White Paper

Abstract
This white paper is an introduction to EMC VFCache. It
describes the implementation details of the product and
provides performance, usage considerations, and major
customer benefits when using VFCache.

February 2012




INTRODUCTION TO EMC VFCACHE

VFCache is a server Flash-caching solution
VFCache accelerates reads and ensures data protection
VFCache extends EMC FAST Suite to server

2 Introduction to EMC VFCache






















Copyright 2012 EMC Corporation. All Rights Reserved.

EMC believes the information in this publication is accurate as
of its publication date. The information is subject to change
without notice.

The information in this publication is provided as is. EMC
Corporation makes no representations or warranties of any kind
with respect to the information in this publication, and
specifically disclaims implied warranties of merchantability or
fitness for a particular purpose.

Use, copying, and distribution of any EMC software described in
this publication requires an applicable software license.

For the most up-to-date listing of EMC product names, see EMC
Corporation Trademarks on EMC.com.

VMware, ESX, VMware vCenter, and VMware vSphere are
registered trademarks or trademarks of VMware, Inc. in the
United States and/or other jurisdictions. All other trademarks
used herein are the property of their respective owners.

Part Number: H10502.1

3 Introduction to EMC VFCache
Table of Contents
Executive summary.................................................................................................. 4
EMC VFCache solution ........................................................................................................ 4
Introduction ............................................................................................................ 5
Audience ............................................................................................................................ 5
Terminology ....................................................................................................................... 5
Use cases of Flash technology ................................................................................. 6
VFCache advantages over DAS ....................................................................................... 6
Cache in the storage array .............................................................................................. 7
Flash cell architecture ........................................................................................................ 7
VFCache design concepts ...................................................................................... 10
Business benefits ............................................................................................................. 13
Implementation details ......................................................................................... 15
Read Hit example ......................................................................................................... 15
Read Miss example ...................................................................................................... 16
Write example .............................................................................................................. 17
VMware implementation .................................................................................................. 18
Split-card feature .............................................................................................................. 20
VFCache management ...................................................................................................... 21
Performance Considerations .................................................................................. 22
Locality of reference ......................................................................................................... 22
Warm-up time ................................................................................................................... 22
Workload characteristics .................................................................................................. 23
Throughput versus latency ............................................................................................... 24
Other bottlenecks in the environment .............................................................................. 24
Write performance dependent on back-end array ............................................................. 24
Usage guidelines and characteristics ..................................................................... 25
Specifications .................................................................................................................. 25
Constraints ....................................................................................................................... 26
Stale data ......................................................................................................................... 26
Application use case and performance ................................................................... 28
Test results ....................................................................................................................... 29
Conclusion ............................................................................................................ 30
References ............................................................................................................ 31



4 Introduction to EMC VFCache
Executive summary
Since the first deployment of Flash technology in disk modules (commonly known as
solid-state drives or SSDs) by EMC in enterprise arrays, it has been EMCs goal to
expand the use of this technology throughout the storage environment.
The combination of the requirement for high performance and the rapidly falling cost-
per-gigabyte of Flash technology has led to the concept of a caching tier. A caching
tier is a large-capacity secondary cache using Flash technology that is positioned
between the server application and the storage media.
EMC VFCache solution
EMC VFCache is a server Flash-caching solution that reduces latency and
accelerates throughput to dramatically improve application performance by using
intelligent caching software and PCIe Flash technology.
VFCache accelerates reads and protects data by using a write-through cache to the
networked storage to deliver persistent high availability, integrity, and disaster
recovery.
VFCache coupled with array-based EMC

FAST software provides the most efficient

and intelligent I/O path from the application to the data store. The result is a
networked infrastructure that is dynamically optimized for performance, intelligence,
and protection for both physical and virtual environments.
Major VFCache benefits include:
Provides performance acceleration for read-intensive workloads
As a write-through cache, enables accelerated performance with the protection of
the back-end, networked storage array
Provides an intelligent path for the I/O and ensures that the right data is in the
right place at the right time
In split-card mode, enables you to use part of the server Flash for cache and the
other part as direct-attached storage (DAS) for temporary data
By offloading Flash and wear-level management onto the PCIe card, uses minimal
CPU and memory resources from the server
Works in both physical and virtual environments
As VFCache is installed in a greater number of servers in the environment, more
processing is offloaded from the storage array to the server. This provides a highly
scalable performance model in the storage environment.


5 Introduction to EMC VFCache
Introduction
This white paper provides an introduction to VFCache. Flash technology provides an
opportunity to improve application performance by using it in different ways in a
customer environment. Topics covered in this white paper include implementation in
physical and virtual environments, performance considerations, best practices, usage
guidelines and characteristics, and some application-specific uses cases.
Audience
This white paper is intended for EMC customers, partners, and employees who are
considering the use of VFCache in their storage environment. It assumes a basic
understanding of Flash technology and its benefits.
Terminology
Cache page The smallest unit of allocation inside the cache, typically a few
kilobytes in size. The VFCache cache page size is 8 KB.
Cache warm-up The process of promoting new pages into the VFCache after they
have been referenced, or a change in the application access profile that begins to
reference an entirely new set of data.
Cache promotion The process of copying data from the SAN storage in the back end
to VFCache.
Hot spot A busy area in a source volume.
Spatial locality of reference The concept that different logical blocks located close
to each other will be accessed within a certain time interval.
Temporal locality of reference The concept that different logical blocks will be
accessed within a certain time interval.
Working set A collection of information that is accessed frequently by the
application over a period of time.


6 Introduction to EMC VFCache
Use cases of Flash technology
There are different ways in which Flash technology can be used in a customer
environment depending on the use case, application, and customer requirements.
EMCs architectural approach is to use the right technology in the right place at the
right time. This includes using Flash:
In the storage array
As a cache
As a tier
As a singular tier for the entire application
In addition, there are different types of Flash, with different cost structures, different
reliability considerations, and different performance characteristics. All of these
different types of Flash have a proper place in the vast use cases continuum. These
technologies range from PCIe to SSD, as well as single-level cell (SLC) to multi-level
cell (MLC) Flash architecture.
Some of the use cases in which Flash can be used (and some of them may overlap)in
a customer environment are:
Applications with high performance and protection requirements that may be read
heavy are a perfect fit for using PCIe Flash in the server as a cache, for example,
VFCache.
Applications with performance requirements without protection requirements,
such as temporary data, may be a good fit for PCIe Flash in the server as a data
store (direct-attached storage (DAS)), for example, the split-card feature in
VFCache.
Applications with performance and protection requirements that are read and
write heavy may be a good fit for Flash in the array as a cache, for example, EMC
FAST Suite on an EMC VNX storage system.
Applications with mixed workloads and changing data temperature is a perfect
fit for Flash as part of a tiering strategy, for example, Fully Automated Storage
Tiering for Virtual Pools (FAST VP) on an EMC storage system.
Applications requiring high consistent performance may be a good fit for Flash as
the single tier of storage, for example, using SSDs in an EMC storage system.
VFCache advantages over DAS
One option to use PCIe Flash technology in the server is to use it as a DAS device
where the application data is stored on the Flash. Advantages of using VFCache over
DAS solutions include:
DAS solutions do not provide performance with protection, since they are not
storing the data on an array in the back end. If the server or the Flash card is
faulted, you run the risk of data unavailability or even data loss. VFCache,

7 Introduction to EMC VFCache
however, provides read acceleration to the application, and at the same time
mirrors application writes to the back-end storage array, thereby providing
protection.
DAS solutions are limited by the size of the installed Flash capacity and do not
adapt to working sets of larger datasets. In contrast, once the working set of
the application has been promoted into VFCache, application performance is
accelerated. Then, when the working set of the application changes, VFCache
adapts to it and promotes the new working set into Flash over a period of time.
DAS solutions lead to stranded sets of data in your environment that have to
be managed manually. This is in contrast to application deployment on a
storage array where data is consolidated and can be centrally managed.
Cache in the storage array
Another way in which some solutions use PCIe Flash technology is to use it as a cache
in the storage array. However, VFCache uses PCIe Flash in the server and is much
closer to the application in the I/O stack. VFCache does not have the latency
associated with the I/O travelling over the network to access the data.
Flash cell architecture
In general there are two major technologies used in all Flash drives:
Single-level cell (SLC) NAND-based Flash cell
Multi-level cell (MLC) NAND-based Flash cell
A cell is the smallest unit of storage in any Flash technology and is used to hold a
certain amount of electronic charge. The amount of this charge is used to store binary
information.
NAND Flash cells have a very compact architecture; their cell size is almost half the
size of a comparable NOR Flash cell. This characteristic, when combined with a
simpler production process, enables a NAND Flash cell to offer higher densities with
more memory on a given semiconductor die size. This results in a lower cost per
gigabyte. With smaller, more precise manufacturing processes being used, their price
is expected to fall even further.
A NAND Flash cell has faster erase and write times compared to NOR Flash cells,
which provides improved performance. It has higher endurance limits for each cell, a
feature that provides the reliability required for enterprise-class applications.
Flash storage devices store information in a collection of Flash cells made from
floating gate transistors. SLC devices store only one bit of information in each Flash
cell (binary), whereas MLC devices store more than one bit per Flash cell by choosing
between multiple levels of electrical charge to apply to its floating gates in the
transistors (Figure 1).

8 Introduction to EMC VFCache

Figure 1: Comparison between SLC and MLC Flash cell data storage
1
Since each cell in MLC Flash has more information bits, an MLC Flash-based storage
device offers increased storage density compared to an SLC Flash-based version;
however, MLC NAND has lower performance and endurance because of its inherent
architectural tradeoffs. Higher functionality further complicates the use of MLC NAND,
which makes it necessary for you to implement more advanced Flash management
algorithms and controllers.
SLC NAND and MLC NAND offer capabilities that serve two very different types of
applications those requiring high performance at an attractive cost per bit (MLC),
and those who are less cost sensitive and seeking even higher performance over time
(SLC).
Taking into account the kind of I/O profiles in enterprise applications and their
requirements, EMC VFCache uses the SLC NAND Flash architecture.
Table 1 compares the SLC and MLC Flash characteristics (typical values).







1
Kaplan, Francois, Flash Memory Moves from Niche to Mainstream, Chip Design Magazine, April/May 2006


9 Introduction to EMC VFCache
Table 1: SLC and MLC Flash comparison
Features MLC SLC
Bits per cell 2 1
Endurance (Erase/Write cycles) about 10 K about 100 K
Read service time (Avg) 129 us 38 us
Write service time (Avg) 1,375 us 377 us
Block Erase (Avg) 4,500 us 1,400 us
Although SLC NAND Flash offers a lower density, it also provides an enhanced level of
performance in the form of faster reads and writes. Because SLC NAND Flash stores
only one bit per cell, the need for error correction is reduced. SLC also allows for
higher write/erase cycle endurance, making it a better fit for use in applications
requiring increased endurance and viability in multi-year product life cycles.
For more details on various Flash cell architectures, refer to the Considerations for
Choosing SLC versus MLC Flash technical notes on Powerlink.


10 Introduction to EMC VFCache
VFCache design concepts
Over the past decade, server processing technology has continued to advance along
the Moores Law curve. Every 18 months, memory and processing power have
doubled, but disk drive technology has not. Spinning drives continue to spin at the
same rate. This has caused a bottleneck in the I/O stack whereby the server and the
application have capacity to process more I/O than the disk drives can deliver. This is
referred to as the I/O gap, as shown in Figure 2.

Figure 2: I/O gap between the processor and storage sub-systems
Flash drives in the storage system have helped to close this gap, and EMC is a very
successful industry leader in Flash drives. Flash is a silicon technology, not
mechanical, and therefore can enjoy the same Moores Law curve.
Flash technology itself can be used in different ways in the storage environment.
Figure 3 shows a comparison of different storage technologies based on the I/O per
second (IOPS) per gigabyte (GB) of storage that they offer.

11 Introduction to EMC VFCache

Figure 3: Comparison of storage technologies
Mechanical spinning drives provide a great dollar-per-gigabyte economic value to
cold datasets, but they do not provide the best performance. Putting Flash drives in
the array provides an order of magnitude better performance. Putting Flash in the
server on a PCIe card can accelerate performance by even another order of magnitude
over Flash drives.
FAST technology on EMC storage arrays can help place the application data in the
right storage tier based on the frequency with which data is being accessed. VFCache
extends FAST technology from the storage array into the server by identifying the most
frequently accessed data and promoting it into a tier that is closest to the application.
EMC VFCache is a hardware and software server caching solution that dramatically
improves your application response time and delivers more IOPS. It intelligently
determines, through a fully automated tiering (FAST) algorithm, which data is the
hottest data and would benefit by sitting in the server on PCIe Flash and closer to
the application. This avoids the latencies associated with I/O accesses over the
network through to the storage array. Once enough data from the application working
set has been promoted into VFCache, future accesses to the data will be at very low
latencies. This results in an increase of performance by up to 300 percent and a
decrease in latency by as much as 50 percent in certain applications.
Since the processing power required for an applications most frequently referenced
data is offloaded from the back-end storage to the PCIe card, the storage array can
allocate greater processing power to other applications. While one application is
accelerated, the performance of other applications is maintained or even slightly
accelerated.
EMC VFCache is EMCs newest intelligent software technology which extends EMC
FAST into the server. When coupled with FAST, VFCache creates the most efficient
and intelligent I/O path from the application to the data store. With both

12 Introduction to EMC VFCache
technologies, EMC provides an end-to-end tiering solution to optimize application
capacity and performance from the server to the storage. As a result of the VFCache
intelligence, a copy of the hottest data automatically resides on the PCIe card in the
server for maximum speed. As the data slowly ages and cools, this copy is discarded
and FAST automatically moves the data to the appropriate tier of the storage array
from Flash drives to FC/SAS drives and SATA/NL-SAS drives over time.
VFCache ensures the protection of data by making sure that all changes to the data
continue to persist down at the storage array, and uses the high availability and end-
to-end data integrity check that a networked storage array provides. Figure 4 shows a
VFCache deployment in a typical environment.

Figure 4: Typical EMC VFCache deployment
VFCache is designed to follow these basic principles:
Performance: Reduce latency and increase throughput to dramatically improve
application performance
Intelligence: Add another tier of intelligence by extending FAST into the server
Protection: Deliver performance with protection by using the high availability and
disaster recovery features of EMC networked storage

13 Introduction to EMC VFCache
Business benefits
VFCache provides the following business benefits:
Because of the way VFCache works, a portion of I/O processing is offloaded from
the storage array to the server where VFCache is installed. As VFCache is installed
on more servers in the environment, more I/O processing is offloaded from the
storage array to the servers. The result is a highly scalable I/O processing storage
modela storage environment with higher performance capability.
As VFCache helps in offloading workload from the storage array, the disk drives
may become less busy and can be reclaimed and used for other applications.
Note This should be done only after carefully studying the workload patterns
and current utilization of disk drives.
VFCache increases the performance and reduces the response time of
applications. For some businesses, this translates into an ability to do faster
transactions or searches, and more of them.
For example, a financial trading company may be limited in the number of
transactions it can process because of the number of IOPS that the storage
environment can provide. VFCache increases throughout to allow for more trades,
thereby generating more revenue for the company.
As another example, visitors to an e-commerce website may experience delays
because of the speed at which data can be read from the back-end storage. With
reduced latencies from VFCache, searches will be faster and web pages will load
in less time, which in turn improves the user experience of the site.
Typical customer environments might have multiple applications accessing the
same storage system in the back end. If some of these applications are more
important than others, you want to get the best performance for these
applications while making sure that the other non-critical applications continue to
get good enough performance.
Since VFCache is installed in the server instead of the storage, it provides this
flexibility. With multiple applications accessing the same storage, VFCache
improves the performance of the application on the server where it is installed,
while other applications on other servers continue to get good performance from
the storage system. In fact, they might get a small performance boost because
part of the back-end storage systems workload gets offloaded to VFCache, and
the storage system has more processing power available for these applications.
VFCache also provides you with the capability to configure VFCache at the server
volume level. If there are certain volumes, like application logs, which do not
need to be accelerated by VFCache, those specific devices can be excluded from
the list of VFCache-accelerated volumes.
In a virtual environment, VFCache provides the flexibility to choose the virtual
machines and their source volumes that you want to accelerate using VFCache.

14 Introduction to EMC VFCache
VFCache is a server-based cache and therefore completely infrastructure agnostic.
It does not require any changes to the application above it, or the storage systems
below it. Introducing VFCache in a storage environment does not require you to
make any changes to the application or storage system layouts.
Since VFCache is a caching solution and not a storage solution, you do not have to
move the data. Therefore data is not at risk of being inaccessible if the server or
the PCIe card fails.
VFCache does not require any significant memory or CPU footprint, as all Flash
and wear-level management is done on the PCIe card and does not use server
resources. There is no significant overhead from using VFCache on the server
resources, unlike other PCIe solutions.
Split-card mode in VFCache allows you to use part of the server Flash for cache
and the other part as DAS for temporary data.


15 Introduction to EMC VFCache
Implementation details
This section of the white paper provides details about how I/O operations are
handled when VFCache is installed on the server. In any implementation of VFCache,
the following components need to be installed in your environment:
Physical VFCache card
VFCache vendor driver
VFCache software
In a physical environment (non-virtualized), all the components need to be installed
on the server where VFCache is being used to accelerate application performance. For
more information abou the installation of these components, see the EMC VFCache
Installation and Administration Guide for Windows and Linux.
Figure 5 shows a simplified form of VFCache architecture. The server consists of two
components the green section on top shows the application layer, and the blue
section on the bottom shows the VFCache components in the server (SAN HBA shown
in the figure is not part of VFCache).
VFCache hardware is inserted in a PCIe Gen2, x8 slot in the server. VFCache software
is implemented as an I/O filter driver in the I/O path inside the operating system. One
or more back-end storage LUNs or logical volume manager volumes are configured to
be accelerated by the VFCache. Every I/O from the application to an accelerated LUN
or volume is intercepted by this filter driver. Further course of action for the
application I/O depends on the particular scenario when the I/O is intercepted.
In the following examples, if the application I/O is for a source volume on which
VFCache has not been enabled, the VFCache driver is transparent to the application
I/O, and it gets executed in exactly the same manner as if there was no VFCache
driver in the server I/O stack. You can assume that the application I/O is meant for a
source volume which is being accelerated by VFCache in the following examples.
Read Hit example
In this example, you can assume that the VFCache has been running for some time,
and the application working set has already been promoted into VFCache. The
application issues a read request, and the data is present in VFCache. This process is
called Read Hit. The sequence of steps is detailed below Figure 5.

16 Introduction to EMC VFCache

Figure 5: Read Hit example with VFCache
1. The application issues a read request that is intercepted by the VFCache
driver.
2. Since the application working set has already been promoted into VFCache,
the VFCache driver determines that the data being requested by the
application already exists in the VFCache. The read request is therefore
forwarded to the PCIe VFCache card, rather than to the back-end storage.
3. Data is read from the VFCache and returned back to the application.
This use case provides all the throughput and latency benefits to the application,
since the read request is satisfied within the server itself rather than incurring all the
latencies of going over the network to the back-end storage.
Read Miss example
In this example, you can assume that the application issues a read request, and that
data is not present in VFCache. This process is called Read Miss. The data might not
be in VFCache because the card has just been installed in the server, or the
application working set has changed so that this data has not yet been referenced by
the application. The sequence of steps is detailed below Figure 6.


17 Introduction to EMC VFCache


Figure 6: Read Miss example with VFCache
1. The application issues a read request that is intercepted by the VFCache
driver.
2. The VFCache driver determines that the requested data is not in VFCache and
forwards the request to the back-end storage.
3. The data is read from the back-end storage and returned back to the
application.
4. Once the application read request is completed, the requested data is written
by the VFCache driver to the VFCache card. This process is called Promotion.
This means that when the application reads the same data again in future, it
will be a Read Hit for VFCache, as explained previously.
If all cache pages in VFCache are already used, VFCache uses a least-recently-used
(LRU) algorithm to write new data into itself. If needed, data that is least likely to be
used in future is discarded out of VFCache first to create space for the new VFCache
promotions.
Write example
In this example, you can assume that the application has issued a write request. The
sequence of steps is detailed below Figure 7.

