> > 2009/12/4 Smruti <smrutiman...@gmail.com>: > > Hi, > > > > We have a few hundred physical *nix servers and thinking of virtualizing > > them. So, could you please tell how to go about virtulizating them. > > > > The first step would be obviously to calculate the actual utilization of > > hardware for each server. So, what are the things to consider while > > calculating the utilization of a server. > > > > Thanks & Regards, > > Smruti >
On Fri, Dec 4, 2009 at 7:02 AM, Rahul Bhargava <rhlbh...@gmail.com> wrote: > Hi Smruti, > > It is possible to dynamically resize both the processing and storage > clusters, so it isn't necessary to start with estimates of utilization > necessarily. Though, where the application requires a lot of storage > or quick disk access, this will affect the types of disks used and > protocols, that is, whether you require a SAN using iSCSI or Fibre > Channel with SCSI disks at what rpm or something faster. Hi Rahul, First of all thanks a lot for the detailed mail. It will help me a lot. But what I am concerned about right now, is to find out the actual utilization of each physical server. This is how I can determine how many applications I can club into what kind of machines. Some applications are home made and some are like SAP and Oracle DBs. I am planning to calculate the utilization of each server over a week and month's time. Also want to calculate the peak load it gets and when does it get that. I am looking at resources such as CPU, Physical M/m, swap space, NIC and FC Card utilizations. Please add whatever else comes to your mind that I should consider too. Assign some reliable high-end or dedicated hardware nodes as cluster > fences. Fencing protects your information in the case of a node > failure in the cluster. > Yes, every global zone will be clustered. Determine if you'd like Network Attached Storage, for network > available storage and the filesystem, or a SAN, if you'd like to see > block devices and let the clients handle the filesystem. Source some > dedicated hardware or make the appropriate partitions on each node and > make these network visible. > About the storage, all the user and application data is already in NAS and SAN. So, I won't be much worried about that. RedHat Enterprise Linux has some software for managing clusters, > Conga, with support for fencing and dynamic resizing, for example. > Meet luci and rici. > And most of the machines are Solaris, so the chosen way to go is Zones. Thanks again for the quick response. You guys rock!! Regards, Smruti -- I took the red pill. _______________________________________________ ilugd mailinglist -- ilugd@lists.linux-delhi.org http://frodo.hserus.net/mailman/listinfo/ilugd Archives at: http://news.gmane.org/gmane.user-groups.linux.delhi http://www.mail-archive.com/ilugd@lists.linux-delhi.org/