A data center infrastructure management implementation that integrates cable management enhances its worth.
By Michael Phares, Cormant Inc. and David Veneski, Fluke Networks
Several new dynamics are affecting the management of enterprise data centers. One is the increased requirement for energy efficiency. Another is the need to optimize the investment in infrastructure. A third is the revolution in data center architecture brought about by virtualization and its cousin, cloud computing. Overseeing these dynamics is an innovative discipline called data center infrastructure management (DCIM). DCIM often is aimed at controlling the physical facilities: power, cooling, racks and space.
But DCIM is also a comprehensive repository for information and a logical means to manage other key pieces of data center infrastructure, including the cabling that interconnects servers, storage, switches, routers, firewalls and other infrastructure. By including structured cabling within DCIM, holistic and proactive data center infrastructure management can be achieved.
|Flowcharting milestones in the DCIM-adoption process increases the chances of the adoption's success. This flowchart template follows a path for a new data center before commissioning.|
DCIM is application software that inventories data center facility and IT assets and merges physical layer data with monitoring data collected from those assets. DCIM may include hardware, although ideally no new hardware should be required by a DCIM solution to manage existing assets.
DCIM performs the following functions.
• Tracks the entirety of a data center's infrastructure, assets and connectivity
• Provides a consolidated view of the physical and queried data while assuring data accuracy
• Allows facility managers and IT managers to collaborate and plan the future of the data center
• Provides a closed-loop process for executing planned change
DCIM removes the need to keep infrastructure records in tedious-to-update, error-prone files such as spreadsheets, drawing tools and home-grown databases on network shares. Chief information officers (CIOs) embrace DCIM for their data center operations because it is more comprehensive, more accurate and faster than logbooks, desktop software and dedicated local databases.
Many vendors offer DCIM capabilities. Manufacturers of data center hardware such as power systems, racks and cooling offer DCIM as an accessory. Some DCIM vendors are software specialists whose products are hardware-independent. System management companies also offer DCIM modules for their operations-management software. Lack of options is not a challenge for prospective DCIM users, though choosing the best option may be.
A number of points should be considered when selecting a DCIM product. The following 11 questions bring up some of the most critical of them.
1) Are the advantages in a hardware-based solution salient or do I want hardware-independence?
2) How easy is it to import inventory data into my DCIM candidates and then audit it for accuracy?
3) How should I weigh the relative importance of information management, information reporting, operations optimization and long-term planning?
4) Who in the organization will own DCIM?
5) What is the total cost of ownership? Take into the consideration employees' new process-learning time as well as updating/documenting moves/adds/changes (MACs) moving forward.
6) Are connections and links for both power and data encompassed in the DCIM system?
7) Will the selected solution capture MACs when they occur, where they occur? Or will I need to wait until I have access to a PC to enter data?
8) Is the system extendable to store the data I want to store?
9) Is DCIM a core business or a sideline product for the vendor/developer and how long has the solution been around?
10) Can I easily visualize cabinet and room layout changes in the DCIM system?
11) Can I see in one place key environmental and statistical information such as temperature, available RU space, actual and derated power, and other information?
Adopting DCIM may be complex. Flowcharting the milestones will markedly increase the chance of success.
Your DCIM flowchart should include as much detail as necessary to define, evaluate, select and install the system.
|The process described in this flowchart creates a base of knowledge from which IT management can effectively act and react when integrating cable management into DCIM.|
Who, what and why
A DCIM implementation involves several stakeholders. While the data center IT manager traditionally is tasked with improving the efficiency of data center infrastructure management and is often the first point of contact for DCIM, DCIM benefits all stakeholders by answering questions such as the following.
• What do I own?
• Where is it?
• When was it installed?
• What did I pay for it?
• What revision of software is it running?
• If I have to move an asset, where will it fit?
• Which ports connect to what?
• How are devices and ports connected?
When the features of DCIM are understood, the IT and facilities teams interact to discover how it will increase the efficiency of both teams, reduce energy costs and achieve a better Power Usage Effectiveness (PUE) ratio and/or Data Center Infrastructure Efficiency (DCIE) rating. DCIM is a "single pane of glass" that turns disparate data into focused, actionable information.
