Saturday, January 29, 2011

To COLO or Not To COLO Part II

There are many valid reasons to COLO or outsource part of your data processing and storage requirements, but we are finding that there are many misconceptions about cost benefits in making a decision to COLO and that cost is typically the determining factor even though there is no real savings. What is often overlooked in evaluating data center strategy options; owning and operating a data center versus COLO space is that even if I outsource the processing and data storage I cannot outsource the need for a local network and facility support infrastructure so I still need an environmentally controlled data center with conditioned power and back-up to support my local network, WAN connectivity, security & phone systems. You can never outsource the entire facilities mission critical infrastructure can you?

For a true comparison, we need to look at hosted space for my processing & data storage while owning a small data center to support my local network, facility and safety equipment with power & bandwidth costs for both the local & COLO spaces versus owning the data center to accommodate my processing, data storage, network, facility and safety equipment with its operating costs to support everything in that single facility. With the COLO option we can reduce CAPEX from having to expand the mission critical facility by hosting the need for additional servers & data storage, but building with modular scalable data center solutions can accomplish that goal with financing and an added bonus of tax depreciation. There are cases where costs for power in a location are over .20 per kWr that hosting becomes more attractive for my processing and data storage, but it would still be less costly to relocate your processing and data storage to an area with lower utility costs and continue to own as hosting facilities always have a mark-up on at least one facet of space, power, bandwidth and support. While COLO has a lower initial CAPEX, its higher OPEX absolutely ensures the COLO model will always be more expensive in the long run. So if COLO isn’t less expensive in the long run, why are COLO facilities popping up like rabbits in springtime?

The 3 real reasons to host & the cause of the COLO boom are:
1.) We can't keep up with the expansion demand; we’re going to run out of space, power or cooling for our processing and data storage before we can alter our facilities to accommodate the growth
2.) We don't have the internal expertise to effectively plan, build, manage and operate our own data centers to the availability requirements of our businesses. I'll expand on this one a little to say that many organizations haven't effectively planned, designed or engineered their data centers in the past so they only got 3 years out of their 10 year data center plan. They built structured cabling or power infrastructure to meet their needs for bandwidth and power today so their data center quickly became outdated. For organizations like this data centers were a bad investment. Perhaps they should look to make improvements in their decision making in this area or rely more on effective consulting engineers.
3.) We don't want to be in the business of owning and operating a data center and want to focus our attention to our core business. Careful with this one as I've yet to see an organization operate a facility without a network, security system and phone system which require a small data center, of course we can outsource the operations and maintenance of a small data center’s operation but not the responsibility.

If we are doing an effective job with management and decision making, it will always be less expensive in the long run to own and operate our data centers. Stakeholders and decision makers should be more careful in their dreams of getting out of the data center business as well because nowadays it is the business. COLO facilities don't alleviate us of the responsibility for effectively protecting and managing our mission critical assets. COLO facilities can only reduce the data processing & storage components to deliver what might be unobtainable in our existing facilities or difficult to obtain in time given an aggressive IT expansion in our own facilities. Yes there are numerous ways to effectively shed some of the responsibility, with hosting effectively shedding some of the processing, data storage and DR responsibilities, but we will never get away from all of the responsibility for our data center or the ultimate responsibility.

Tuesday, January 04, 2011

Energy Efficient Ethernet?

With just about every IT and Data Center project today including an efficiency initiative, I’m wondering how many are considering Energy Efficient Ethernet in their network & cabling plant designs?

Did you know that only copper not fiber can take advantage of energy savings from WOL – Wake on LAN and LPI – Low Power Idle, which are features covered in IEEE’s ratification of 802.3az for Energy Efficient Ethernet? Did you know that only copper, not fiber, can take advantage of existing power saving features like Power Back Off in today’s Ethernet standard?

With Data Centers (1) IT watt saved typically results in (2.8) watts saved overall due to the costs and losses incurred running the support infrastructure. So should we look to design our cabling infrastructure using copper standards? With FCOE, Infiniband over Ethernet and converged networks also running VoIP and video, it is clear the long range plan should be to continue support Ethernet standards. It is also best practice to design cabling infrastructure to support 2 levels of bandwidth ahead of our requirements today. Consider that CAT7a supports 40Gbps and in any cabling or IT design for that matter we also look to be backwards compatible, which is why ISO has adopted the GG45 connector that supports CAT7a with backwards compatibility to CAT5. Of course for distances over 100M we must design with fiber, but in many cases we should be considering MPO/MTP fiber as there are fan out cables to support the LC 10Gbps connections we need today where MPO will support the 40Gbps & 100Gbps we need to future proof my cabling infrastructure. For shorter distances the choice has typically been copper distribution and with the existing power saving features like Power Back Off and those to come in IEEE 802.3az WOL – Wake On LAN and LPI – Low Power Idle, I believe the preferred cabling distribution under 100M for Ethernet will continue to be copper. What do you think?

I’m really looking forward to hearing what industry experts from Berk-Tek, JDSU, and The Siemon Company have to say about running higher than 10Gbps transmissions over twisted paid later this month, perhaps you should check it out as well?

Here's an excellent article on Energy Efficient Ethernet in CIO magazine: