The WHIR is reporting live from Germany at WebhostingDay 2010. Stay tuned to our news, features, blogs and WHIR tv for more updates from the event.
(WEB HOST INDUSTRY REVIEW) — At WebhostingDay 2010, serial entrepreneur Kristof De Spiegeleer presented his vision of what’s wrong in data center and storage, and it all boils down to simplicity. Active in the space since 1985, and now as chairman of Amplidata and Racktivity, and director of SUN Microsystems Cloud Strategy, “15 years later and not much has changed,” a contentious statement, considering the crowd.
In his presentation, “Cloud 2.0: a vision on the next generation of cloud products,” De Spiegeleer presented some very compelling ideas on how to improve the entire data center environment — starting from the physical layer. He explains that most IT effort in the cloud is focused around upper layers, but there is little attention paid to automation, preventing of disasters, remote control, monitoring, auditing, power optimization, and a host of other problems.
“Physical layer is sometimes a forgotten layer.” For instance, in a 2,500 watt data center, only 1,000 watts are typically delivered out to the IT load because with each power conversion, power is lost, and after that, 90 percent of CPU capacity in data centers is not being used.
De Spiegeleer said there needs to be better infrastructure management. The older systems with raised floors are not efficient — but even in modern data centers, hardware can pose problems. For instance, some old switching equipment may be hanging around a data center. It performs its function, but it’s easy to ignore the fact that it’s consuming many times more energy than a modern component, and its replacement would deliver a full return on investment in months.
Data center operators must plan their power capacity to prevent blown fuses, which can easily result in the loss of a full rack. It’s also important to be aware of broken power supplies that draw loads of electricity. So it boils down to being aware of all hardware and having data on power use at a per-device level.
De Spiegeleer opened his next topic with, again, a contentious statement: “Storage technologies are not good today.” De Spiegeleer has a problem with the current volume-based approach to storage which remains the most popular, despite harkening back to an era when disks were extremely small. Now, they’re simply unreliable because the percentage of bit errors is too high. Also, solutions made to correct for these such as replication require too much bandwidth, which is growing much much slower than storage. “So that problem is only going to get worse,” he concluded.
His solution? Storage needs to be virtual. A solid-state drives should used for caching data in a storage system that is slower, but reliable, and it doesn’t have to be particularly fast because active data is a small percentage of all storage in any data center. This way a system can have unlimited backups, everything can be cloned, and you’re one step closer to an unbreakable storage system.
If this sounds too hard, De Spiegeleer introduced his company Amplidata. “It’s my fourth attempt at truly unbreakable storage,” he said, noting that this time it worked. At WebhostingDay, Amplidata gave its first presentation of Amplidata 2.0, which provides ubiquitous data availability at drastically lower costs.
Amplidata is a turnkey implementation of the innovative Distributed Storage System technology, which solves the scalability and reliability problems traditional storage systems face with the introduction of large capacity SATA drives and solid state disk. By storing data across a selection of disks that are widely distributed across storage nodes, racks and sites, data availability is dramatically improved.
Amplidata promises to reduce the risk for data loss from one event in years — as measured on current storage systems — to one in thousands of years. The DSS architecture ensures that a failure of a component such as a disk, storage node or even a full rack has no impact on data availability and a minimal impact on data redundancy. The entire system is permanently monitored and data integrity is constantly checked. As a result, bit errors on disks are proactively healed before they become an issue to the user.
Amplidata dovetails really well with other De Spiegeleer’s other companies Racktivity, which gives users full remote control over a physical data center, and A-Server, which specializes in ready-to-go cloud computing products for service providers.
After spending so much time in this fast growing industry, De Spiegeleer hopes he’s finally making progress.