A network manager is asked to prove the network infrastructure can support the rollout of 500 virtual-machine servers. By using connectivity plus DCIM data the manager can confirm that the physical cabling infrastructure (often referred to as information technology systems or information transport systems [ITS]) will support the network traffic, then view and plan network port capacity to the individual racks.
Network managers also play a crucial role in defining the physical connections between active network devices, servers and storage. These connections are the ITS-structured copper and fiber cabling-that must support an evolving range of traffic types and speeds. Network managers have an interest in DCIM because it will give them an accurate and comprehensive catalog of network links. This is especially important as increasing bandwidth demands pose questions about the need to upgrade or replace. In addition the recently ratified TIA-606-B standard makes specific reference to electronic management of connectivity records and a good DCIM solution will provide this management.
Cabling and DCIM
Cabling is a sizable investment in data center infrastructure. It is also a highly leveraged asset in that poorly managed structured cabling inhibits flexibility while well-managed cabling facilitates growth. And one cannot forget MACs-a data center constant to which cabling must conform. Unlike power supplies, chillers, racks, servers and switches, structured cabling is a nearly invisible piece of data center infrastructure. Cabling is expected to work and it usually does. But when it doesn't, or when faced with changes, not having a complete history of the structured cabling may lead to unnecessary field queries, retesting, recabling or delays in service deployment.
As we know, data centers are subject to technology shifts. While many are independent of the infrastructure, three current shifts could affect data center cabling.
Virtualized servers-Where discrete servers used 1-Gbit/sec links to switches, dense virtualized servers make that technology passé. The impact of this trend is that top-of-rack switches are becoming more commonplace. And the cabling challenge presented is that top-of-rack switching requires higher-bit-rate uplinks with large-scale in-rack patching.
Parallel optics-Manifest for cabling-system managers in the form of the MPO connector, parallel optics will be de rigueur in data centers for storage and high-speed networking. For many 10-Gbit/sec data center fiber connections, and most or all connections higher than 10 Gbits/sec, parallel optics will be the industry-standard connectivity method. The challenge to the cabling infrastructure is to ensure the adequacy of the cabling system to support parallel-optic transmission.
Single-tier switch fabrics-Switch fabrics are the most advanced technology as they allow single-hop connectivity between any data center asset. Their 40- and 100-Gbit/sec interfaces necessitate adoption of the highest-bandwidth fiber technologies. More and more, three-tier switching is being replaced with these single-tier fabrics, and hand-in-hand with this shift is that fiber displaces copper in the data center cabling plant. This challenges those responsible for a data center's cabling system to provide a future-ready fiber plant.
Integrating cable management into DCIM will make daily work and supporting the future far more tractable. A DCIM implementation that follows the process outlined in the flowchart on page 16 enhances its worth from the IT and facilities assets to encompass a critical part of the infrastructure-the aforementioned ITS. It provides a view of limitations, potential problems and potential growth because certification yields a wealth of data.
With this type of DCIM system, one can answer questions such as:
• Will the twisted-pair link between Rack 34 and Rack 18 support 10-Gbit/sec traffic?
• Was the optical interface on Port 6 of my storage area network switch clean?
• Was the fiber link from a top-of-rack switch to the aggregation switch tested for loss?
• Are my garden-variety patch cords really suitable for virtualized application server connections?
• Will my standard fiber trunk migrate from 10- to 40-Gbits/sec? To 100-Gbits/sec?
• Can I change a cable without risking an inadvertent disconnect of a service?
Some DCIM solutions today include structured cabling under the umbrella of managed assets. Getting answers to the questions posed throughout this article can help you on your way to obtaining the right DCIM system for your needs.
Editor's note: This article is derived from a white paper entitled "The need for connectivity management as part of a DCIM environment."
Michael Phares leads marketing and business development at Cormant Inc. (cormant.com). David Veneski is director of business development at Fluke Networks (flukenetworks.com).