A:
Performance in CAS systems is not just about max. read/write throughput under ideal conditions. It is also about attaining that same throughput if you are storing 10,000 small objects rather than 10 large ones: it may not seem difficult, but it is one of the toughest challenges for the design, yet very relevant in many CAS applications that need to store small objects such as email messages or high-volume scanned forms. Architectures layered on old technology tend to perform miserably in this respect - as expected. CAStor was built "on top of the bare metal" as well as specifically structured to maximize this very important benchmark.
Keep in mind that performance is more than just faster reads and writes. What happens when a disk fails? Many systems slow down to a crawl to recover — if they can recover at all. CAStor actually recovers faster as cluster size grows. At Caringo, we look at performance not as a single benchmark, but as a holistic thing. Time spent installing software, interfacing to it, installing nodes, adding nodes, managing and maintaining the system. Our development organization models the use of a CAStor system from design to deployment to operations to maintenance and expandability. Our symmetric architecture that allows adding nodes and disks while up and running improves aggregate throughput and recovery time. Our HTTP interface reduces the time it takes to deploy applications. Even our embedded operating system is trimmed down to its most efficient scale. We feel that performance is an essential factor to a storage system's success and we will continually seek to optimize the work CAStor performs.
|
|
email us © Copyright 2008 Caringo, Inc. All rights reserved.
