Transformation of the datacentre
Date: Tue, 09/04/2012 - 18:54
The datacentre is so critical to many operations that the first question must be whether to attempt any major transformation or whether it is just too risky to shift the foundations of a working enterprise. How do different vendors go beyond delivering the kit, to ensuring a smooth transition?
In previous NetEvents datacentre debates it was generally agreed that the enterprise needs open solutions – clearly a touchy topic, as it led to several “we are opener than you” scraps. In particular, those who argued the case for fabrics to meet the “East-West” communication demands of a virtual datacentre, were accused of stealth-promoting proprietary lock-in. So, what’s it to be: fabric or a better hierarchy?
Another question: every vendor promises more processing power, but what about the customers’ need to use less electric power – and other resources? And what about the use of public as well as private cloud – where do we draw the line?
Faced with these and many other balls for the CIO to juggle, Daniel Beazer will present our team of vendors at NetEvents EMEA press Summit, Garmisch, with some typical but contrasting enterprise scenarios. What would be their suggested deployment, and why is it better than their competitors’ offerings? They get the chance to fight it out, and you get the chance to vote on the outcome.
Panellists: Andreas Lemke, Senior Manager of Cloud Solution Marketing, Alcatel-Lucent; Sven Klindworth, Head of BT Advise Compute, BT; John Rollason, Senior Manager Product & Solutions EMEA, NetApp; Paul Griffiths, Global Consulting Engineer, Riverbed Technology
Introduced and Chaired by: Daniel Beazer, Senior Research Analyst, EMEA, Tier1 Research
Good morning everyone. My name is Daniel Beazer, as it says on the slide. I work for Tier1 Research and we are part of 451 Group. I cover the internet infrastructure, the datacentre stack – we have started a datacentre specialising in hosting, CDN, Cloud etc. I am here to talk about the transformation datacentre. I will just show you a few slides, a couple of high level observations that we have seen as a group about the datacentre market, especially in Europe. We will then bring the panellists up to the stage to discuss datacentre transformation.
Here are a couple of things that we are seeing in the market. I went round our group (we had a reasonably large size datacentre group operating in Europe before I came here) and the one thing that really popped out is that although there are a lot of datacentres being built, we do not know of any enterprise ones being built. It is difficult to prove a negative of course and being a cautious person, I would not say no absolutely there. We know of very few being built so there is a [trend] in datacentre market but it seems to be in the third party area particularly.
I put retail up there for those who are less familiar than I am with the datacentre world – which may be lucky. There is a difference between retail and wholesale in colo, generally taken to be the 5,000sq.ft. level. So, if your retail business is just someone saying, “I want a rack and 100meg”, that is the [Equinix] of [indiscernible] well the wholesale players are people like Digital Reality Trust. It is really a tale of two cities there. Also markets really stall for want of a better word and there is a very simple reason for that, which is the safest capital market. If you want to build out a 5,000sq.ft. datacentre or a 5,000sq.ft. pod in someone else‟s datacentre obviously [indiscernible] is very large.
It is interesting that one exception we have to the slowing in this market is a company in the UK called Infinity SDC and they are almost more a finance company than a datacentre. They can approach companies like Cable & Wireless, with whom they have just signed a big deal, and say to them, “We will do you a finance deal so this will show up as opex rather than capex on your books. It will be a wholesale deal, except it will not be capex.” That is the model they are using there and that is what I refer to down at the bottom.
I have there „West Wales growing fast‟. Mike Tobin, who runs TeleCity, probably Europe's largest carrier neutral datacentre operation (certainly in his opinion the largest carrier neutral datacentre operation) refers to West Wales very scathingly as anything that is outside London. Much to his chagrin, this particular section of the market that is slightly outside metro market seems to be doing really quite well, particularly when it is clustered round [slap] where there is a lot of interest.
That are a couple of high level pointers I am seeing. I just thought I would share with you some numbers. We do a lot of market sizing, there is a lot of talk about Cloud and I am afraid this section is going to be no exception. I thought I would share with you where we see the market. I do not know how visible this is to people at the back there but that big red block there is managed hosting, with dedicated hosting underneath. That little green bit at the top, which is pretty small even in 2011, is Cloud.