18 Introduction to EMC VFCache

Figure 7: Write example with VFCache
1. The application issues a write request that is intercepted by the VFCache
driver.
2. Since this is a write request, the VFCache driver passes this request to the
back-end storage for completion.
3. Once the write operation is completed on the back-end storage, an
acknowledgment for the write request is sent back to the application.
4. The data in the write request is written to the VFCache card. If the application
is writing to a storage area that has already been promoted into VFCache, the
copy of that data in VFCache is overwritten. The application therefore will not
receive a stale or old version of data from the VFCache. VFCache algorithms
ensure that if the application writes some data and then reads the same data
later on, the read requests will find the requested data in VFCache.
The process of promoting new data into VFCache as explained in the previous two
examples is called Cache Warmup. Any cache needs to be warmed up with the
application working set before the application starts seeing the performance benefits.
When the working set of the application changes, the cache will automatically warm
up with the new data over a period of time.
VMware implementation
The implementation of VFCache in a VMware environment is slightly different from the
implementation in a physical environment. In a virtualized environment, multiple
virtual machines on the same server may share the performance advantages of a
single VFCache card. This is shown in Figure 8.

19 Introduction to EMC VFCache

Figure 8: VFCache implementation in a VMware environment
VFCache implementation in a VMware environment consists of the following
components:
Physical VFCache card on the VMware ESX

server
VFCache vendor driver on the ESX server
VFCache software in each virtual machine that needs to be accelerated using
VFCache
In a VMware environment, the VFCache software includes the VFCache driver, CLI
package, and VFCache Agent. The VFCache software does not need to be installed
on all the virtual machines in the server. Only those virtual machines that need to
be accelerated using VFCache need to have VFCache software installed.
VFCache VSI Plug-in for VFCache management in the VMware vCenter client
This is usually the laptop that the administrator uses for connecting to the vCenter
server.
You have to create a datastore using the VFCache hardware on the ESX server. Once
the VFCache datastore has been created, the rest of the setup can be managed using
the VFCache VSI plug-in. In order for a virtual machine to use the VFCache datastore, a
virtual disk (vDisk) for the virtual machines cache device must be created within the
VFCache datastore. vDisks can be created either through the VFCache VSI plug-in or
directly using the vSphere client. This virtual disk needs to be added to the virtual
machine.
The cache configuration and management steps from this point on are similar to the
steps that you would follow in a physical server environment. These can be done

20 Introduction to EMC VFCache
using either the CLI on the virtual machine or the VSI plug-in on the vCenter client.
More details on installation of VFCache in VMware environments can be found in
VFCache Installation Guide for VMware and VFCache VMware Plug-in Administration
Guide available on Powerlink.
Depending on the cache size required on each virtual machine, an appropriate sized
cache vDisk can be created from the VFCache datastore and assigned to the virtual
machine. If you want to change the size of VFCache on a particular virtual machine,
you need to do the following:
1. Shut down the virtual machine.
2. Increase the size of the cache vDisk assigned to the virtual machine.
3. Restart the virtual machine.
VFCache is a local resource at the virtual machine level in the ESX server. This has the
same consequences as any other local resource on a server. For example, you cannot
configure an automatic failover for a virtual machine that has VFCache. You cannot
use features like VMware vCenter Distributed Resource Scheduler (vCenter DRS)for
clusters or VMware vCenter Site Recovery Manager (vCenter SRM) for replication.
You cannot use VFCache in a cluster that balances application workloads by
automatically performing vMotion from heavily used hosts to less-utilized hosts. If
you are planning to use vMotion functionality, you should:
1. Stop VFCache on the source virtual machine.
2. Remove VFCache from the source virtual machine.
3. Perform vMotion from the source virtual machine to the destination virtual
machine.
4. Restore caching on the destination virtual machine.
Both RDM and VMFS volumes are supported with VFCache.
Split-card feature
EMC VFCache has a unique "split-card" feature, which allows you to use part of the
server Flash as a cache and another part of the server Flash as DAS. When using the
DAS portion of this feature, both read and write operations from the application are
done directly on the PCIe Flash capacity in the server.
The contents of the DAS portion do not persist to any storage array. Therefore, it is
highly recommended that you use the DAS portion only for temporary data, such as
operating system swap space and temp file space. This feature provides an option for
you to simultaneously use the card as a caching device and as a storage device for
temporary data.
When this functionality is used, the same Flash capacity and PCIe resources are
shared between the cache and DAS portions. Therefore, the cache performance may
be less compared to when the PCIe card is being used solely as a caching solution.


21 Introduction to EMC VFCache
VFCache management
VFCache does not require sophisticated management software. However, there is a
CLI for management of the product. There is also an option of using a VSI plug-in for
VFCache management in VMware environments.


