Thin provisioning is becoming increasingly popular among data center administrators, but still many of us are afraid of using it in our production environments, fearing performance-related issues. However, if planned judiciously, a better performance can be extracted from thin provisioning rather than by provisioning thick storage. In this tip, we highlight a few
• Thin provisioning was initially introduced at the storage level. However, recently VMware and Hyper V introduced thin provisioning at the Hypervisor level. Storage virtual provisioning gives an impression to the server that an ‘x’ tera byte (TB) of storage has been allocated, when actually it is less than ‘x’ TB. Thus, when the storage administrator sets threshold storage, it can be extended well in advance, in case he utilizes it. Thin disks don’t give any illusion to the server i.e. the amount of storage visible to the HyperVisor Server is equal to the actual storage allocated. Hence, as a thin provisioning best practice, we first need to zero down on the level of thin provision to implement and for what applications.
• In virtual server environments, system administrators can provision thin as well as thick disks. Space required for a thin-provisioned virtual disk is allocated and zeroed upon demand, as opposed to upon creation in thick disks. There is a higher I/O penalty during the first write to an unwritten file block, but the same performance as an eager-zeroed thick disk on subsequent writes. VMware thin disks should be allocated to test servers as there is a possibility of a sudden data burst from an application. This could lead to an application outage, which is unacceptable in production environments, as the storage administrator will not get enough time to provision extra storage.
• As a thin provisioning best practice, a dedicated DataStore should be used for thin disks so that the performance hungry applications are not affected, as the calculation of metadata in thin disks adds an overhead.
• In case of storage virtual provisioning, we can get a better performance as the storage pool will comprise ‘n’ number of disks, which implies more disk spindles to serve the I/O requests. Therefore, storage from thin pools can be allocated to the Hypervisor, with threshold ratios for the alerts set wisely so that data bursts can be served and storage can be extended.
• For guest virtual machines installed with Oracle or any application where RDM devices need to be presented, thin provisioned disks for test servers can be used as a thin provisioning best practice.
• For laboratories and testing environments, thin provisioned disks can be used to utilize the storage efficiently.
• As a thin provisioning best practice, storage thin provisioning can be used for applications where chances of data burst are minimal, therefore optimally utilizing the storage.
Implementing thin provisioning in the virtual server environments necessitates categorizing the servers and applications intelligently, according to their data usage trends and criticality.
About the author: Anuj Sharma is an EMC Certified and NetApp accredited professional. Sharma has experience in handling implementation projects related to SAN, NAS and BURA. One of his articles was published globally by EMC, and titled the Best of EMC Networker during last year's EMC World held at Orlando, US.
This was first published in February 2011