By Cloud we follow the NIST guidelines, which means we have a pretty tight infrastructure as a service [indiscernible] in Cloud and we are not seeing it as massive, 1.7 billion in 2011. This pretty much tells the same story, but what it brings across is that it is small but growing. Obviously in percentage terms it is growing hugely, but in cash terms probably not massively, only nibbling. Why is this? These numbers are a bit old and I should refresh them, but if you spin the numbers from the latest Amazon pricing you will get something similar. The truth is it is expensive. There is a very good reason for that. Take the hire care analogy, if you buy a car on a three year lease it is going to be much cheaper than hiring one for the weekend or day by day. The [indiscernible] there is almost [2X]. These figures here by the way are showing a workload running at 100%. That is extremely fiscal. It is running 100% all the time and they are factoring out the personnel involved [2N] (I think there are two people involved in both case as part of their infrastructure).
Finally, what we are seeing is not so much a massive enterprise adoption of public Cloud and infrastructure as a service. What we are seeing is the rise of a hybrid model and what that means is that, in our view, people are running a dedicated environment in tandem with a public Cloud or a private Cloud environment. For the typical enterprise workload, the things that are predictable and solid (payroll, billing and the like), they are using a dedicated environment either on premises, they are buying their own hardware.
Then for the spikier, the web scale applications, the unpredictable applications, they might spin up extreme instances using their hosting provider or using a public Cloud like Amazon. The classic use case of this is Domino Pizza in the UK. Some of those pizzas in the UK are very interesting. They have 600,000 people using Facebook now to order pizzas. They are saying that they take more orders on the web than they do by telephone, I think that came out last week. Of course although Domino‟s orders are quite spiky, they are also quite inherently predictable. When the Olympics are here this summer, or during the football world cup or cricket, they know they are going to have a spike in orders. In those cases they spike up with their provider, [Blackspace] in Slough. They spike up some extra instances to deal with those extra workloads and then just wipe them, take them down again once the Olympics are over.
To discuss those observations amongst other things, I would invite my panellists up to the stage.
The first question I want to ask is a Cloud question. Do you think in 10 or 15 years time all applications will be run from public Cloud, or Amazon, or a couple of other providers, and you vendors will only have two or three customers?
There are statistics that by 2014, 80% of applications will be in Cloud so we definitely see a strong trend of all the applications moving into Cloud. Therefore we see a strong need to develop the technology to rapidly set these Cloud application ups, both at the datacentre level and at the network level. This means that we have to fulfil the requirements of the customers. We have done a study with 3,800 IT decision makers and asked them what their major concerns are. The first three were performance (the same performance as local applications), security and ease of use. These requirements need to be met to generate that.
What about that, John? Do you think the rise of public Cloud means that people like Sven here at BT is going to be taking all your customers, your only customer?
We certainly have certainly seen a move towards broadly Cloud. If you think Cloud equals Amazon web service today, that is a very small part of all our IT markets in general. If you define Cloud as shared virtual infrastructure of some sort, whether on premise or off premise, then that market is growing phenomenally at the moment.
From a storage point of view, data keeps on growing. That is fabulous for a storage vendor but not so good for the end users. What we are seeing is that we will be taking more orders, we will be partnering more with the likes of BT because end users do not want to have to own and run their infrastructures. They are not ready to go and move to an infrastructure that has no service level agreement or guarantees. So yes, we will have less numbers of direct customers but in general, the workloads [indiscernible] into Cloud based infrastructure using that definition will continue to [indiscernible].
I would also like to take that out to the audience to see if there are any comments or suggestions. Maybe the thought of vendors disappearing, or at least any having two customers who [crush] them whipped everyone into a frenzy of excitement. Does anyone want to add anything or give their own predictions? Maybe by a show of hands, do we think all applications are going to be run from the Cloud in 15 years time? Do we think it is never going to happen? I personally think it is never going to happen.
My personal view on it is that I understand that outsourcing has only got a 25% share of the market. At the point when we were talking about every enterprise is going to outsource, here we are 10 – 15 years later and it has only got 25%. I think it will have a role but as to a huge one, I am not so sure.
The next thing I want to talk about is the datacentre, which is the route of all this Cloud. It is focussed around the datacentre and I would like to talk about the rise of these massive data factories that we are seeing with the likes of Amazon in Dublin. Microsoft in Dublin announced yesterday that they are going to double the size of their datacentre in Dublin which serves Office 365, Azure, Bing and many other Cloud properties. I think this has massive implications for the industry. In fact it has massive implications for everyone in this room because it is a very big change in the way that networks and data are delivered.
I would like to ask the panel what their opinion is of that, starting with Paul from Riverbed. How do you see that affecting your world?
