IT Infrastructure Services Ltd Holborn Gate, 330 High Holborn, London, WC1V 7QT Telephone: +44 (0)20 7849 6848 Fax: +44 (0)20 7203 6701 Email: info@itisltd.co.uk www.itisltd.com Introducing Computational Fluid Dynamics Virtual Facility 6SigmaDC 2009 IT Infrastructure Services Ltd
The Concept 6SigmaDC software was developed by Future Facilities to build and create a virtual representation of the Data Centre facility and the IT infrastructure deployed within them. These simulation techniques replace inadequate estimations and rules of thumb with scientific precision in order to manage and predict the performance of the data room facility. ITIS s aim is to reduce thermally induced equipment downtimes and poor manageability, and improve the performance and operational efficiency of the data centre. The key to our methodology is the Virtual Facility (VF). The VF provides a full 3 dimensional, mathematical representation of any data centre at any point in time. The custom built VF provides an accurate and holistic perspective, allowing the operator to track past events, investigate the present status and predict the impact of future changes, with no risk to the real facility. 2009 IT Infrastructure Services Ltd 2
The increasing power demands of IT components, in particular blade architecture; affects the data centre environment and the Mechanical and Electrical infrastructure in place to support the facility. Due to a lack of effective communication between IT suppliers and the Building Services industry, this produces problems for Facility Managers. Data Room Facility Equipment Manufacturer Communications Breakdown Facilities Manager Our aim is to provide a communication path between the associated disciplines to enhance the understanding and maximise the performance of a data room facility. 2009 IT Infrastructure Services Ltd 3
Data Centre Design Whether you are creating a design concept or simulating an existing facility, 6SigmaDC software uses the same design concepts to create the VF. The VF contains the entire room layout and configuration; this includes all items of equipment and infrastructure deployed within the facility. These items are provisioned for an entire library of generic and specific models The audit of a data room facility includes every detail within the room; structural dimensions, ACU s, PDU s, cabinets, cabling routes, lighting detail, fire suppression, floor void and even floor grilles, down to the IT equipment deployed within the cabinet is documented and imported into the VF. The audit detail is critical to the accuracy of the VF to ensure that we create as accurate a representation as possible. 2009 IT Infrastructure Services Ltd 4
Construct the model The four stages to construct the VF: Stage 1 Create the basic room infrastructure Room size, floor void depth, columns, beams... Stage 2 Introduce cabinets and testing layout Stage 3 Add ACU s, PDU s, floor grilles, fire suppression and cabling Stage 4 Introduce equipment and produce results 2009 IT Infrastructure Services Ltd 5
Simulation results Once the VF has been created we can examine the model and highlight areas of concern. Room temperature reference points are taken at 0.2m, 1.0m and 2.0m above the floor. The following model illustrates jets of hot exhaust are clearly visible entering neighbouring cabinets which will be causing thermal issues for the equipment housed within. This measurement was produce at 1.0m and shows air exhaust at temperatures exceeding 30 c 2009 IT Infrastructure Services Ltd 6
The average inlet air temperature is another characteristic that is interrogated. This plot shows the average inlet air temperature for the equipment in each cabinet. In this instance many cabinets are predicted to have equipment breathing in air over 35 C. For card chassis and Cisco equipment this may not be a problem but for typical server technology this could cause thermal shutdown. Each cabinet that has a potential problem is investigated and the thermal characteristics of the cabinet can be fully analysed. This plot shows that internal issues of the cabinets are increasing the inlet temperatures of the equipment. This cooling air is supplied to the equipment at low level and then exhausted into the cabinet. In this scenario the rear door is solid; the hot air has no way to escape so rises through the cabinet supplying equipment in higher U slots. The temperature of the air steadily rises as it moves through the equipment, causing the high inlet temperatures. 2009 IT Infrastructure Services Ltd 7
The cabinet may not only have a thermal issue but it may have air exhaust issues as well This plot shows that the problem is worsened by the complex air paths involved to cool the equipment deployed within the cabinet. This scenario shows a Cisco 6509 reliant on front to back and side to side cooling. When placed in cabinets with no blanking or perforated doors, the exhaust air re-circulates internally raising the inlet air temperature to the equipment. Cabinets without side panels also allow air to move between cabinets adding to the problems. 2009 IT Infrastructure Services Ltd 8
