All stock codes associated to this product
S26361F5243L670, S26361-F5243-L670, 32S26361F5243L670, 4549210167340
Fujitsu S26361-F5243-L670 White Paper RAID Advanced
Software Options
Overview
This technical documentation is intended to help you become acquainted with
the RAID Advanced Software Options which are called LSI CacheCade® and Fast
Path.
Many companies are faced with a growing amount of data to manage combined
with an ever increasing demand for better performance. This problem is further
compounded by the ever-widening gap between compute processing and storage
processing. This gap, as illustrated in Figure 1 below, has increased over time
due to dramatic improvements in CPU processing capabilities and the advent of
multicore CPUs compared with only minimal improvements in hard disk drive (HDD)
performance, resulting in poorly performing databases. This poor performance is
a direct result of the amount of time CPUs spend waiting for data, with typical
CPU utilization rates commonly trapped at less than 25%. This problem is
exacerbated each time theoretical CPU performance is increased. Quite simply,
traditional storage has not been able to supply data quickly enough to keep up
with server CPUs.
Usually the way to increase the storage performance and capacity of a storage
array is to increase the number of disks in the array. Each drive in the array
is capable of reading and writing a given amount of data per second. The more
drives there are in the overall array, the more aggregate the data that can be
written to or read from the array. However, the problems with this solution are
significant. First, even though more drives offer better performance and greater
capacity, more drives also mean more complexity and increased management costs.
Second, the potential for failure increases, as additional system components are
added.
It turned out that a better way to increase performance is not to add more
HDDs or to short stroke the ones that are there. A good solution is to add SSDs.
There are two effective ways to add SSDs.
Fast Path
The first is to use SSDs to create a complete storage array. Using SSDs for
the storage array can offer extreme high-performance read and write data access.
Power and cooling requirements for this type of storage can be a fraction of
that required for traditional HDDs resulting in a green solution with
significantly improved ROI. Fast Path software is designed to get the best
possible performance from a pure SSD array. Fast Path software is a
high-performance IO accelerator for Solid State Drive (SSD) arrays connected to
a RAID controller card. This advanced software is an optimized version of LSI
MegaRAID® technology that can dramatically boost storage subsystem and overall
application performance — particularly those that demonstrate high random
read/write operation workloads — when deployed with a 6Gb/s MegaRAID SATA+SAS
controller connected to SSDs.
Range of Application and Performance
Application workloads that benefit most from Fast Path software with SSD
volumes are those with small and random IO patterns requiring high transactional
throughput, such as OLTP. The following two charts demonstrate read and write
performance in RAID 5 configurations with and without Fast Path.
The drawback to a pure SSD array is that it offers significantly less
available capacity than a HDD-based array with a comparable number of drives at
a given price point.
CacheCade
The second option for adding SSDs to a storage environment is to use them as
a cache for a traditional HDD array. This allows the SSDs to be used for
high-performance access to cached data, while still offering the higher capacity
of traditional hard drives. CacheCade software offers the perfect combination of
HDD capacity and SSD speed. CacheCade software is designed to improve the
performance of a server’s existing drive volume(s) by dynamically utilizing SSDs
as a dedicated pool of RAID controller caches to maximize random read and write
performance.
White paper RAID Advanced Software Options
In the following figure you can see how CacheCade software works in
particular. Frequently accessed data sets are copied from the HDD to lower
latency flash SSDs to improve system performance in subsequent read requests.
Range of Application and Performance
- CacheCade software is best suited for highly random, small I/O
block-oriented applications (such as web and SMB OLTP) that frequently read
from a small working data set. Applications with much larger active data sets
or more write-intensive profiles show limited improvement with this software.
Fast Path software is typically a better fit for this type of data set.
- CacheCade can deliver up to a 50X performance improvement in
read-intensive applications with relatively small working data set sizes, such
as web and file servers and online transaction processing.*
- The following two charts demonstrate the performance gains in Response
Time and in Transactions per Second (TPS) when implementing CacheCade
software.
- CacheCade Performance considerations: Hot Spot One essential prerequisite
for CacheCade 2.0 is a hot spot with random data access. Recognizing this as
such and estimating the size is one of the first steps for the planning of the
CacheCade 2.0 cache. Beside access to the hot spot, other accesses are
possible, but the proportion of the accesses to the hot spot should be 85% or
more. Over time a hot spot may relocate, whereby lower transaction rates are
set during this transition.
- Transaction rate improvement Further important aspect with CacheCade 2.0
is the difference between the transaction rates of the HDD array and the SSD
array. If it is relatively large, the use of CacheCade 2.0 will be
particularly effective. If the HDD array is on the other hand already designed
for a high transaction rate, e. g. due to a large number of HDDs, the
difference will be minor.
- Time factor Ultimately, the time factor is of interest, i. e. how long it
takes before CacheCade 2.0 provides higher transaction rates. This can be the
case after a few minutes for a small hot spot. However, in the case of a large
data volume or a HDD array with a low transaction rate it can take several
hours.
- Transaction rate falls when the flash memory cells are not empty. It is
only possible to write on flash memory cells if they are empty. Re-use
therefore requires the prior erasure of the flash memory cell. This operation
is known as the program/erase cycle and is associated with higher response
times.
- An example of CacheCade performance
- Profile File copy is Random Access 100%, Read 50% write 50%, 64 Kb block
size
- In practice the maximum transaction rate of the SSDs is not achieved, as a
certain overhead for CacheCade 2.0 has to be taken into account This shows
that performance drops until the Hot Spot is found.
