AMI StorTrends Optimizes WANs for Primary Data
One expects primary storage to be fast. AMI StorTrends networked storage delivers performance in spades, especially for intensive data IO. However, the same optimization that works for high performance iSCSI storage over the LAN does not have the same effect over the WAN. This can present a real challenge for administrators who replicate primary data to remote data centers or DR sites. They need fast and secure data transport, but the WAN’s distance-related challenges requires separate optimization from the LAN.
StorTrends is first optimized for primary storage on the LAN where speed is of the essence. StorTrends optimizes the TCP stack for a fast rate of data IOs moving over the local network to its iSCSI storage devices. However, the same optimization that works so well over the LAN sputters over the WAN, where TCP/IP latency and packet loss remediation become real problems.
Certainly LANs have their own delays – there is as yet no instantaneous data transport outside of science fiction – but it rarely impacts applications or users. However, WAN delays can range from an acceptable 2 milliseconds (ms) across a campus or city to 80-100 ms and higher across countries and across the world. Ultimately even replication latencies multiply, increasing the time the operation takes and the risk of not having the replicated data complete by the time you need it. This risk grows with the geographical distance that the WAN must travel, and the business priority of the replicated data. (Note that bigger bandwidth makes relatively less difference given increasingly longer distances. The pipes may be big but latency gets worse the more distance the data covers, which retards the data transfer even over fast bandwidth.)
AMI StorTrends understands this problem very well and set out to optimize performance over the WAN as well as the LAN. To provide timely asynchronous WAN replication, StorTrends uses iTX Data Storage Software. iTX software combines their WDS (WAN-optimization Data Services) feature with IP protocols to optimize data transfer speeds across WAN bandwidth. Part of the WAN solution is also to avoid transferring data in the first place. StorTrends uses delta snapshots to transfer only data that has changed between replication passes. StorTrends also compresses, dedupes and encrypts data at the source to further relieve WAN congestion and latency. Compression includes dynamically matching compression rates to current workloads on the WAN. The iTX deduplication process identifies data identical to previous replications, and avoids re-replicating them over the WAN.
For example, a storage administrator has had bad luck with a replication program taking much too long. They had to compromise both on amount of data and frequency of replication, and felt they were risking critical data. Even when they installed more bandwidth, the remotest of their sites frequently took longer than its allotted time. The administrator then deploys StorTrends arrays in the data center, two DR sites and in a branch office. He schedules one-to-many replication from the central StorTrends array to two DR sites residing on the east and west coasts. Working with the branch office’s general IT, he also sets up a one-to-one replication from the office to the corporate data center. Now iTX software optimizes all three WAN paths so replication easily makes its windows. The administrator now has the option to replicate more data and to do it more frequently, and plans to expand replication to more branch offices.
Optimizing the WAN for primary data replication can be easier said than done, but AMI StorTrends has managed it. They have integrated WAN optimization for one-to-one and many to one replication. This results in highly efficient ROBO and DR replication that is fully integrated with fast primary storage.
There are no comments to display. Scroll down to leave your own!