The data centre airflow characteristic from the AHU are evaluated. The following plot shows a typical air path from the perimeter CRAC units into the main area of the facility. This shows that air is heating up as it travels into the centre of the room. Air is supplied to the centre cabinets are 5 C warmer then the edge cabinets. There are areas of the data hall receiving low levels of cold air from the CRAC supplies. 2009 IT Infrastructure Services Ltd 9
Making changes at the cabinet level and room level can provide significant overall energy savings. By ensuring air takes the most efficient path from CRAC supply to equipment and back to CRAC return. By adhering to best practise the following savings could be made:- > Lower airflow requirement thus reduced fan power > Higher CRAC supply temperatures to reduce chilled water system power > Higher chilled water set points to reduce chilled water system power 2009 IT Infrastructure Services Ltd 10
Comparison of Virtual Facility with Real Facility During the survey physical measurements of temperature and airflows are taken to allow comparison between the real and virtual facility. I-buttons are placed on the inlets and outlets of all the CRAC units to measure the air temperature entering and leaving the units. Temperatures within the room are also monitored using I-buttons. Airflow measurements are made at the return grilles on the CRAC and these measured flow rate are used in the simulation. Using a balometer, temperatures and flow rate measurements through each floor grilles are recorded. Pressure measurements are also taken in the floor void. 2009 IT Infrastructure Services Ltd 11
CRAC performance is monitored A summary of the measured and simulated CRAC return and supply Temperatures are produced. The match between the simulated and measured results in the example table is very good. Only two units are supplying outside their range. Since we use an average return temperature to control the units in simulation we are likely to see some variation against the real units which will be taking a single sensor value. In both these cases if we reduce the return temperature by 1.5 C we will be inside both the supply and return ranges for all CRAC units. 2009 IT Infrastructure Services Ltd 12
Floor Void Pressure is Monitored The pressure difference across a floor is measured in various areas of the facility to ensure a good match between the Virtual Facility and the real facility. 2009 IT Infrastructure Services Ltd 13
The Service Our services are provided by a dedicated team of consultant engineers, who will undertake surveying, data collection, equipment inventory and modelling in order to ensure your virtual facility matches your requirements as exactly as possible. The service is provided over two phases, the audit phase and the reporting phase. Audit Phase The audit information required for the thermal model is critical to its accuracy in order to control cost and keep them as low as possible. ITIS will audit all cabinets within the data hall facility, produce an equipment inventory and input into a CSV format which will be imported into the thermal model. A snap shot of the load within the cabinets will be documented and imported into the thermal model; if this information is not available a clamp test of all the cabinets will be undertaken. It will also form part of the review of the load balancing across the phases. Import all essential data into the 6SigmaDC software database and construct a thermal model and present findings. All of these works will be documented and managed by the ITIS consultant management team Report Phase To evaluate the thermal model and produce a Recommendation document for the data hall facility. This report will contain the following information:- > CFD virtual facility conclusions > Evaluate any changes that need to be made to the physical cabinet infrastructure, produce report detailing critical timeline, resource and budgetary requirements > Evaluate any server/infrastructure restack changes that need to made, produce report detailing critical timeline, implementation plan, resource and budgetary requirements > Evaluate and assess the Mechanical and Electrical services within the data hall facility, produce report on findings detailing estimated energy consumption savings, potential payback in terms of low years and an estimate PUE rating > Recommend thermal model reiterations Specialist Areas Covered The assessment will be carried out on the Mechanical and Electrical services as well as the IT Infrastructure services. Areas that we will focus on to achieve improved efficiency are as follows: > Cabinet/rack data room layout > Cabinet internal configuration air flow efficiency > Under floor airflow efficiency > Server deployment strategy > Chilled water temperatures > Free cooling > M&E infrastructure redundancy > Ambient data hall temperatures > Evaluate Variable Speed Drive efficiency > CRAC control strategy 2009 IT Infrastructure Services Ltd 14