22 Introduction to EMC VFCache
Performance Considerations
VFCache is a write-through caching product rather than a Flash storage solution, so
there are certain things that need to be considered when evaluating VFCache
performance.
Locality of reference
The key to maximizing VFCache performance is the locality of reference in the
application workload. Applications that reference a small area of storage with very
high frequency will benefit the most from using VFCache. Examples of this are
database indices and reference tables. If the locality of reference is low, the
application may get less benefit after promoting a data chunk into VFCache. Very low
locality will result in few or no promotions and thus no benefit.
Warm-up time
VFCache needs some warm-up time before it shows significant performance
improvement. Warm-up time consists mostly of promotion operations into VFCache.
This happens when the VFCache has just been installed and is empty. This also
happens when the working data set of the application has changed dramatically, and
the current VFCache data is no longer being referenced. During this phase, the
VFCache read-hit rate is low, so the response time is more like that of the SAN
storage. As the VFCache hit rate increases, the performance starts improving and
stabilizes when a large part of the application working set has been promoted into
VFCache. In internal tests using a 1.2 TB Oracle database, the throughput increased
to more than twice the baseline values in 30 minutes when TPC-C-like workload was
used.
Among other things, warm-up time depends on the number and type of storage media
in the back-end SAN storage. For example, a setup of 80 SAS drives will have a
shorter warm-up time than a setup with 20 SAS drives. Similarly, a setup with SAS
hard-disk drives (HDDs) in the back end will warm up faster than with NL-SAS HDDs in
the back end. This is because NL-SAS drives typically have a higher response time
than SAS drives. When you are designing application layouts, it is important to
remember that there is a warm-up time before stable VFCache performance is
reached.
In a demo or a Proof of Concept, the warm-up time can be speeded up by reading
sequentially through the test area in 64 KB I/O size. Once the working set has been
promoted, the benchmark test can be run again to compare the numbers with the
baseline numbers. CLI commands can be used to find out how many pages have been
promoted into VFCache. This gives you an idea of what percentage of the working set
has been promoted into the cache.
If you are comparing the performance against PCIe Flash DAS solutions, the initial
performance numbers of VFCache will be less because the cache needs to warm up
before the stable performance numbers are shown. In the case of DAS solutions, all
read and write operations happen from the PCIe Flash and there is no warm-up phase.

23 Introduction to EMC VFCache
Therefore, initial performance numbers should not be compared between a caching
and a DAS solution.
Workload characteristics
The final performance benefit that you can expect from VFCache depends on the
application workload characteristics. EMC recommends that you do not enable
VFCache for storage volumes that do not have a suitable workload profile. This
enables you to have more caching resources available for those volumes that are a
good fit for VFCache. For example:
Read/write ratio
VFCache provides read acceleration, so the higher the read/write ratio of the
workload, the more performance benefit you get.
Working set size
You should have an idea of the working set size of the application relative to the
cache size. If the working set is smaller than the cache size, the whole working set
will get promoted into the cache and you will see very good performance
numbers. However, if the working set is much bigger than the cache, the
performance benefit will be less. The maximum performance benefit is for those
workloads where the same data is read multiple times or where the application
reads the same data multiple times after writing it once.
Random versus sequential workloads
An EMC storage array is very efficient in processing sequential workloads from
your applications. The storage array uses its own cache and other mechanisms
like prefetching to accomplish this. However, if there is any randomness in the
workload pattern, the performance is lower because of the seek times involved
with accessing data on mechanical drives. The storage array cache is also of
limited use in this case because different applications using the storage array will
compete for the same storage array cache resource. Flash technology does not
have any latency associated with seek times to access the data. VFCache will
therefore show maximum performance difference when the application workload
has a high degree of random component.
Concurrency
Mechanical drives in the storage array have only one or two read/write heads,
which means that only limited number I/Os can be processed at any one point in
time from one disk. So when there are multiple threads in the application trying to
access data from the storage array, response times tend to go up because the
I/Os need to wait in the queue before they are processed. However, storage and
caching devices using Flash technology typically have multiple channels
internally that can process multiple I/Os at the same time. Therefore, VFCache
shows the maximum performance difference when the application workload has a
high degree of concurrency. The application should request multiple I/Os at the
same time.

24 Introduction to EMC VFCache
I/O Size
Large I/O sizes tend to be bandwidth-driven and reduce the performance gap
between Flash technology and non-Flash technologies. Applications with smaller
I/O sizes (for example, 8 KB) show the maximum performance benefit when using
VFCache.
Throughput versus latency
There are some applications that can push the storage environment to the limit to
provide as many IOPS as possible. Using VFCache in those application environments
will show very high IOPS at very low response times. However, there are also
applications that do not require very high IOPS, but they require very low response
times.
You can see the benefit of using VFCache in these application environments. Even
though the application issues relatively few I/Os, whenever the I/Os are issued, they
will be serviced with a very low response time. For example, a web application may
not have a lot of activity in general, but whenever a user issues a request, the
response will be very quick.
Other bottlenecks in the environment
VFCache helps improve throughput and reduce latencies for the applications.
However, any drastic improvement in application throughput may expose new
underlying performance bottlenecks and/or anomalies. Addressing these may include
application tuning, such as increasing buffer cache sizes or other changes that
increase concurrency. For example, in a typical customer deployment, a Microsoft SQL
Server administrator should not enable VFCache on the log files. An inefficient
storage layout design of the log files may be exposed as a bottleneck when VFCache
improves the throughput and latency of the SQL Server database.
Write performance dependent on back-end array
VFCache provides acceleration to read I/Os from the application. Any write operations
that the application issues still happens at the best speed that the back-end storage
array can provide. At a fixed read/write ratio from an application, this tends to limit
the net potential increase in read throughput. For example, if the storage array is
overloaded and is processing write operations at a very slow rate, VFCache will not be
able to accelerate additional application reads.
Once VFCache has been enabled on a particular source volume, every I/O from the
application needs to access the VFCache card, whether it is a read or a write
operation. In most cases, the processing capability of VFCache will be much greater
than what the storage array can provide, therefore VFCache will not be a performance
bottleneck in the data path. However, if a very large number of disks on the storage
array are dedicated to a single host application, and they are fully utilized in terms of
IOPS, the throughput that the storage array could provide without VFCache might be
more than what VFCache can process. In this scenario, VFCache may provide minimal
performance benefit to the application.

