0
Q
Assigned a Logical Unit Number (LUN) and are presented to a host, or hosts, as a physical device.
A
Logical disks
1
Q
Utilizes an Intelligent Storage System to group disks together and then partitions those physical disks into discrete logical disks.
A
Classic Storage Model
2
Q
How can RAID be applied?
A
At the physical disk layer or the logical disk layer
3
Q
4 Limitations of the Classic Storage Model
A
- There is a maximum number of physical disks that can be combined to form a logical disk.
- To expand a logical disk beyond this limit, a MetaLUN (or Meta) must be formed, adding another layer of management.
- Expanding the MetaLUN could be difficult and/or time consuming,
- The amount of storage provisioned in the Classic model is often greater than what is actually needed.
4
Q
Factors limiting the maximum number of physical disks that can be combined to form a logical disk
A
- Array limitation
2. Practical limitation
5
Q
Example of a practical limitation on the maximum number of physical disks that can be combined to form a logical disk
A
The chance of multiple disk failures increases as the number of disks increases, and thus increasing the chance of data loss or corruption.
6
Q
How is a logical disk expanded beyond the limit of the maximum number of physical disks?
A
To expand a logical disk beyond this limit, multiple logical disks are created and then chained together, forming what is often referred to as a MetaLUN (or Meta).
7
Q
Downside to MetaLUNs
A
Add another layer of management
8
Q
Multiple logical disks chained together
A
MetaLUN (or Meta)
9
Q
In the Classic Storage model, why is the amount of storage provisioned often greater than what is actually needed?
A
This is done to prevent application downtime to expand the storage, or due to unknown growth patterns.
10
Q
In the Classic Storage model, results in storage sitting idle for some time.
A
Overprovisioning of storage
11
Q
Virtual representations of portions of physical disks within a storage system
A
LUNs (or devices)
12
Q
An aggregation of a portion of physical disks that are presented to a host as a physical disk
A
LUN
13
Q
Advantage of a LUN
A
Instead of writing data to a single disk, it is spread over the disks that comprise the LUN.
This provides the ability to provide fault tolerance and performance improvements without having to deal with complex configurations at the host level.
14
Q
2 Basic Categories of LUNs
A
- Traditional LUNs
2. Pool LUNs
15
Q
Traditional LUNs offer what?
A
Fixed performance based on the RAID and disk type.
16
Q
Good choice for storage that does not have large growth
A
Traditional LUNs
17
Q
LUN type that has been standard for many years
A
Traditional LUN
18
Q
Traditional LUN Configuration
A
When a traditional LUN is created, the number of disks that are utilized to create the LUN correspond to the desired RAID type LUN.
19
Q
Type of LUN which utilizes a larger grouping of disks to create LUNs
A
Pool LUNs
20
Q
Pool LUN Configuration
A
While the physical disks that comprise the pool members are configured with a RAID mechanism, when the LUN is created using a Pool, the LUN is built across all of the disks in the pool.
21
Q
LUN type where the LUN is built across all of the disks in the pool.
A
Pool LUN
22
Q
Allows the creation of larger LUNs without sacrificing availability
A
Pool LUN
23
Q
Downside of Pool LUN approach
A
It introduces some variability in performance - it is more likely that a larger number of applications will share storage in a pool.
24
2 Type of Pool LUNs
1. Thin
| 2. Thick
25
Another name for Virtual Provisioning
Thin Provisioning
26
Another name for Thin Provisioning
Virtual Provisioning
27
Ability to present a LUN to a compute system with more capacity than what is physically allocated to the LUN.
Virtual Provisioning (Thin Provisioning)
28
Advantage of Virtual Provisioning (Thin Provisioning)
Capacity-on-demand from a shared storage pool, called the thin pool.
29
Characteristics of Virtual Provisioning (Thin Provisioning)
Capacity-on-demand from a shared storage pool, called thin pool.
Physical storage is allocated only when the compute requires it.
Provisioning decisions not bound by currently available storage.
30
Layers where Virtual Provisioning (Thin Provisioning) can be implemented
```
Storage Layer
Compute Layer (virtual provisioning for virtual disk)
```
31
How do administrators typically allocate storage space?
Based on anticipated storage growth.
32
Overprovisioning of storage capacity leads to what?
Higher costs
Increased power and cooling
Increased floor space requirements
Lower capacity utilization
33
One of the biggest challenges for storage administrators
Balancing the storage space required by various applications in their data centers
34
Ability to present a logical unit (Thin LUN) to a compute system, with more capacity that what is physically allocated to the LUN on the storage array.
