Even More Wicked Fast Data Transport And Cloud Disaster Recovery - Forbes
Published in Forbes, March 12, 2019
At the end of the recent blog I did on Zettar and their high performance transfer on the Energy Sciences Network I asked if anyone can do data transfers faster than 1 PB of data transferred in 29 hours. Zettar had earlier responded to a piece I did on Data Expedition. To my surprise, I got a reply back from a company called Vcinity who transferred 1 PB of file sets across a single 100 Gbps WAN connection with 70 ms latency in about 23 hours and 16 minutes over 4,350 miles.
Vcinity was formed in 2018 and Russ Davis, the CTO/COO told me that what is most interesting, besides the fast data transport, is that their solution offers direct memory access of remote data (really remote data). Thus, they can do memory to memory copies or allow accessing data without moving or replicating it. The company provides federated compute and storage through distributed file services. The figure below shows their concept for long distance data access through their unified data fabric. In addition to using remote direct memory acceess (RDMA) this fabric uses advanced traffic engineering and flow control methods that work across any distance.
The company says it provides a simplified integration with open and well understood interfaces and API for existing IT architectures and uses an industry proven parallel file system designed for high performance I/O on this global fabric as well as global namespace views and distributed seamless access controls. The company says using RDMA and storage-based fabrics allows for a lossless and deterministic extension of local fabrics across global distances.
As a consequence, the company says it can reach 0-97% of the available bandwidth from WAN entry to WAN exit, compared to the minimal TCP/IP bandwidth utilization of about 30% and better than UDP based extreme file transfer solutions. Virtual point-to-point or point-to-multi-point tunnels, traffic engineering and flow control are applied end to end minimizing the risk of errors and dropped packets. Vcinity says that its fabric extension technology supports layer 2 or layer 3, concurrent, lossless, multi-fabric extension over a single connection at distances exceeding 250,000km for 10G or 25,000km for 100 G clients.
I also made a visit to cloud-based disaster recovery company Druva, based in Silicon Valley. They recently announced their Disaster Recovery-as-a-service (DRaaS) 2.0 release. According to the company their new capabilities include one-click failover to the cloud, simplified orchestration and testing, less than a 10 minute recovery time objective (RTO) and improved automation capabilities, while reducing costs by up to 60 percent. The Druva product is built on AWS storage. The recovery point objective (RPO) is an hour across any AWS Region. The figure below illustrates Druva’s disaster recovery process with both cloud and on-premises data centers.
Druva says their solution supports hybrid workload failback, such as to VMware Cloud on AWS or on-premises data centers. They can also capture data within customer’s AWS account and clone it across regions for testing and compliance. They can also use all AWS storage tiers to optimize cost versus performance. Enterprises can automate the disaster recovery testing process, ensuring their people and services are prepared for a potential disaster and meeting compliance and audit requirements. Users can also replicate virtual machines, clone full VPCs and move them cross-region for testing and development. The new product will be generally available in Q2 2019.
So, dare I ask it? Can anybody top 1 PB across a high speed long distance network in less than 23 hours and 16 minutes? Cloud storage use grows as data transport and direct memory access allow working with distant data and with cloud-based disaster recovery services.
Written by Tom Couglin, Contributor, Forbes