distributing the share of costs of ownership? | SQL Server Performance Forums

SQL Server Performance Forum – Threads Archive

distributing the share of costs of ownership?

Hello DBAs! <img src=’/community/emoticons/emotion-1.gif’ alt=’:)‘ /><br /><br />We’re running two nodes in a Microsoft Cluster/SQL Server Cluster which get their disk space provided by some SAN storage device. (Also we’re still running several dedicated SQL Servers on Win NT 4, whose DBs are to be migrated to the cluster systems.) All DBs of all affiliated companies shall be held on the cluster system.<br /><br />The consolidation leads me to a new task: dividing the total cost of the system and charging the individual cost center.<br /><br />Well, I’m not an accountant. [<img src=’/community/emoticons/emotion-2.gif’ alt=’:D‘ />]<br />I hope you can give me some hints on how to handle that.<br />(edit: I am not asking for prices, but for ideas how to handle cost ditribution and how to measure consumption of resources.)<br /><br /><br />Here are my facts/costs:<br /><br />Stuff you pay once and then (hopefully) use for some years:<br /><ul><br /><li>MS SQL Server Enterprise Server: $20,000 per CPU</li><br /><li>MS Cluster Software: ?</li><br /><li>Windows 2003 OS: ?</li><br /><li>Hardware / rack space / etc</li><br /></ul><br /><br />Stuff I will be charged for, so I can pass the cost to my ‘customers’, or stuff that is billed some other way (e.g. the backup):<br /><ul><br /><li>SAN Storage*</li><br /><li>Backup to TSM</li><br /></ul><br /><br />* I’ll have to pay for the SAN space in the whole and I’ll have to distribute the cost. Dividing the expenses according to the (max) size of the DBs for DB/backup device and the (max) size of the logs will be fair enough (no I/O consideration).<br /><br />Other server resources that are limited:<br /><ul><br /><li>CPU load %</li><br /><li>RAM used %</li><br /><li>?</li><br /></ul><br />Each node has 4 CPUs and 4 GB of RAM. I have absolutely no idea how to divide this in a fair way. I mean, if the one new small database is the one that leads to install a new node/instance, I cannot charge those poor guys the whole server.<br /><br />I also do not have an idea how to measure the above ‘consumptions’ or if there are more things I should take into account.<br /><br />Thanks in advance for your ideas. <img src=’/community/emoticons/emotion-1.gif’ alt=’:)‘ />
I’m surprised they’re making you do this. Normally, the costs are put into an operations bucket and distributed out at a much higher level. If you need to do this, make sure you have all the original POs. Otherwise, the cost centers you are assigning these too are going to get very grouchy with you. I wouldn’t worry about dividing the server up by CPU and GB. If you have to divide it up at all, take the total amount of server space for each DB on the server and spread the cost that way. You can do the same thing with the SAN and other storage space. The Enterprise Edition licenses aren’t going to be $20k per CPU if you have any kind of a good agreement with MS or your vendor. You need to lower that. I would be extremely worried about allocating costs this way. Most hardware and software is going to capitalized over a three to five year period. You would allocate this down to the cost center at that capitalized level, so the individual cost centers aren’t taking the entire hit up front. Does corporate not have a set amount in your budget for infrastructure and SMO costs that falls outside the detail cost center allocation model?
[email protected] When life gives you a lemon, fire the DBA.
Well, it will not be me who writes the bills. [<img src=’/community/emoticons/emotion-2.gif’ alt=’:D‘ />] The definitive charges will be set by the companies accountants I guess. But if I went to ‘my’ accountant now (I did that) of course he asked me for suggestions how to divide the usage.<br /><br />I see the big advantage of an easy measurement taking into account only the (max) sizes of data, log and dump space used.<br /><br />On the other hand I’d like the departments to pay for rotten software. The kind of software which uses unnecessary complex queries, has a bad database design and so on. (Did you know? Oracle in the meanwhile gives you the possibility to define aliases to queries: Bad incoming query is beeing substituted by a more performant one.) In the past months I came across several products using bad design. When I ask the department which bought and uses the software to put some pressure on the programmers to enhance their products in the future releases, I find that many of them do not care at all. So we keep buying faster and faster hardware, need more instances. Then they fire the next guys in order to reduce IT costs… well, if I could tell the departments that they could save money on their cost centers they would hopefully be a bit more interested.<br /><br />Right now I have no idea on how to measure either usage. But I’d like to tell my accountant some possibilities so that he can work out a fair model.<br /><br />What does PO and SMO mean?<br /><br />Right now we have a model that there are some basic services (eg networking, ADS, DNS and so on) which’s costs are thrown in one big pot and then are divided according to the number of client computers the individual company owns. Unfortunately this leads to an unrestrained usage of resources.<br />We do not have these problems in the mainframe environment or with phones, because each boss gets billed every month. Now they are trying to get more and more services billed in a fair way. This will also keep the companies to think that they have to pay for the other ones, too.<br /><br />My company is an IT service provider exclusively for the other affiliates in the <br />There are about 20,000 employees on office duty and about 40,000 field staff working for about 15 companies that consist of some three digit number of departments and countless cost centers. A high degree of transparency could help to reduce the total costs and help the individual groups feel billed just.<br /><br />Thanks for the hint with the licenses. As I do not buy hardware or software myself I do not know the prices paid and used the fees from the Microsoft web site.
if you wanted to charge by CPU, assuming each company runs in a separate instance, you could track the CPU used by each instance, then charge by relative CPU usage, but this is not easy or fool proof, because it is a pain figuring which process belongs to each instance
PO = purchase order
SMO = Support, Maintenance, and Operations MeanOldDBA
[email protected] When life gives you a lemon, fire the DBA.
Thanks. <img src=’/community/emoticons/emotion-1.gif’ alt=’:)‘ /><br /><br />Seems like I’ll use/suggest solely the ‘used disk space approch’.