Thoughts on Cloudera and Cisco UCS reference architecture for Apache Hadoop
Cloudera and Cisco jointly announced a reference architecture for running Cloudera’s Distribution Including Apache Hadoop (CDH) and Cloudera Manager on Cisco’s Unified Computing System (UCS) last November. It was the first Apache Hadoop reference architecture assembled by Cisco, and is proudly certified by Cloudera.
I bring a different perspective on the Cloudera-Cisco relationship, as I worked for over five years in Cisco on the software powering the Nexus 5000 series switches and the Cisco Virtual Interface Card. I now work at Cloudera on the HBase team, and can fully appreciate the synergies that the Cloudera and Cisco reference architecture brings to the table.
I know that Cisco is actively investing in development and support for the rack servers and switches that comprise UCS. Working at Cloudera now, I see the tireless effort that goes into each CDH and Cloudera Manager release from leaders in the Apache Hadoop community.
The combination of CDH and UCS is a natural marriage of best-of-breed technologies that delivers clear advantages to the big data end-user:
- CDH is a supported and proven big data platform, with clear advantages in ease of deployment and no vendor lock-in since it is 100% open source
- Cloudera Manager provides easy and powerful Hadoop cluster management
- UCS provides massive cross-sectional network bandwidth between servers when using 10 gigabit Ethernet adapter cards. More bandwidth means faster execution of network-heavy jobs, especially within a rack.
- UCS provides very good cluster management capabilities through UCS Manager. UCS Manager empowers the network and server adminstrators with fully-integrated control and monitoring of compute, storage, and networking elements from management GUI, CLI, XML, and SNMP interfaces.
- The reference architrecture also specifies up to 16 TB of storage per node for large MapReduce workloads, and a wide range of Intel Xeon CPUs, with timely upgrades to Romley and beyond.