PCIe Going Mainstream

Solid-state storage analyst Jim Handy of Objective Analysis told eWEEK his firm is forecasting that the PCIe interface will become dominant in the enterprise SSD market in 2012, with unit shipments greater than the combined shipments of its SAS and Fibre Channel counterparts. PCIe (peripheral component interconnect express) is a computer expansion-card standard based on point-to-point serial links rather than shared parallel bus architecture. It is designed to replace the older PCI, PCI-X and AGP standards.

Oracle has a leg up on this one. Its new VirtualBox 4.1 includes a new virtual machine cloning facility — one of the first on the market. A regular VM snapshot is a child of the current virtual machine, but that’s not something that can independently grow afterwards. With a clone, you have a new entity that can then have its own life and, subsequently, its own data snapshots, said Wim Coekaerts, Oracle’s senior vice president of Linux and Virtualization Engineering.

Getting data stores reconnected with systems and getting those systems back up and running after a power outage is a bear that can take days. Whereas in the past this process was manually done, the software now is smart enough to get large portions of a virtualized system back online much faster and with less effort. So enterprises are checking this out very closely. The ability to have automated, fully tested disaster recovery is one of the key drivers for many organizations to virtualize their most important applications.

This came to the fore in 2010 at Sepaton and is now gaining momentum. Pooling is an approach to storage virtualization that delineates specific areas of the storage system to be dedicated to specific data flows to enable more efficient multitenant service deployments. Virtualized storage systems break files into chunks of data that are dispersed into numerous data center or storage locations and reassemble them on demand. Keeping data file chunks closer together in pools is said to provide faster reassembly of file chunks.

As different vendors clamor to be part of the cloud, unified management of the entire technology stack is critical. Whether public or private, tying together the different infrastructure layers—including applications, VMs, systems, networks and storage—with a comprehensive set of management tools reduces complexity by providing end-to-end service visibility, performance monitoring and automated provisioning.

Users discover that server virtualization can actually drive up higher storage costs, due to the fact that I/O patterns in servers that host VMs are significantly more random—and write-intensive than in the dedicated physical application servers. Increased randomness causes HDDs to run slower in terms of IOPS. This is why a storage configuration that seemed to perform fine when it was attached to physical servers all of a sudden appears to slow down when it’s attached to VMs. So you start adding more spindles and storage costs go up. (Thanks to storage analyst Eric Burgener for this one.)

The number of applications and amount of data in virtual environments will grow significantly in 2011, increasing the need for disaster recovery systems that protect these applications. A recent survey found little more than half of the data within virtual systems is regularly backed up, leaving plenty of room for improvement.

Storage administrators must lose their “pack rat” mentality and categorize what information is most important. The near-infinite level of data retention is causing storage costs to skyrocket. This factor vastly increases data recovery times and causes e-discovery nightmares across enterprises of all sizes. Enterprises are re-evaluating their retention needs and automate their information management strategy to keep backups for 30 to 60 days, archive for long-term storage and delete everything else.

The cloud is greatly changing the way services are now delivered. More enterprises will leverage public and private clouds as they become highly available. As we get deeper into the year, enterprises will require the ability to manage storage resources, whether they’re local, campus-wide, multi-campus, global or in the cloud.

The hybrid cloud archiving model is being adopted to allow enterprises to use hosted messaging services while keeping their archives on-premises to drive cost out of the discovery process, maintain strict access to data and define who is searching it and where they are sending requests.