Riverbed Ramps Up Disaster Recovery
Riverbed's two new SteelHead appliance models, the 7050-L (low) and 7050-M (medium) add to the upper end of the company's appliance line, increasing WAN throughput and connection performance as well as enhancing disaster recovery fault tolerant options. The 7050-L lists for $179,995 and the 7050-M lists for $234,995 and will be available in Q1, 2010.
February 9, 2010
Riverbed's two new SteelHead appliance models, the 7050-L (low) and 7050-M (medium) add to the upper end of the company's appliance line, increasing WAN throughput and connection performance as well as enhancing disaster recovery fault tolerant options. The 7050-L lists for $179,995 and the 7050-M lists for $234,995 and will be available in Q1, 2010.
The Steelhead 7050 appliances are both 3U units that support up to 4Gbps LAN bandwidth and 1 Gbps WAN bandwidth. There is a 100GB partition set aside for the RiOS Services Platform (RSP), which runs virtual machines on the Steelheads. The 7050-L appliance ships with 14 hot-swappable 160GB SSD drives, providing 2.2TB of data store capacity and 32GB of RAM. The 7050-L can support 75,000 TCP connections. The 7050-M has 28 160GB SSD drives for 4.4TB of data storage and 48 GB of RAM, and it can support 100,000 TCP sessions. The appliances have 10Gb Ethernet NICs needed to connect to a 10Gb switch.
The Steelheads use a fault-tolerant data store. If one or more SSD drives fail, the Steelheads will stop using them for data storage. The purpose of the drives is to simply cache network data and not persistent storage. If a drive fails, you simply lose the cached data that will have to be re-warmed. The SSDs are hot-swappable through the front of the appliances, so replacing failed drives is easy and doesn't impact network traffic. In addition, Steelheads can be clustered side-by-side or in-line with the data stores synchronizing, so in the event of an appliance failure, the other Steelhead can take over.
Riverbed increased performance through the use of SSD drives, which unlike disk drives, don't have to contend with seek times to locate data on disk. Reading and writing to an SSD is fast with Input/Output Operations Per Second (IOPS) typically over 20 thousand, while hard disk drives are in the hundreds. The relative large amounts of RAM provide faster memory access for deduplication. In addition, the 7050s may be positioned to scale higher than 4Gb.
Riverbed is positioning the 7050 appliances for disaster recovery scenarios where data needs to be replicated between data centers over a large WAN pipe. Using Riverbeds Adaptive Data-Streamlining technology, the Steelheads dynamically recognize different traffic types and apply the appropriate optimization methods. Other WAN optimization products from vendors like Expand and Silverpeak have similar adaptive features features. Using Riverbeds clustering product, Interceptor, multiple Steelheads can be combined to increase traffic flow.The 7050's are bigger, faster appliances that will be a fit for a relatively small set of enterprises that have high capacity WAN links and need to keep data replicated between data centers. The Steelheads work with existing storage networks, though the products still have to be blessed by storage vendors via a product qualification process before the storage vendors will support a SAN with the Steelheads in place, but Riverbed qualifications are in progress.
While many storage networking vendors support data deduplication already, WAN optimization products like Riverbed's Steelheads can further deduplicate already deduplicated storage traffic. Storage deduplication tends to work larger 4KB blocks, whereas WAN deduplication works on smaller sizes in the hundreds of bytes. Storage and network deduplication works similarly. Repeating bit patterns in data is replaced with a reference or symbol that references the original bit pattern. The larger the block size used in matching, the fewer repeating blocks there are, and the less deduplication results. Since network deduplication found in WAN optimization appliances work on smaller blocks of data, the likelihood of repeating blocks is higher and greater deduplication takes place. Data that has already been deduplicated by a SAN can still be deduplicated via network deduplication.
You May Also Like