Virtual Provisioning
35
How is physical storage allocated in a virtual provisioning (thin provisioning) environment?
Physical storage is allocated to the application "on-demand" from a shared pool of physical capacity.
36
Provides more efficient utilization of storage by reducing the amount of allocated, but unused physical storage.
Allocating physical storage to the application "on-demand" from a shared pool of physical capacity. (Virtual Provisioning / This Provisioning)
37
Other capability of Thin Provisioning
Ability to overallocate resources
38
Example of ability to overallocate resources
A pool that has 100 GB of physical capacity can actually virtual LUNs with a higher capacity - 500 GB for example.
As capacity is used by the hosts, the available capacity of the pool will diminsh.
The administrator will need to add physical storage to the pool before the 100 GB is exhausted, or the VMs and hypervisors will experience errors when trying to write data.
39
What happens if the physical storage pool is exhausted?
The VMs and hypervisors will experience error when trying to write data.
40
Establishes a hierarchy of storage types, and identifies the candidate data to relocate them to the appropriate storage type to meet service level requirements at a minimal cost.
Storage Tiering
41
Each storage tier is optimized for what?
A specific characteristic, such as: performance, availability, or cost
42
Efficient storage tiering requires implementation of what?
Policies
43
Storage tiering policies may be based on what parameters?
File type, frequency of access, etc.
44
Two types of storage tiering implementations
1. Manual Storage Tiering
| 2. Automated Storage Tiering
45
Not a cost efficient solution for growing data storage needs
Buying more high-end storage
46
Reason for storage tiering
Organizations require solutions that enable storing the right data, at the right cost, with the right access.
47
Helps identify active or inactive data to relocate it to an appropriate storage type
Storage tiering
48
May be configured as Tier 1 storage for frequently accessed data to improve performance
High performance FC drives
49
May be configured as Tier 2 storage for less frequently accessed data
Low cost SATA drives
50
Improves application performance
Moving the active data (frequently used data) to Solid-State Drives (SSDs) or FC.
51
Can free up storage capacity in high performance drives and reduce the cost of storage
Moving the inactive data (less frequently used data) to SATA.
52
Movement of data in a storage tiering environment happens is based on what?
Defined tiering policies
53
Tiering policy may be based on what parameters?
File type
Frequency of access
Performance
Etc.
54
Example of a storage tiering policy
"Move the files which are not accessed for the last 30 days to lower tier."
55
Traditional method of storage tiering
Manual storage tiering
56
Manual, repetitive and time consuming tiering
A traditional storage tiering process
57
Addresses the challenges presented by traditional storage tiering
Automated storage tiering
58
Automates the storage tiering process WITHIN an array
Automated Storage Tiering - Intra Array
59
Enables efficient use of SSDs and Near-Line Serial Attached SCSI (NL SAS) drive technologies
Automated Storage Tiering - Intra Array
60
Performs data movements between tiers at the sub-LUN level
Automated Storage Tiering - Intra Array
61
Employs cache tiering to improve application performance further
Automated Storage Tiering - Intra Array
62
How does Automated Storage Tiering - Intra Array work?
1. Moves active data to high performance SSD tier.
| 2. Moves inactive data to higher capacity lower performance NL SAS drives tier.
63
Three types of storage in an array used in Automated Storage Tiering - Intra Array
1. SSD
2. SAS
3. NL SAS
64
In an Automated Storage Tiering - Intra Array environment, how are data movements between tiers handled?
Data movements executed between tiers can be performed at the sub-LUN level.
65
Creates a large capacity secondary cache using SSDs
Cache Tiering
66
Enables tiering between DRAM cache, SSD drives (secondary cache).
Cache Tiering
67
How are most reads handled in a cache tiering environment?
Most reads are now served directly from high performance tiered cache.
68
Two benefits of cache tiering
1. Provides excellent performance benefit during peak workload.
2. Non-disruptive and transparent to applications.
69
Benefit of a large cache in a storage array.
Improves performance by retaining frequently accessed data for a long period of time.
70
Drawback to a large cache in a storage array.
Involves more cost
71
Alternative way to increase the cache size
Utilize the existing SSDs on the storage array to extend the size of the cache configured in the array.
