Shared IT Infrastructure with the NetApp FAS6200 Series

Up to a 3.6X Performance Boost

When we set out to refresh the FAS6000 series, we knew we needed to push the envelope in almost every dimension. Our goal was to allow you to consolidate more storage workloads than ever before onto a single storage system to support shared infrastructure or other needs, while still leaving lots of headroom for important storage tasks like data protection. In head-to-head comparisons using enough spindles, hosts, and host connections to eliminate the possibility of artificial bottlenecks, the FAS6280 handily outperformed the FAS6080 on every benchmark, including:

• 3.6x improvement in sequential read performance
• 2x improvement in sequential write performance
• 2x improvement in overall NFS performance

These performance improvements are the result of innovations throughout the architecture including more and faster processing cores, more and faster memory, more NVRAM, support for greater quantities of Flash Cache, and an increase in the maximum number of spindles. The FAS6200 series uses the Iatest Intel® Westmere and Nehalem microarchitecture along with DDR3 memory to delivery industry-leading performance.

As with the FAS6000 series, the FAS6200 uses a 64-bit architecture. However, the new Westmere and Nehalem processors are capable of providing much higher data throughput.

The DDR3-1066MHz memory in the FAS6200 also offers much faster memory transfers versus the DDR-333MHz memory used in the FAS6000 series, so that that FAS6200 processor cores aren’t left waiting for the completion of memory transfers.

We also added TCP/IP stateless offload to all onboard 10GbE network ports so that storage system cores don’t get bogged down with network processing. Overall internal bandwidth was increased to 90 gigabytes per second from 64 gigabytes per second using the new PCIe Gen 2 architecture and providing more I/O pathways to onboard connections and expansion slots. PCIe v2.0 (Gen 2) x8 architecture delivers bandwidth up to 4GB/sec to each slot versus 2.5GB/sec for the previous-generation systems.

FAS3200 Dramatically Boosts Expandability

Collaboration with Intel for Performance

The FAS/V3210 and FAS/V3270 use Intel® Wolfdale processors while the FAS/V3240 uses Intel Harpertown. NetApp Engineering worked directly with Intel to modify the Intel chipset to allow NetApp NVRAM to be part of main memory.

As you probably already know, all NetApp systems use nonvolatile memory to log incoming write requests as a means of accelerating write response times. This has typically been provided as a separate card or a dedicated portion of the motherboard. Providing NVRAM from main memory is faster and also decreases cost.

NetApp Engineering worked with Intel to create a self-refresh sleep mode in the Intel chipset called Asynchronous DRAM Refresh (ADR) that allows a portion of main memory to be backed by an on-board battery. This gives the NVRAM the same high bandwidth as main memory and simplifies the motherboard design.

The performance gains we’ve seen with the FAS/V3200 line are due in part to this change as well as to the great performance we’ve seen from the Intel processors. For instance, on the SPECsfs benchmark (SPECsfs2008_nfs.v3), the standard benchmark for measuring NFS performance, the FAS3270 delivers 101,183 ops at 1.66ms ORT while the previous FAS3160 midrange submission was 60,507 ops at 1.58ms ORT, providing almost a 70% increase in SFS throughput.


For out-of-band management of NetApp SAS disk shelves, NetApp has created its Alternate Control Path (ACP) technology. ACP enhances data availability by giving a storage controller the ability to reset a storage channel without having to communicate over that channel. If a channel is down or misbehaving, a quick reset can bring it back online without external intervention. ACP allows a storage system to recover from faults that might otherwise require it to reboot, and that’s a big advantage.

ACP gives you a back door into your disk shelves. It is completely separate from the SAS data path and provides new options for nondisruptive recovery of shelf modules, including the ability to reset or power-cycle an individual I/O module or an entire domain. We designed in the ability to power-cycle an entire shelf as well. ACP technology enhances the ability of Data ONTAP to automatically reset a misbehaving component in order to return it to a fully operational mode without disruption.

Zdroj:
https://communities.netapp.com/docs/DOC-9215

Spúšťam blog o virtualizácii ukladania dát

Pozdravujem všetkých nadšencov virtualizácie. Postupne budem pridávať nové zaujímavé postrehy z oblasti virtualizácie ukladania dát.