The centre of attention

08 October 2018

According to UKFast, there are three vital factors that make a “great” data centre: fire suppression; a high level of organisation and cabling standards; and power supply plus redundancy.

According to UKFast, there are three vital factors that make a “great” data centre: fire suppression; a high level of organisation and cabling standards; and power supply plus redundancy.

The colocation data centre market in Europe saw “exceptional” growth during the second quarter of this year and is set to have a “mammoth” year, according to CBRE.

In its EMEA Data Centre MarketView for Q2 18, the company – which claims to be the world’s largest commercial real estate services and investment firm – said that 177MW of new colo supply will be brought online during 2018, and that take-up during the first half of the year hit record levels at 87MW.

Across Europe’s four largest colo markets – Frankfurt, London, Amsterdam and Paris (FLAP) – CBRE found that total supply reached 1,256MW at the end of the quarter, and that 50 per cent of new supply landed in London. As a result, the report states: “Any political and economic uncertainty in the UK, with regard to an exit from the EU, has not yet been felt on the London colocation market”.

So when considering colo services (and indeed any other data centre services) should you look for providers based in the capital?

Steve Hone, CEO of the Data Centre Alliance (DCA), says while London is still a popular data centre location, the quality and quantity of power in the city is starting to become an issue, resulting in data centre providers opting to build further into the commuter belt. "The attraction of these out of town locations is simple: lower land/real estate costs and more power availability."

Hone points out that at any one time, the National Grid only has about three per cent capacity held in reserve, and while this has so far not led to brownouts, any unplanned strain on the grid caused, for example, by a cold winter spike, could leave more than just the capital susceptible to unplanned power outages. 

He continues by saying availability to international telco networks continue to make London popular in terms of a data centre location. But generally, with so much choice now on the outskirts of the city offering both competitive pricing and the promise of higher power density per rack, Hone believes it would be "mad" to automatically make a beeline to just Docklands unless your business needs absolutely demanded it.

He also notes that while most data centre providers remain busy, there are some who appear to be struggling.

"These tend to fall into two camps. There are legacy data centres who, despite operating good facilities, simply have too much space and not enough power to deliver the higher rack densities clients are now seeking. This can only be solved by giving them 200 sq. ft. of colo space just to host five 6kV racks, which is far from ideal.

"The second group is the single site operators who also have high quality facilities with all the power required but find it difficult to offer secondary facilities or backup/redundancy locations. To counter this threat, these independent operators have been partnering up so they can offer the client an end-to-end solution and compete on a level playing field with the ‘big boys’ who already have multiple DC location offerings."

One of those “big boys” is US-based Equinix. It operates more than 180 centres globally including a dozen facilities in Manchester and London. When it comes to choosing a data centre, the company’s UK MD Russell Poole says major enterprises like to see service providers offering a wide spread of locations especially in major metropolitan areas. “By situating a data centre at the digital edge – where the physical and virtual worlds come together – enterprises can be close to their customers and therefore deliver fast and secure services.” 

As an example, Poole says Equinix’s Slough campus is “strategically situated to be near to Canary Wharf” and City firms. “The recent decision to expand ourLD4data centre at this campus came as a direct response to the increased demand from financial services companies, both from the UK and all over the world, wanting to connect to their customers, partners and suppliers in a single financial ecosystem.”

He adds that Manchester is an ideal data centre hub in the UK because it not only sits at the intersection of the country’s highly resilient ‘figure-of-eight’ fibre optic network, but also because it is a comms hub linking other major UK cities as well as the US and Europe. 

Volta MD Jon Arnold also believes that, along with cost, security, reliability and connectivity, location is another decisive factor businesses will need to consider when choosing a data centre. Arnold reckons Volta’s “many” customers were attracted to the company because of its data centre which is at Great Sutton Street in London. “The facility is uniquely designed to provide a flexible, secure and sustainable IT environment to cater to businesses requiring a Central London location – such as those in the financial, media and content industries – as well as ultra-low latency. Our location has meant working with the world’s leading carriers to deliver ultra-fast, resilient connectivity to users of latency-sensitive services.”

