No matter where we are with the datacenter transformation and how they are built now, one thing still remains almost the same; and that is scaling. I do not know why companies are not concentrating on this since this is not that hard to do. I wish i had the resources or atleast the time to do it but i don't. Lemme give you an example.
Let us say there is a cisco ucs and some 5 blades with vmware on it. Client now want to scale it to 50 more blades. It usually takes 3 guys (storage, network/compute, virtualization/windows) engineers to get this done over a long period of time. weeks or may be a month too where as this stuff shouldnt take more than a day if automated. You can do most of the stuff with cisco automation tool powertool that you would do via GUI. So these are the break down of the tasks that I would do if I ever try to do this.
1) an auto loop script detects the new blade(s) and compares the firmware with the last registered (In our case 5th blade) working blade which is in production and upgrades all the new ones to keep them in sync with the old ones.
2) It takes the service profile of the 5th blade, creates a copy of it, does a +1 to all the variables including the name of the service profile and applies it to the new blade. Repeats the same till all the blades are applied with a service profile.
3) Now Once all these blades are ready the powertool triggers another VMware auto deployment script (appliance or anything else) which installs the esxi on all these blades and does a +1 to the variables like ip address of the vmk0.
4) now the vmware auto loop script compares the standard and distributed switchs for this hosts and creates them accordingly including portgroups. If the portgroups are vmkernel portgroups then it will do a +1 to the ip address of the previous host for that portgroup and applies it to the next host in series. This is not very hard to do but just a matter of time.
5) Now the vmware auto loop script will compares the advanced settings on the existing hosts and applies the same to all the new hosts.
This can easily be done in just one regular shift if it is automated the way it has been laid out here and the tools needed for this are very much existing.
Let us say there is a cisco ucs and some 5 blades with vmware on it. Client now want to scale it to 50 more blades. It usually takes 3 guys (storage, network/compute, virtualization/windows) engineers to get this done over a long period of time. weeks or may be a month too where as this stuff shouldnt take more than a day if automated. You can do most of the stuff with cisco automation tool powertool that you would do via GUI. So these are the break down of the tasks that I would do if I ever try to do this.
1) an auto loop script detects the new blade(s) and compares the firmware with the last registered (In our case 5th blade) working blade which is in production and upgrades all the new ones to keep them in sync with the old ones.
2) It takes the service profile of the 5th blade, creates a copy of it, does a +1 to all the variables including the name of the service profile and applies it to the new blade. Repeats the same till all the blades are applied with a service profile.
3) Now Once all these blades are ready the powertool triggers another VMware auto deployment script (appliance or anything else) which installs the esxi on all these blades and does a +1 to the variables like ip address of the vmk0.
4) now the vmware auto loop script compares the standard and distributed switchs for this hosts and creates them accordingly including portgroups. If the portgroups are vmkernel portgroups then it will do a +1 to the ip address of the previous host for that portgroup and applies it to the next host in series. This is not very hard to do but just a matter of time.
5) Now the vmware auto loop script will compares the advanced settings on the existing hosts and applies the same to all the new hosts.
This can easily be done in just one regular shift if it is automated the way it has been laid out here and the tools needed for this are very much existing.
No comments:
Post a Comment