What'll we do tonight, Kieran? Same thing we do every night, Tintri....
... try to take over your public cloudy worlds (one array at a time)
Newly-IPO’d Tintri has made its storage array bigger, better and faster by extending it out to the AWS and IBM clouds, speeding specifically Storage vMotion, upping its scalability limit and predicting future capacity and compute needs.
It says its on-premises arrays meet autonomous services, rapid elasticity or expansion, automation, and resource pooling cloud service definitions listed by NIST so it is, effectively, a public cloud-like service on-premises.
Having its cloud cake and eating it, the company is introducing a Cloud Connect product to send array snapshot data at the VM or container level to either AWS, via S3, or the IBM COS public clouds. Such a cloud backup can be restored to any Tintri VMstore array.
It says this makes it a hybrid cloud storage provider and the public clouds can be used for data protection, archiving or disaster recovery. There is a claimed single click restore capability.
CTO and co-founder Kieran Harty said: “The logical space in the public cloud is 10 times or more of what you actually consume there.”
Alongside this cloud capability it is adding hard multi-tenant quotas for better quality of service provision. This includes tenant separation, data encryption, service assurance and per-VM analytics for chargeback.
The array system can now scale out to support 480,000 VMs or containers and 40PB of capacity managed though its Global Center 3.6 console looking after 64 Tintri storage systems.
It has replication technology for array-to-array data transfer that can speed up a VMware Storage vMotion exercise, or Hyper-V equivalent, by a claimed 10 to 30 times compared to other arrays, because it moves deduplicated and compressed array-taken snapshots with no impact on the accessing (host) servers’ CPUs. The data is encrypted as well. A hypervisor snapshot is taken at VMware level and that is moved across as well.
Harty said a large vMotion exercise that could take an hour or more before can now complete in minutes. Customers can move workloads with per-VM snapshots, QoS and policies intact.
Previously moving a 1TB VM with vMotion involved a host CPU pumping all the data, even the zero-provisioned bytes, to the target system; it is very inefficient. Doing it at the Tintri array level is much faster.
Harty claims: “We just move whatever has been written and compress and dedupe it.”
Tintri is planning to update its predictive analytics to say what anticipated compute as well as storage needs will be required in the future. It says its analytics engine collects and analyses up to three years of historical data from millions of VMs, and applies algorithms to predict resource needs up to 18 months into the future.
Tintri Analytics with Compute Analytics, Storage File System 4.4, Global Center 3.6, VM Scale-out software with array offloading, and the Cloud Connector will all be available in the August-October 2017 timeframe. ®