Arnold goes on to say that depending on the needs of the business, data centre location can be a deciding factor as proximity can have wider consequences that need to be taken into account. These could include ease of access for the company’s engineers, seeing if there are plenty of diverse fibre paths to the building, if the centre is in a flood plain or on a flight path, and also how close the facility is to the customer and its end users if Edge computing is required.

But Arthur Howie, product manager at aql telecoms, points out that while regulatory requirements around security may affect location, this has otherwise become less important. “Some organisations consider that distance from the service provider is a key issue, and whilst this may be true for some applications the trend towards cloud services and colocation show that this is sometimes an unnecessary fear. 

“The high capacity, low latency and non-disruptive scalability of advanced networks, combined with the enhanced reliability of equipment hosted off site in environments with professional installation standards, reliable power and cooling, means the need for equipment access is rare. ‘Smart hands’ supplier services can also offer maintenance 24/7, eliminating the need for staff call-outs and travel time and costs.”

Paul Hanson, director of Birmingham-based IT consultancy eSpida, agrees here: “Depending upon what services are needed, location may or may not have an impact. For example, if you need to carry out your own patching and running updates across the estate, can this be undertaken remotely and what would be the implications if you needed to be in physical attendance? Do you purchase the smart hands service, or invest in your own to attend the premises? If you do attend for regular maintenance cycles, could you reach the centre if we get another ‘Beast from the East’? These are all important considerations for businesses to make.”

Choosing your provider

The DCA defines commercial data centres (as opposed to private facilities) as those that offer the traditional services of space and power, managed services, hosted services, and cloud-based services (although it is noted that the latter providers rent space in existing facilities operated by others). Hone says hyper-scalers, such as Amazon, Facebook, Microsoft, can also be included but they do not as yet exist in the UK. “So what we’re looking at are the colo, cloud and managed service providers. Following consolidation in the market over the last few years, we estimate that the UK currently has around 130 commercially-facing data centres, which are Tier II and Tier III. These are spread across 80 providers of which only a handful, such as Equinix and Pulsant for example, have more than four or five centres in their portfolios.”

And if there is one thing the industry agrees on it is that of these data centres and service providers, no two are the same. “There are so many different variables that it can be daunting for a network manager to know what to look for, particularly when so many companies are now born in the cloud and have never owned their own kit,” says Node4’s datacentre operations manager, Rik Williams.

According to Neil Lathwood, CTO with UKFast, there are several factors that make a great data centre but three are absolutely essential: fire suppression; a high level of organisation and cabling standards; power supply and redundancy (the higher the level of redundancy, the more reliable the centre).

Some of the other factors he says network managers should consider include power generation and certifications. “UKFast data centres are supplied by mains power but also have a backup generator on-site that provides a minimum of 48 hours power in the event of a grid outage. All good data centres will have similar measures in place.

“A good data centre will have relevant industry certifications such as ISO and government certifications. For example, for security, a good provider is ISO 27001-certified, PCI-compliant and secured to government standards.

Furthermore, Lathwood believes a provider should ideally own the land and building its data centre is in, as this makes the service and supply chain more secure over the long and short-term.

Williams reckons this last point is possibly the most important differentiator, as data centre providers that not only own the data centre, but the entire managed IT platformand network have much more control. “Having total ownership of infrastructurethat is 100 per cent designed, built, owned and managed by the data centre provider means they have absolute control over the quality and flexibility of solutions for customers, something that network managers will be particularly interested in.”

Naturally, different enterprises and organisations have different needs, so as Equinix’s Poole points out, it’s imperative that they weigh up the options and choose a service that fits them. “Different data centres suit different partners. For example, some have increased power capabilities so they are an ideal fit for companies that have high intensity data demands running around the clock. Most enterprise data centres are built to cater for power and cooling loads of around 4kW, but most Equinix facilities can handle up to around 18kW before we must start building custom solutions. Businesses should also be able to choose a solution that fits their individual needs.” 