It is a very interesting world and one of the things about those types of offerings from the likes of Amazon and Microsoft that you mention is, fundamentally, people are looking for choice. They will want to choose where they actually host their applications for the business, whether it is on a temporary basis, an as needed basis, or even in terms of test and development or, in some cases, for their actual production applications. The key gold four for us in these environments is to follow the customer‟s data, wherever they want to host those applications and the data that is associated with them. Wherever the users are geographically located and however they want to access that data, what we are able to do for them is provide them with a [lamb like] experience. Many customers will have a mixture of their own managed datacentres as well as public and private Cloud. We want to be able to allow them to get the same response of the applications.
What we certainly see in the market is a trend towards much larger datacentres than in the past. The Amazon, Facebook and Microsoft examples are really huge, but BT has just opened a new datacentre in Frankfurt for managed hosting and Cloud services.
The trend you have shown is 5,000sq.m. of rack space and these are huge datacentres.
What I would not agree on is that it is enough to have just one datacentre per Continent. What we see in the German market for example, and I know this is different across geographies, even if customers use the managed service or Cloud service, they want to be able to see the datacentres. They want to see the flashing lights on the servers and storage system and be able to go there once a year with their auditor and look at whether data is actually hosted and managed.
There are also many discussions about compliance and laws, for example do US companies have to comply with the Patriot Act, how does that contradict with new Regulations, etc? So, yes, I believe in large datacentres per country, not one being enough for worldwide traffic.
That raises some interesting issues. First of all, the Patriot Act. There are a lot of service providers out there saying, “We offer you protection from all these nasty regulations floating around there from the US”, but I would like to point out to everyone that, as mega upload proves quite dramatically, when the Feds come knocking on the door, service providers are very happy to roll over the do what they ask them to do, even if they come from America. This undermines what the likes of [Teesis] and others tell the likes of me all the time. I think data protection is fascinating but it is not the topic of conversation here.
A more interesting this is the server hugger business. We are always hearing about these Luddites, these server huggers, these bad people who stand in the way of this great march forward to the glorious hero of Cloud. In my experience, people do things for a reason and I am always trying to get a handle on why people are server huggers. The only real example I have is from my previous life when I worked in the industry for a hosting company. When looking at their datacentre, people very often use geographical criteria. It was not that they liked looking at lights in the datacentre because I think they were a bit more sane than that. Remarkably, what they did have was tape back up. More often than not it was because they were sending someone over once a week to back up to tape. That person had to drive over to the datacentre in Docklands.
When we talk about server hugging, why does it happen? Is it a problem? Is it something that is going to go away?
John Rollason [speaker not identified]
There are two aspects to that.
Despite what we all talk about every day, IT change has proved to be extremely slow.
There is no logical or technical reason to back up to tape. There is no economical reason, 25% of restores do not work but it is what people have done. It takes a lot for an organisation to move away from something that they have seen as belt and braces.
So I think that reluctance to change is something that we should never overlook.
The second point from what we see is that over the next 10 years everybody has different forecasts, but our rough estimate is that there are going to be 4000% growth in data. That is a whole other topic of what is driving that and I am sure that will part of the conference.
We were looking earlier at your comparison between the server on premise and the Cloud server, or whatever else it may be, it is not a very fair comparison because the on-premise infrastructure is only typically utilised at between 10% and 30% because people buy three times as much stuff because they are used to waiting. If you think of efficiencies at that level, typically flat to down budgets, with 4000% data growth there is no choice. Things have to change and that is what is going to compel people to change.
To your point about the cost model, I think the reason why 30% is thrown around as a number frequently as what their utilisation is, is just because they have to account for capacity of the spike so the beauty of the Cloud model means that they can run at 60, 70 or 100 [indiscernible].
Sven, is there something you want to say?
I would like to add a short, additional comment. Many of our customers come from the finance and insurance industries and they use some of the latest technology, infrastructures as servers, managed hosting, UPI applications, etc., but the CIOs tell us that all of the new technology, and it may be hosted with a third party, always come on top, but no old system ever gets switched off. They have systems for managing their customers, for sending out financial statements, etc. which were programmed 20, 30 years ago. They still run on old mainframes and what they basically say is “We want to buy all the latest technology from you”, but the old one never gets switched off.
The thing that really stuck out for me from Carolyn‟s presentation was that 70% of them may be stuck in branch offices. That is an extraordinary number but it does not surprise me. I think this goes back to the server hugging business in that they are not being irrational. I have worked in the field office of a multi-national company, be it a small one, and I know what it is like to have head office with its insane plans that do not work for you (or only work on their timescale). If you can have a server in a cupboard, you are really happy about it.
