Shaved Kenneth (Kenneth Brill)

Shaved Kenneth (Kenneth Brill)

The person who thought up multi-layer system of assessment of DPC

What pushed you to development of this multi-layer system?

KB: Everything began about 18 years ago. We worked as consultants in United Postal Service. Till 1989 UPS was delivery company of sendings. Calculations were only support function.
However everything changed with express delivery introduction. Suddenly calculations became crucial for a company performance. The new recently constructed data center of UPS was outdated the next day. It was intended for accomplishment of support functions for the mail company and did not conform to requirements of air mail at all.
It is important to remember that the worst time of failure at aviadelivery – night. At night all IT infrastructure should work uninterruptedly without fail. Almost on all airlines there are mainframes that the company did not suffer serious losses because of any failures.
If to consider UPS, then night failure of IT will lead to the fact that two million sendings will not be delivered next day. Taking into account that UPS pays compensation in $10 for each sending which is not delivered next day, that the potential damage from each hardware failure is 20 million dollars. Couple of such failures – and on the missed money it would be possible to construct new data center.
Let's return to the end of the 80th when business was run absolutely in a different way. All sector of IT came down only to service. It is simply amazing that only twenty years ago the situation was thus. We were employed as consultants. We did the same, as all others in the analysis of a problem – tried to look at it from the new party. As a result we developed the standard which began to be applied in all industry.
Our work led to the fact that energy consumption and cooling stopped being the only criterion in the industry.

What forced you to publish the standard and to protect it copyright?

KB: We published the first standard in 1996. The representative of Hewlett Packard was one of our clients. He should have explained well to the management why it needs 50 million dollars on construction of data center. Here that really interestingly in this standard. It is really aimed at the user. The majority of standards is intended for those who sell something. The same standard is intended for buyers.

How often the standard is used?

KB: This standard is used and apply very many people. Any own declaration of conformity to any level should be carefully checked. All data-centers certified by us are listed on the website of institute. We developed process of assessment of to what level there corresponds the data center. In the analysis of the data centers which declared any level we often find out that their actual level is one lower if not on two positions. It can strongly upset the owner who spent considerable money, but did not receive due result.
Often the situation could be corrected, having got advice at the initial stages – usually as a result cheaper to make everything correctly.

Will not you give an example of a widespread error?

KB: In data center 3 levels the possibility of service of any element of mechanical or electric infrastructure without shutdown of IT should be implemented. It is necessary to have an opportunity to point to any component of data center and to tell that further work is possible also without it – that's all. However in many data centers ranking themselves as the 3rd level it is necessary to disconnect IT infrastructure for maintenance. It means that the most part of time this equipment is not serviced. At its failure effects can be catastrophic. Or other case – recently I investigated one data center. The power desirable to the operator was 10 MW. An object is implemented in such a way that at simultaneous service they can use only 6 MW from 10 – i.e. costs for excess 4 MW were vain. Now I am not sure even that most of clients can just detect a similar problem. The personnel of our company investigate 50-100 proyet a year.

Why other data centers do not address Uptime Institute for official certification? It is expensive?

KB: The price is less than a percent from project cost. We do not earn money, and we perform certification.
It should be noted that extremely few enterprises really need data center of the maximum fourth level assuming absolute fault tolerance. We do not believe even that effects of failure of most data centers will justify compliance to the 4th level. Losses should make the amount in the amount of a quarterly profit.
However, many companies need the 3rd level. Google uses 1 level – and this absolutely right decision. Our own small server corresponds to the first level, and to us it is enough.
We are skeptical about the level necessary for you. If you aimed at the 4th level, then you should understand that it implies. Will be cheaper to pay for it the small amount.

You certify only an object? Or workflow too?

KB: 70% of failures fall not on the equipment, and on a human factor. If on an object there is no good management, it is fraught with troubles. One of the widespread mistakes which are most made by the person in data centers 4 levels – an error in connection of two sets of spare equipment. So server A is connected to system B, but not to A. Or such situation – the tenant works in data center, without following rules, and inadvertently switches-off the equipment.

Whether the main data centers of Australia are certified?

KB: Australia, and in general the Pacific region, are very perspective. We have flowing projects here.

Do you have personnel in this region which can inspect and advise local data centers?

KB: We are a part of a system of certification of LEED (Leadership in Energy Efficient Design is Leadership in energy efficient design), which name consultants can use. The consultant becomes the certified designer of levels according to a certain standard. Within the last 12 months we certified 184 engineers, and we are going to give the following course on Taiwan. In Asia there is only one certified specialist on Taiwan so next week we will conduct the course there. We expect to receive from this 15-20 specialists.

Set of the Australian data centers are container...

KB: Sorry, at once I will interrupt you. Kontenyerny data center – initially incorrect term. In fact, it is container server. The data center consists of two-three different systems – mechanical and electric which ensure its functioning, the computer floor and the cable infrastructure transferring information. On server 20 and less percent from the total cost of data center are necessary. 80% leave on another.
And so, container server is inexpensive. However for its work the mechanical and electric systems are necessary. Minimization of volume of server will not save a lot of money.

What will you tell in Sydney about?

KB: I will speak about reduction of the advantages caused by Moore's law. According to Moore's law, everything should become cheaper constantly, however at the same time is not considered that energy consumption does not fall with the same speed with which the proizvoditelsnost grows. We spend one energy more and more - it is a basic reason of strong growth of cost of data centers. The reason of growth of cost IT infrastruktry consists in the cost of data centers.

What main costs? Operating expenses like payment of the electric power or a capital expenditure on an object?

KB: Pay attention to operating expenses very important, but not so, as for capital costs. The space for server concerns them, it is resistant for the server. The server rack costs an average of $1300 - $2000 for the server. Besides, the building and the earth under an object belong to a capital expenditure, supply systems and coolings is $16.000 for the server – i.e. are 8 times more. The electric power for a power supply of the server will cost $500 a year. Thus, capital costs for a rack for $16.000 at life in 15 years will make $1000 a year. It is the cost of utilities. It is difficult to determine average total cost as it depends on very many factors.

So you will speak about efficiency of data centers?

KB: I am going to tell about a system from five steps which will save $120.000 for a rack within four years. I call it the strategy of NNDC – No New Data Centre (is not present to new data centers). It means lack of need for construction of new data centers. And these steps are quite painless and do not assume big costs.

Why it is not used already now?

KB: The efficiency has not enough supporters. At an average term of work of the director of IT in 30 months, it simply has no time for creation of really effective IT infrastructure as it takes 3-4 years. Increase in efficiency due to cost reduction can even upset owners.
It is quite real to cut energy consumption of each component of data center by half. However methods can be painful.
Painful, but simple. So simple that you are surprised much, having heard them.

Will not you give one example?

KB: OK, only one. For example, to switch-off idle servers. Some of them assume to consume energy empty. Be more attentive, and you will be able to reduce consumption by 10-30%. Floor server is covered with gold nuggets – it is necessary to have desire to bend down and lift them only.

Source: ^[1]