Richard Blanford, founder and MD at Fordway, echoes this point to an extent when he says that as well as checking out technical details such as SLAs, security etc., you need to be confident that the way the provider works fits the way your organisation needs to operate. “You’re unlikely to be able to persuade larger providers to revise their processes to suit you, so if you have very specific requirements you’ll be better off talking to smaller providers. 

“It’s also important to think about the cultural fit between your organisation and a potential provider. This may seem trivial, but when you’re considering a multi-year agreement which will impact the services you offer your end users, it helps to ensure that all parties are aligned.”

Arnold advises readers to look for a number of key differentiators when choosing a data centre provider, including: connectivity (“the more carriers available, the better choice for the end user”); 100 per cent uptime; location; and security. 

He also recommends seeking out an operator that is carrier and cloud neutral: “It is important in the hybrid world that data centres offer multiple access to various cloud platforms and continue to grow the cloud ecosystem. Being placed within a data centre that is both carrier and cloud neutral will give organisations the freedom to choose which provider they would like to connect to.”

While cloud connectivity remains key, Poole points out that some customers will not need links to clouds all around the world. In such cases, he says enterprises should be able to select a cloud service that gives them connectivity in their specific market.

So when it comes to choosing individual services such as, for example, SaaS/PaaS/IaaS, etc., are there different considerations for network managers to take into account?

“A key consideration is how much work you want to do yourself,” says Blanford. “With colo, all you’re getting is racks, power and hopefully more easily provisioned network and internet connectivity – the rest is up to you. If you choose IaaS on public cloud, for example, you only get hosting, including host and hypervisor patching and proactive infrastructure security monitoring. So you’ll still need to provide patching, resilience, backup, security and application support, and maintenance inside the instance. 

“If you don’t want to do this you need to choose managed IaaS which typically includes provision of the operating system, monitoring, patching, authentication and specific security, including a dedicated firewall. Managed IaaS services effectively offer your application back to you as an SaaS.”

The pitfalls to avoid before signing on the dotted line

According to Gary Kilmister, head of data centres at Pulsant, looking at a provider that has in-house skills and a portfolio of services and partnership to offer full lifecycle solutions is becoming more prevalent. 

But here, eSpida’s Hanson warns: “It is all too easy to move to a data centre supplier with the view of reducing your internal management overhead in supporting your infrastructure and services. But be aware that you could also be depleting your internal IT skills, and so if there is any need to disengage with the data centre you must be certain that you can cope with reverting back to on-premise systems at relatively short notice.”

Kilmister continues by saying that as companies enter the IT journey, they need to ensure that their provider has the resources, utility services and space to be able to meet ongoing needs, and that these may well change especially if moving from off-premise colo to managed/cloud or hybrid solutions.

He also advises readers to look at what resources are available with the data centre provider: “Are these matched to your current and ongoing requirements? Are they able to offer consultancy/professional services to assist with the roadmap of the business? This is becoming more of a consideration moving forward especially with on-premise to off-premise to hybrid solutions.”

Blanford says it’s vital to carry out due diligence to ensure that the provider is suitably financially secure and can continue to deliver services for many years to come. He adds: “It also helps to ensure that you know the supply chain of the elements you are signing the contract for, as in many cases the data centre owner or manager is not the organisation which is selling you the contract.”

On the subject of contracts, what should network managers look for in the SLAs that they develop with their chosen data centre providers?

Volta’s Arnold says that after having identified what they want,  businesses should also consider the factors that want to avoid or be aware of. Some things to consider are the data centre’s historical performance (number of outages), if there are costly additional charges such as cross-connects, and whether there is additional capacity for growth.”

aql’s Howie supports this when he says that the SLAs should be comprehensive in their coverage of all things that may influence the business now or in the future. “The use of established frameworks procured by trusted experts will guide readers in what to look for in the SLAs. Clearly, SLAs carry a cost, so try to negotiate around the price, service level and risk you require. Service credits should be seen as sufficient motivation to secure and maintain good service, but not as a means of reducing cost through an expensive confrontational approach to suppliers.”

