren’t higher – but I think it’s because I’m a little focused in the now. In most places around the globe, our arrays are going out with between 50% and 100% flash attach – but then again, if you look at the ratio of “current gen platform” vs. “N-1 gen” – perhaps that’s the root of it. Great to see customers using deduplication/compression features.
Leaving the storage arrays for a bit – I asked what features people are using at the VMware layer. Also great to see “Thin Provisioning” so prevalent. Remember, if your array can’t do it – do it at the VMware layer. And remember – “Thin on Thin” is perfectly OK – just make sure to setup monitoring at both layers. svMotion is helping customers every day (met several today) – so great to see that there. Also of course not surprised to see VMware snapshots so widely used. What DID surprise me on the one hand that SDRS is as high as it is (remember that only ~60% of respondents are using vSphere 5), and at the same time SIOC isn’t higher. My view – both are awesome features (so hey it’s great that SDRS adoption is so fast) – and there is little to no reason not to use SIOC if you’re on vSphere 4.1.
My favorite bit is where I ask what customers like/don’t like about various tasks/things. More than features – these are use cases – and what they find easy and what makes them “want to poke their eyes out” is very informative. In the table below, I’ve highlighted any "bad category” where it was north of 10%, and any “good category” where it was more than 20%. (BTW – for those of you who are interested – fascinating to compare with last year – which is here).
Storage cost is still one of the blockers to the customers – which highlights how important storage efficiency is. A fascinating run through the data would be to try to correlate with storage vendor choice… perhaps will do that.
I should stop being surprised – but determining capacity use, and more importantly capacity forecasting remains a problem. I wonder if people know how much capacity use reporting they can get in the datastore reports in vCenter. Capacity IQ now has storage capacity reporting – and expect to see more of this in vCOPs in the future. Also, ECC customers should check out ProSphere – which is (I believe a free upgrade and a ground up re-write).
Ok – onwards and upwards through the study! So – next I was curious about where customers where on their journey to cloud… which is partially correlated with use of “advanced” VMware capabilities and products, and partially correlated with virtualizing mission critical apps, deploying IaaS and PaaS models. So first of all – “advanced VMware products use”… Very interesting summary in the chart below. Think of the combination of the red, green and yellow as being some degree of use. The blue bars represent various degrees of non-use. Amazing to see that more than 1/3 of customers responded as using View, and when you mix in intent – it’s over 50%. The pattern is similar across the rest of VMware’s “infrastructure” related capabilities… Yet, when you go to the “2nd layer in the VMware layer cake” (cloud application layer), it drops off fast and furious. Also look how wide the “don’t know about it” bands become. To me this is interesting – as some of the capabilities in those areas are more transcendental. Perhaps this highlights a VMware challenge – and why at VMworld, some of the keynote content has people scratching their heads, and the app sessions are relatively lightly attended. Do infrastructure folks “get” that side of things? Should they?
Moving on to a more classic (but still critical) question – How are you doing re: “virtualizing mission critical apps”?
Again – to me, this is just fascinating (remember, red, green, yellow all indicate various degrees of use). Also, remember that in this case, the lightest blue means “not applicable”. The data is clear – people are virtualizing their SQL Server instances, Sharepoint, and Exchange servers – all mostly problem free. Of the customers who actually HAVE Oracle or SAP – about half are virtualizing them. With Oracle you can see that there is a bit higher ratio of people who had problems – but relative to those who “love it”, it’s a minority.
…On to VDI… You can see that about 50% of the customers are doing some degree of VDI – amongst this sample – a little skewed to View. I suspect that this is a little sample bias (personally out there, I tend to see about as much XenDesktop as I do View).
These last couple of questions were some of the most interesting to me. While we industry folks are pounding on the table saying “get to your hybrid cloud!” (and I firmly believe this should be an imperative for any IT org), it’s clearly still early days. Also, this chart makes me laugh whenever I hear people claim that clouds can ONLY be public, or conversely that public clouds can NOT service the enterprise (usually due to the scare-tactics of security and performance). If you look at the below – about 1/3 of the respondents already have a private cloud offering infrastructure as a service, and LOVE IT. Fewer are using external public clouds, but if you add both vCloud and non-vCloud options – it’s about 1/9th of the respondents who know that they are using it – and most are positive. I wonder how much “shadow IT” there is in this case the respondents don’t know about….
The same question asked about PaaS highlights the challenge that exists (similar to the earlier observation about use of the Cloud Application stack at VMware) – most people said “it’s not applicable”. This has me just scratching my head. It seems that PaaS is at least as important to any enterprise – after all, it’s the job of infrastructure to support apps – apps support business, and apps are what people touch. I’m going to contemplate over the holidays – is it possible for infrastructure-centric folks to “get this”, and be part of the answer, or will it bypass them completely?
There you have it. The “Official Unofficial VMware/Storage survey – 2011 edition”!
Thank you to all who participated – very much appreciated. I hope you find the results as interesting and thought provoking as I did!
Merry Christmas, Happy Holidays, and Happy New Year to all – have a safe holiday. Thank you to all the EMC and VMware customers out there. I’m still heads-down on 2012 planning work – but trust me – it’s going to be a fun and exciting year – we have a lot in store!