| |
| |
|
|
 |
Over the past decade, while the rest of IT infrastructure has improved exponentially in performance, storage has been virtually stagnant, increasing in capacity but not in performance. The table below shows how CPU, DRAM, network, and bus performance have outpaced storage by 16 to 400 times. To have kept pace, hard drives would need to spin at 6 million rpm today. Instead, storage is now the prime bottleneck in datacenters.
| |
2000 |
2005 |
2010 |
Today |
| CPU |
1 x
Pentium 4 1.5 GHz |
5 x
Pentium D 2.6 GHz |
15 x
Nehalem Quad 2.6 GHz |
30 x
Octal 2.6 GHz |
| DRAM |
1 x
DDR1 PC-2100 |
4 x
DDR2 PC2-4200 |
8 x
DDR3 PC3-8500 |
16 x
DDR3 PC3-17000 |
| Network |
1 x
100Mb Ethernet |
10 x
Gigabit Ethernet |
100 x
10 Gigabit Ethernet |
400 x
40 GbE |
| Bus |
1 x
PCI 32-bit/33 MHz |
15 x
PCIe Gen1 x8 |
30 x
PCIe Gen2 x8 |
60 x
PCIe Gen3 x8 |
| Disk |
1 x
15K rpm hard drive
|
1 x
15K rpm hard drive
|
1 x
15K rpm hard drive
|
|
|
|
 |
At the same time, power, cooling, rackspace, and datacenter maintenance costs have reached record levels. According to Gartner, storage consumes 37% of overall datacenter power. Organizations face relentless data growth without the facilities or operating budget to support them. Despite achievements in areal density, disk storage remains based on power-hungry, hot, large and comparatively unreliable technology. |
 |
Caching, spindle short-stroking, and most recently primary storage tiering have emerged as possible solutions to these challenges. Unfortunately, all of these approaches have significant drawbacks.
| |
Strengths |
Weaknesses |
| Caching |
| • |
Accelerates reads and sometimes writes |
|
| • |
Helps only if data happens to be in cache |
| • |
Small cache (usually less than 3% of storage capacity) means cache hits are rare |
| • |
Inconsistent performance |
| • |
No reduction in power consumption |
| • |
No cooling or rackspace savings |
|
| Short-stroking |
| • |
Uses outer edge of platters to boost speed |
|
| • |
Wastes 70% or more of the storage capacity |
| • |
Does not improve performance substantially |
| • |
Reduces efficiency significantly |
| • |
Increases hardware and operating cost |
| • |
Increases rackspace and failure likelihood |
|
| Tiering |
| • |
Moves data between SSD, 15K and 10K rpm disk |
|
| • |
Data is constantly in motion, increasing power consumption and hardware wear |
| • |
Data location may not be clear when there is a failure during tiering action |
| • |
Recipe for vendor lock-in, since all storage tiers must be procured from the same vendor |
| • |
Non-automated solutions require manual effort, complicating administration |
| • |
Insignificant gains in power or rack efficiency |
|
|
|
 |
Nimbus Sustainable Storage® provides the most intelligent and efficient solid state storage platform to enable enterprises and datacenters to solve storage performance and associated operating cost challenges. Nimbus Sustainable Storage systems, including the modular S-Class system and the cloud-scale E-Class system, represent a revolutionary departure from spinning disk arrays, delivering purpose-built all-flash storage with comprehensive unified storage management and data protection software built-in.
| Performance |
| • |
Delivers consistent and predictable high performance by eliminating rotational latency, dramatically accelerating both random and sequential workloads. |
|
| Efficiency |
| • |
Consumes one-fifth the power of disk arrays of the same capacity, cutting operating costs while enabling greener, denser, and longer-life datacenters. |
|
| Reliability |
| • |
Eliminates sensitive moving parts, minimizing the risk of mechanical or environmentally-induced failures to improve application availability. |
|
| Intelligence |
| • |
Provides complete storage software for unified provisioning, end-to-end data protection, and efficient capacity management in one elegant solution. |
|
|
|
|
|
|