25 Introduction to EMC VFCache
Usage guidelines and characteristics
This section provides some of the usage guidelines and salient features of VFCache.
Since VFCache does not store any data that has not already been written on
the storage array, the application data is protected and is persisted on the
storage array if anything happens to VFCache on the server. However, the
cache would need to be warmed up again after the server starts up.
In a physical environment, you can enable or disable VFCache at the source
volume level or LUN level. In a virtual environment, the VFCache capacity
needs to be partitioned for individual virtual machines, as applicable. This
allocated cache capacity inside the virtual machine can then be configured at
vDisk-level granularity. The minimum size for the cache vDisk is 20 GB.
There is no hard limit on the maximum number of server volumes on which
VFCache can be enabled. However, if you enable it on a very large number of
volumes, that may create resource starvation for those volumes that could
actually benefit from VFCache. EMC recommends that VFCache not be enabled
for those volumes that are least likely to gain any performance benefit from
VFCache. This allows other volumes that are a good fit for VFCache to get the
maximum processing and cache capacity resources.
PowerPath optimizes the use of multiple data paths between supported
servers and storage systems, providing a performance boost by doing load
balancing between the paths. VFCache improves the application performance
even further by helping to move the most frequently accessed data closer to
the application by using PCIe Flash technology for write-through caching.
PowerPath and VFCache are complementary EMC products for scaling mission-
critical applications in virtual and physical environments, including cloud
deployments. Additionally, since VFCache sits above the multipathing
software in the I/O stack, it can work with any multipathing solution on the
market. PowerPath and VFCache are purchased separately.
VFCache is complementary to FAST VP and FAST Cache features on the storage
array. However, it is not required to have FAST VP or FAST Cache on the storage
array to use VFCache.
VFCache only accelerates read operations from the application. The write
operations will be limited by the speed with which the back-end array can
process the writes.
Specifications
The cache page size that is used internally in VFCache is 8 KB, but it will work
seamlessly with applications where the predominant I/O size is other than 8
KB. The cache page size is fixed and is not customizable.
VFCache needs to be installed in PCIe Gen2, x8 slots in the server. It can be
installed in x16 PCIe slots also, but only 8 channels will be used by VFCache.

26 Introduction to EMC VFCache
Similarly, if it is installed in an x4 PCIe slot in the server, VFCache will perform
sub-optimally.
VFCache cards are available in 300 GB capacity.
Only one VFCache card can be used per server.
VFCache supports the following connection protocols between the server and
the storage array:
o 4 Gb/s Fibre Channel
o 8 Gb/s Fibre Channel
VFCache is compliant with the Trade Agreements Act (TAA). The following main
requirements are certified as not applicable to VFCache:
o FIPS 140-2
o Common Criteria
o Platform Hardening
o Research Remote Access
Constraints
VFCache does not provide connectivity between the server and the SAN storage
array. You still need to use an HBA card to connect to the back-end storage array
where the data is eventually stored.
VFCache is not supported on blade servers. Blade servers require a customized
version of the card. For the most current list of supported operating systems and
servers, refer to E-Lab Interoperability Navigator.
VFCache is currently not supported in shared-disk environments or active/active
clusters. However, shared disk clusters in VMware environments are supported
since VFCache is implemented at the virtual machine level rather than the ESX
server level.
By default, there is a maximum I/O size of 64 KB, which VFCache driver intercepts.
Any I/O larger than 64 KB will not be intercepted by VFCache. Application with
larger I/O sizes are typically bandwidth sensitive and have sequential workloads,
which would not benefit from a caching solution like VFCache.
Stale data
Stale data because of storage array snapshots
If any operations modify the application data without the knowledge of the server,
it is possible to have stale data in VFCache. For example, if a LUN snapshot were
taken on the array and later used to roll back changes on the source LUN, the
server would have no knowledge of any changes that had been done on the array,
which would result in VFCache having stale data that had not been updated with
the contents from the snapshot. As a workaround in this case, you need to
manually stop and restart the VFCache software driver for the source volume.