Scott Campbell, business development manager with Surrey-based design consultancy Comtec, adds to this by saying some very high-uptime environments will build credits into their SLA; for example, a colo provider could issue credits if power is unavailable.

He goes on to say that while creating an SLA document can take some time, having a good agreement and establishing clear lines of demarcation are crucial. “Creating an SLA is a partnership between the data centre provider and the customer. Expectations must be clearly laid out to ensure that all performance, recovery and other expectations are met. 

“Often, an SLA can be developed based on the needs of the organisation and what is being hosted within the data centre infrastructure. This means identifying key workloads, applications, servers and more. From there, an organisation can develop base service agreements for uptime, issue resolution, response time and more.”

UKFast’s Lathwood continues with the uptime theme when he says network managers must consider the business RPO (recovery point objective – how much downtime they can afford to have) and the RTO (recovery time objective – the time in which the process must be restored), and work from there.

“Some businesses can afford half a day of downtime, whereas others cannot afford 10 minutes,” he says. “Network managers must also look into when disaster recovery begins – is it when the business rings to say their website is down? Or does the provider always know if their servers are experiencing downtime? Are there staff on site to deal with the problem?”

Blanford believes that defining RPO in advance makes it much easier to negotiate with a service provider. He says a service with an SLA of 99.5 per cent measured annually (i.e. allowing up to 43 hours downtime per year) needs considerably less resilience and is therefore much less expensive to operate and support than one with an SLA of 99.95 per cent measured monthly, which allows a maximum of 21 minutes downtime per month. 

Ultimately, the type of SLA you reach with the data centre provider will depend upon what service you are taking. Blanford says: “For colo, key SLA requirements are [about] physical and remote access, and what remote hands and infrastructure monitoring capability is provided. For IaaS/PaaS/SaaS, before asking providers what SLAs they offer, you need to define the service levels you require, which sets user expectations and enables you to align service cost to the agreement.”

eSpida’s Hanson reiterates some of the points made earlier about the importance of ensuring that SLAs are future-proofed for growth and change. “Consider whether the data centre can accommodate all of your business needs beyond the next 12 to 24 months. In many cases, it’s valuable to look at how the SLA suits the company’s roadmap over the next seven years and understand what impact any significant changes, either to the business or to the data centre’s services, would have on the SLAs.”

Moving forward together

Once the SLA has been agreed and a contract is signed, what should the network manager expect from their chosen data centre provider?

“A keen focus on innovation and ensuring the site is updated in line with industry demands, especially in terms of security, power provision, cooling technologies and connectivity,” says Pulsant’s Kilmister. 

He also says it is equally important for data centre operators to form partnerships with the main manufacturers and make sure that their technology roadmaps are aligned. “[They should be] working on ever-developing network solutions and service offerings that can be explored and implemented into solutions, ensuring customers are able to utilise and maximise technology and service offerings, but supported by the data centre provider.”

Meanwhile, aql’s Howie states that as a provider it is always great to have an open and transparent relationship with clients. He adds that regular access to review the service and early insight into developments of client strategy and upcoming projects is also helpful in order to offer good forecasting, support, service delivery and agility.

But at the same time, Hanson points out that clients must recognise that any significant business changes, for example, new applications, will not just happen overnight. “By outsourcing to a data centre you will need to understand and follow the provider’s protocols, since you are one of thousands of clients for that provider. This must also be factored into the chain of events to undertake such a change when subscribing to these services.”

At the end of the day, and as Volta’s Arnold concludes, expectations from both parties need to be based on what is being offered from both sides. “Flexibility, simple but secure processes, and ease to do business with should sit at the top of a network manager’s expectation list, whilst a cohesive partnership will be an expectation of the data centre provider.”