As organizations wrestle with the challenges of data center expansion, IT executives are increasingly turning to outsourcing options for expansion and cost control. There are
Q. What are the major considerations involved in selecting a data center outsourcing provider? Which considerations are often ignored or overlooked?
Pete Sclafani: In my opinion, the biggest consideration when selecting a data center provider has to be connectivity. If you have limited options or a subpar network infrastructure that requires more effort on your part, it will impact your costs significantly – not just during an outage. This is easily addressed when examining the scope of your outsourcing project, when balancing risk with budget and performance. After the network, a consideration of the facility itself should factor in heavily since you need to feel comfortable on the physical infrastructure being up to par. Here is a summary of the issues often overlooked:
Network performance. It is not only the quantity of the network options available, but also their quality. You probably have expectations regarding network performance, but I highly recommend testing the network from where actual users will be located. This way, you get an IP address of the provider in the facility, and are able to measure latency from your various endpoints. It's not a replacement for testing your application, but it’s a benchmark that can help uncover potential issues in advance.
Facility maintenance records. The physical side of a data center (e.g., cooling, power, electrical and so on) ultimately keeps your applications running. If the facility is not being maintained adequately or the owners are deferring maintenance, you need to know before you sign a contract. I recall one client that I had assisted in data center selection; once we inspected the maintenance records of a prospective data center, we were able to rule them out very quickly because they had deferred maintenance on their UPS units for more than nine months. In an outage, the UPS would not have provided the runtime they were expecting based on the load. The third-party vendor servicing the facility did its job and documented everything, but the operator simply didn't want to spend the money on replacing the worn batteries. Review maintenance records, identify who the data center operator is working with and talk to some other customers in the facility, if possible, to see if their experience mirrors the facilities uptime record.
Support or remote hands. Having someone on site 24/7 is a requirement for your data center, and you don't want any surprises when trouble strikes. It makes sense to give a random late night call to the support line and see how what they pitch differs from what they actually deliver. We have run into a multitude of approaches to support – sometimes it is "24/7 on-call," other times it's "24/7 onsite" and sometimes it’s a combination of the two. The point is, you want to verify that support is available as expected.
Disaster recovery and business continuity planning. While disaster is not guaranteed, this is another key area that impacts your business goals. Understand how the data center operator plans for the unexpected. Some questions to ask might include; What happens when there is an extended outage and generators are running out of fuel? Who do they have fuel supply contracts with? When it comes to power delivery, how have they addressed any single points of failure? Is A+B power really existent or just marketing jargon on their collateral?
Look ahead to IPv6. Finally, if your data center provider is also providing connectivity, make sure they support IPv6. That will make life easier for you down the road as you look at dual stacking and rolling out IPv6 internally.
Q. What should prospective clients expect for pricing and terms? Where are the pricing “gotchas” that clients should beware of?
Pete Sclafani: Pricing is always interesting and there are different pricing considerations to be aware of:
Setup fees might be flexible. First, setup fees are "real." Cabinets have to be built, electrical whips have to be installed and cross connects have to be provisioned. However, if the operator already has cabinets previously vacated and "waiting," then you might be able to negotiate these fees down. Expect a bit more leeway if you’re looking at a contract longer than one year. Otherwise, the setup fees may be tough to negotiate.
Look for some long-term incentives. When weighing month-to-month versus multi-year contracts, you can expect some incentive pricing for the longer term contracts, but don't expect too much of a discount and don't expect to lock in rates for longer than three years (this varies on footprint size and provider). The reason is that electrical prices continue to go up and it’s squeezing the providers’ margins. If you are on a yearly contract, be aware that regular price increases (as stated in your contract) are very real and should not come as a surprise.
Watch the way that you’re billed for power consumption. Does the provider charge you per circuit (the total amount of power that you can safely use), or is it metered so that you pay for only the power you actually use? Depending on the answer, you may have a different hardware strategy. You will most likely be charged every month per cabinet and per electrical circuit (or per kilowatt hour). Depending on your hardware environment, you can begin to test the most cost effective way to lower your operating costs. As an example, by using 30-amp circuits instead of 20-amp circuits, you can reduce the number of cabinets you need and use your space more effectively.
Verify any power per space rating. Be aware that data center operators may call out a "watts per square foot" metric, but it is up to you to identify how that space is actually calculated. So before you assume an equipment density, make sure you get the appropriate circuit counts and get the details on the level of density they can support.
Look at the downtime credits. It’s important for clients to realize that there are aspects of a third-party data center provider that cannot be controlled, so knowing the implications of downtime on your service-level agreement can be critical. In many cases, you may get little (if any) compensation resulting from a service outage, so understand the recourse you have when something goes wrong. This can be as simple as getting credit for the period of the outage. In other cases, you may be allowed to extract yourself from your contract. Watch the fine print closely.
Q. What kind of support should a prospective client expect? What kind of provider behaviors might suggest support problems?
Pete Sclafani: This should be a 24/7 shop. Depending on your uptime requirements, you may require someone on-site versus on-call. One of the biggest issues with support is delineating who is responsible for what. You will hear the term "managed services" and the provider may claim to do it all. A successful vendor relationship will involve your team aligning these services accordingly with your business and drawing lines where needed. It's important to do this prior to the 3 a.m. call when something is down.
Q. Where are the leverage points for prospective client businesses? What can clients do to negotiate the best terms with a provider?
Pete Sclafani: There can be many different leverage points depending on the provider and the type of service that you need, but there are some typical points to consider. Start with the contract term; a month-to-month agreement will almost always be more expensive than a fixed contract term, so be aware of this and plan accordingly. Multiple service bundles are another factor. The more services you engage, the higher discounts may be available. Beware, there are limits and additional services may complicate your relationship with the provider and create additional vendor dependencies that you might not have planned on.
Power consumption can be a powerful discount incentive. Some data center operators are running into power limits on the facility (also impacting cooling capacity) rather than physical space. If the operator provides incentives on lowering your power footprint, you may have the ammo you need to get some more power-efficient hardware.
There are also some location considerations. For example, location isn’t as big as it used to be, but real estate prices are an obvious factor. A data center provider with cheap costs per square foot and inexpensive power will probably reflect those lower operating costs in their pricing. However, if you want to be centrally located in the biggest carrier hotel, don't expect bargains. Look at the existing build out and see what cabinets or locations have recently been vacated. If timing is right, you may be able to reuse existing infrastructure if it meets your needs. The vendor doesn't have to rework the setup and you can move in sooner.
Finally, think about the move in timing and staging. If you’re dealing with a DCaaS provider, consider the logistics of when and how to move equipment. Depending on your footprint and the amount of equipment that is involved in the move, you may be able to access a staging area so you can get your hardware set up prior to moving in.
Q. How should a prospective client “move in” with a data center provider? For example, is a phased approach best? What are the most effective strategies?
Pete Sclafani: Coordinate a move in and migration timeline with the provider. People tend to forget that the clock starts ticking once you sign the contract. You can expect to get billed for setup fees immediately, and start paying your monthly charge as soon as your footprint is "handed over." The space will be ready for you whether you show up or not.
Depending on the role of the data center, it is definitely appropriate to plan out an implementation schedule and identify key personnel. While the provider may have done this multiple times, they do not know your application and the dependencies associated with it.
One strategy we have used for data center migrations is leveraging your disaster recovery strategy to "span" facilities between your current data center and the provider’s data center. This allows you to verify that everything is working normally before cutting over. It also gives your disaster recovery plan a refresh and helps update skills and application dependencies that you wouldn't be aware of without having to setup/migrate to a new environment.
This was first published in November 2011