One of the key things here is to look at the industry in general and understand that this is how you describe them and the other technicians and datacentre managers in our customers in the IT organisation. They are doing their best minimise the risk of any of these application upgrades, any of the move from Legacy applications running on Global or whatever else to newer technology, they desperately want to take advantage of the cost effectiveness of new technology, updated applications and the efficiencies that those can provide for them.. However, they are also mindful of the fact that they are working with their own industry‟s crown jewels and if they lose any of that, not only is their company at risk but as a person their jobs may be at risk.
One of the things that we see quite often is using different types of technology. As John was saying, we have one with regards to backup. People generally want to avoid backing up to tape these days and we have seen that. We have taken some of our intellectual property used in the [indiscernible] optimisation devices and applied to them to the storage of backup environment. This then allows the customers to keep their existing backup application, which they know works, but take advantage of Cloud based storage services to provide them with a much more rapid backup and restore.
I am not trying to give the wrong impression here about server huggers as I believe I was actually one myself and for very good reason.
Germany is famously a place where people like to keep their equipment very close to them. Is this true? Why it is true?
What we have seen is that Europeans are much more conservative than other countries, such as the US obviously but also India, who are embarking much faster towards Cloud than European countries are doing. What we need to do as an industry is take these concerns of security, performance and availability seriously. We need to reassure people that these solutions can give the same performance and security results, or even better, than you have in your local privately owned datacentre.
The other thing is the cost savings. We are presently in a very tough economy and if you can save on administration costs with a Cloud based deployment, people will do it.
That is what we are seeing.
You are saying that there are better security solutions across the board in the datacentre and as you pointed out [Indiscernible]. I do not know if [indiscernible] once you intended them for. My question to you is, what kind of security solutions can you offer that ensure the corporate data of a Dutch company does not end up in the hands of the US government, who will use it for different purposes, for instance giving me extra competition?
What we need is that the customer can define policies such as how their data should be handled, for instance in terms of location as Sven was saying. That may not be so simple, you have maybe a main datacentre in your own country but what if it spikes and you need additional capacity? Do you allow these additional resources to be in France or in the Netherlands instead of Germany? Users need control over their Cloud deployment and that is the key.
In theory you are right, but in practice I am currently researching a case where Dutch data from Dutch citizens hosted in a Dutch datacentre by an American company ended up in the hands of the US government.
You are not talking about mega uploads.
Yet you are proposing a solution of keep the data in your own country. We kept the data in our own country and now when I cross the border, my medical file and other data is available to the custom officer. How can this be and how can you then incite a company to trust the data to the Cloud?
We all know that the Patriot Act compels companies with a US presence to hand over data, even if this data is outside the United States and host it in datacentres in Europe and other places. The first thing it is a matter of choice. Who is your datacentre provider? Who is your Cloud provider? There was a study by TNS Informa recently which indicated that 78%, of companies want to have their data in the local country.
Of course that means the provider also needs to be a company in that jurisdiction and not anywhere else.
You are saying do not use Microsoft.
It is not just Microsoft.
For certain purposes.
We have another question from the back.
Franz Kaiser– Fortinet
The Dutch may be a problem and the [indiscernible] may be a problem, but Germany has a bigger problem. Forget all your policies, we have very strict laws in Germany.
You must be blocked if you are 100% sure that in no circumstances data from a German company with critical issues because German datacentres are outside Germany. That means [indiscernible] and you have to be sure that everybody inside your datacentre has read the German datacentre gazettes otherwise you can forget the Cloud. The question is, are you able to guarantee technically to 100% under no circumstances date will go outside your Cloud, even if you have this Cloud in Germany, you must be sure that under no circumstances can they get out of another datacentre and inform of a catastrophe?
I am not sure I can fully answer the question but there may be a perspective on that.
Many people fall into the trap of assuming the previous server hugger/silo infrastructure is secure. From what we see, I think there is far more risk from your users doing things that you do not want them to do. If you take away the server from the server hugger, they can go and buy something off Amazon and start running it in Amazon. That is a far bigger problem.
So 78% are doing that?
The other thing we see in the UK which is of far more concern to most organisation is pure data loss, which is more down to internal users/internal issues rather than worrying about the US government. I know that does not answer your question, but it is just a perspective.
I do not want to get bogged down on the data protection but it is very interesting. In the hosting industry, and ergo the Cloud industry because it has really spun out of the hosting industry, most hosting companies have very close relationships with the local law enforcement agencies and that is for a very good reason. Very often they will have a shared hosting server that will have 200 customers on it so what they do not want is law enforcement agencies turning up at dawn and whipping the whole rack out.
To go back to mega upload, you were asking what should you do in Holland? Well, do not go to LeaseWare because they very politely just handed everything over when they were asked.
As there are no more questions we will end it there. Thank you very much gentlemen for your time.