vir tualisation ICT
and capacity metrics right from the beginning of a new hardware refresh might seem unnecessary in the short term where there appears to be a plentiful supply of resources from newly-specified hardware. However, designing a virtual IT infrastructure by ensuring understanding of the relationship between the demands on the business and the resulting demands on IT resources is an investment that will pay off during the full lifecycle of the hardware.
Knowing the relationship between the number of users and memory consumption or I/O throughput and CPU will enable an organisation to make predictions that will prevent the undesirable result of knee- jerk hardware procurement when service delivery has deteriorated.
To prepare for this strategic approach, organisations need to ask themselves the following questions during the design and support of virtual IT infrastructure:
1. How will I know when to grow the server estate (scale-up )or to buy additional servers (scale out)?
The CIO needs to be aware of the options for growth for your particular application. You can add more power to your virtualised infrastructure but a single virtual machine (VM) will be limited to the power of a single piece of hardware. Thinking ahead to the possibility of your application demands being many times more than your existing demands will force you to consider the paths you have open for these eventualities.
2. How will I justify this to the board? What data will I need and how will I get it?
The requirement for increased horsepower may be obvious to IT, but unless the business is kept informed of any changes and therefore able to budget well ahead of new demands, senior management will often be reluctant to sign off any upgrades. The business needs to understand IT costs as part of the complete business model, as these costs may need to impact prices down-stream. A continuous approach to capacity planning that keeps the business informed will reduce any unplanned and unbudgeted demands that will be poorly received by the decision makers.
3. How will I know which workload is consuming the most resources and will I be able to keep track of this before it becomes a problem? In a virtualised world you may well have multiple business areas and multiple, different applications sharing resources. The ability to attribute consumption of resources to these units will help the decision making process in the future. You might not want to employ the same levels of service (and therefore cost) for all areas, so whilst it makes sense to consolidate where possible, some services might not make good neighbours on the same hardware.
For example, development systems with hungry consumption of resources might not be best placed on expensive hardware designed for the resilience of production systems.
4. How will I track the flow of workload across virtual machines? With multi-tiered applications, sometimes the resource consumption moves from one tier to another in a serial fashion, which lends itself to co-hosting where the peaks of one meets the lows of the other. Sometimes the planning of workloads according to daily, weekly or monthly patterns might present opportunities for resource sharing.
Some virtualisation platforms have the ability to balance load between nodes in a cluster according to demand, but there will be limitations to how quickly they can react to demand. Well-informed, proactive placement resulting from trending analysis will provide better results.
34
www.dcsuk.info I May 2012 The requirement for increased
horsepower may be obvious to IT, but unless the business is kept informed
of any changes and therefore able to budget well ahead of new demands, senior management will often be reluctant to sign off any upgrades.
The business needs to understand IT costs as part of the complete business model, as these costs may need to impact prices down-stream
5. How will I know when the underlying LAN, SAN or storage array is near saturation in performance and capacity? Vendors will give you a performance guide to the maximum I/O capacity of your disk array in terms of I/O’s per second (IOPS) and (MB/s throughput) but the exact capacity will be dependent on the number and layout of disks in your system and the specific profile of your I/O. The I/O profile will vary between workloads in terms of random versus sequential and read versus write. Different mixes of these profiles will provide a challenge for predicting the saturation point for shared storage. Monitoring disk based metrics alongside server and application metrics will build a full picture from which predictions can be made.
6. How will I make the business areas aware of their resource consumption? How do I share out the bill for consolidated hardware purchase?
Virtualisation and consolidation plans are often not implemented fully or effectively because of the inability for companies to share the cost across departments or projects. Often the only way to get sign off for such purchases is on the back of the next business initiative! IT departments need to communicate with business units early to get buy into the shared model. This will prove much easier through more accurate means of apportioning cost. Using metrics from monitoring presents a good way of achieving this.
All too often, the responsibility for ensuring effective capacity availability and performance management falls through the cracks between different departments and can become neglected. It is possible to achieve considerable savings on unnecessary hardware, by setting a clear policy on capacity planning from the outset and ensuring better communication between IT design consultants and the support team.
Using ITILv3 principles is an effective, risk-managed way to promote long-term planning in your IT and virtualisation strategies, anticipate demand and improve the quality of dialogue between IT and the business. You will be able to maximise the utilisation of your existing systems or quickly be able to justify new spend. It is crucial to ask: How do I solve my problem today AND make sure I’m in a position to keep my IT infrastructure optimal in five plus years’ time?
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52