27 Introduction to EMC VFCache
Note The whole cache device does not need to be stopped, only the source
volume on which the snapshot operations are being done needs to be
stopped. When you restart the VFCache software driver on the source
volume, a new source ID is automatically generated for that source
volume, which invalidates the old VFCache contents for the source volume
and starts caching the new data from the snapshot. The application then
gets access to new data from the snapshot.
Stale data in VMware environments
If you use VMware, you should be careful when the VMware snapshot feature is
being used. VFCache metadata is kept in the virtual machine memory, therefore, it
will be a part of the virtual machine snapshot image when a virtual machine
snapshot is taken. This means that when this snapshot image is used to roll back
the virtual machine, the old metadata is restored and potentially causes data
corruption.
You must purge the VFCache before the virtual machine suspend and resume
operations. This is handled using scripts that are automatically installed when the
VFCache Agent is installed in the virtual machine. These scripts are automatically
invoked when these virtual machine operations are run. In Windows
environments, you should take care to ensure that other programs or installations
in the virtual machine do not change the default suspend/resume scripts in such
a way that the VFCache scripts are not executed on those events. VFCache can
also be purged manually before suspend and resume operations in the virtual
machine, if needed.


28 Introduction to EMC VFCache
Application use case and performance
VFCache helps you boost the performance of your latency and response-time
sensitive applications typically applications such as database applications (like
Oracle, SQL Server and DB2), OLTP applications, web applications, and financial
trading applications. VFCache is not suitable for more write-intensive or sequential
applications such as data warehousing, streaming media, or Big Data applications.
Use cases are shown in Figure 9.

Figure 9: VFCache Use Cases
The horizontal axis represents a typical read/write ratio of an application workload.
The left side represents write-heavy applications such as backups. The right side
represents read-heavy applications such as reporting tools.
The vertical axis represents the locality of reference or skew of the applications
workload. The lower end represents applications that have very low locality of
reference, and the top side represents applications where a majority of the I/Os go to
a very small set of data.
You will achieve the greatest results with VFCache in high-read applications and
applications with a highly concentrated skew of data.


29 Introduction to EMC VFCache
Test results
EMC conducted application-specific tests with VFCache to determine potential
performance benefits when this product is used. Here is a summary of the VFCache
benefits with a couple of applications:
SQL Server
With a TPC-E like workload in a 750 GB Microsoft SQL Server 2008 R2
environment, the number of transactions increased three times and the latency
was reduced by 87 percent when VFCache was introduced in the configuration.
Oracle
With a TPC-C-like workload in a 1.2 TB Oracle 11g R2 physical environment,
the number of transactions increased three times and the latency was reduced
by 50 percent when VFCache was introduced in the configuration. The test
workload had 70 percent reads and 30 percent writes.
In a VMware setup with 1.2 TB Oracle Database 11g R2 and TPC-C-like
workload, the number of transactions increased by 80 percent when VFCache
was introduced in the configuration. The test workload had 70 percent reads
and 30 percent writes.
For more information on application-specific guidelines and test results, refer to the
list of white papers provided in the References section.


30 Introduction to EMC VFCache
Conclusion
There are multiple ways in which Flash technology can be used in a customer
environment today, for example, Flash in the server or the storage array, Flash used
as a cache or a tier. The key, however, is the software that brings all of this together,
using different technologies at the right place and time for the right price.
VFCache uses EMC FAST technology in the storage array and FAST in the server to
provide this benefit most appropriately, as simply and as easily as possible.
VFCache dramatically accelerates the performance of read-intensive applications.
VFCache software caches the most frequently used data on the server-based PCIe
card, which puts the data closer to the application. It extends FAST technology
into the server by ensuring that the right data is placed in the right storage at the
right time.
The intelligent caching algorithms in VFCache promote the most frequently
referenced data into the PCIe server Flash to provide the best possible
performance and latency to the application.
VFCache provides you with the flexibility to use the same PCIe device as a caching
solution as well as a storage solution for temporary data.
VFCache suits many but not all customer environments, and it is important that you
understand the application workload characteristics properly when choosing and
using VFCache.
VFCache protects data by using a write-through algorithm, which means that writes
persist to the back-end storage array. While other vendors promise the performance
of PCIe Flash technology, EMC VFCache provides this performance with protection.



31 Introduction to EMC VFCache
References
The following documents are available on Powerlink:
EMC VFCache Data sheet
VFCache Installation and Administration Guide for Windows and Linux
VFCache Release Notes for Windows and Linux
VFCache Installation Guide for VMware
VFCache Release Notes for VMware
VFCache VMware Plug-in Administration Guide
Considerations for Choosing SLC versus MLC Flash
EMC VFCache Accelerates Oracle - EMC VFCache, EMC Symmetrix VMAX and
VMAXe, Oracle Database 11g
EMC VFCache Accelerates Virtualized Oracle - EMC VFCache, EMC Symmetrix VMAX
and VMAXe, VMware vSphere, Oracle Database 11g
EMC VFCache Accelerates Oracle - EMC VFCache, EMC VNX, EMC FAST Suite,
Oracle Database 11g
EMC VFCache Accelerates Microsoft SQL Server - EMC VFCache, EMC VNX,
Microsoft SQL Server 2008
The following Demartek analyst report and video are available on EMC.com:
EMC VFCache Flash Caching Solution Evaluation