72
Uses SSDs to creat a large capacity secondary cache
Cache Tiering
73
Enables tiering between DRAM (primary cache) and SSDs (secondary cache).
Cache Tiering
74
Also enables the array to store large volumes of frequently accessed data on the cache tier.
Cache Tiering
75
How does cache tiering provide excellent performance during peak workloads?
Most reads are now served directly from the cache tier.
76
Facilitates policy-based data movements between tiers in an Inter-Array Storage Tiering environment.
Policy Engine
77
Automates the identification of active or inactive data to relocate them to different performance / capacity tiers BETWEEN the arrays.
Inter-Array Storage Tiering
78
Facilitates moving inactive or infrequently accessed data from the primary to secondary storage.
Policy engine
79
Some of the prevalent reasons to tier data across arrays
Archival or compliance requirements
80
What does the policy engine do for each file it archives?
For each file it archives, the policy engine leaves behind a small space-saving stub file that points to the real data on the primary storage.
81
Creates an abstraction layer in the SAN, between physical storage resources and virtual volumes presented to the compute systems.
Block-level Storage Virtualization
82
Uses a virtualization appliance to perform mapping operation between physical storage resources and volumes presented to compute
Block-level Storage Virtualization
83
Makes underlying storage infrastructure transparent to compute
Block-level Storage Virtualization
84
What happens with Block-level Storage Virtualization?
Instead of being directed to the LUNs on the individual storage arrays, the compute systems are directed to the virtual volumes on the virtualization appliance.
85
Performs a mapping between the virtual volumes and the LUNs on the arrays.
Virtualization appliance
86
Enables us to combine several LUNs from one or more arrays into a a single virtual volume before presenting it to the compute systems.
Block-level Storage Virtualization
87
Can take a single large LUN from an array, slice it into smaller volumes, and present these volumes to the compute systems.
Block-level Storage Virtualization
88
Supports the dynamic increase of storage volumes.
Block-level Storage Virtualization
89
Supports the consolidation of heterogeneous storage arrays.
Block-level Storage Virtualization
90
Supports transparent volume access
Block-level Storage Virtualization
91
Three things supported by Block-level Storage Virtualization
1. Dynamic increase of storage volumes.
2. Consolidation of heterogeneous storage arrays.
3. Transparent volume access.
92
Enables storage volumes to remain online and accessible while data is being migrated.
Block-level Storage Virtualization
93
With a Block-level Storage Virtualization solution in place, what can handle the migration of data?
The virtualization appliance
94
True or False: In a Block-level Storage Virtualization environment, after data is migrated from one array to another, no physical changes are required because the compute system still points to the same location on the virtualization appliance.
True
95
True or False: In a Block-level Storage Virtualization environment, after data is migrated, the mapping on the virtualization appliance is changed to point to the new location (array).
True
96
True or False: In a Block-level Storage Virtualization environment, mappings on the virtualization appliance can be changed online with no impact to the end user data access.
True
97
Benefits of an Information Lifecycle Management (ILM) strategy
Significant cost and resource optimization
98
Facilitates an Information Lifecycle Management (ILM) strategy, enabling significant cost and resource optimization.
Deploying Block-level Storage Virtualization in a heterogeneous array environment
99
True or False: In a Block-level Storage Virtualization environment, low-value data can be migrated from higher performance to appropriate performance arrays or disks.
True
100
Aside from the virtualization appliance, what devices can also perform in the virtualization function in an Block-level Storage Virtualization environment?
Some storage arrays can also perform the virtualization function.
101
Provides an abstraction in the NAS / File server environment.
File Level Storage Virtualization
102
Enables movement of files between NAS systems without impacting client access.
File Level Storage Virtualization
103
Provides opportunities to optimize storage utilization
File Level Storage Virtualization
104
Storage virtualization method which is implemented using global namespace.
File Level Storage Virtualization
105
Four characteristics of File Level Storage Virtualization
1. Provides an abstraction in the NAS / File server environment.
2. Enables movement of files between NAS systems without impacting client access.
3. Provides opportunities to optimize storage utilization.
4. Implement using global namespace.
106
Eliminate dependencies between the file and its physical location.
File Level Storage Virtualization
107
Before File Level Storage Virtualization, each client had to know what?
The exact location of its file-level resources
108
True or False: In a data center, migrating data from a NAS to another NAS may be required for technology refresh, performance requirements, and non-availability of additional storage capacity.
True
109
Simplifies file mobility in a NAS / file server environment
File Level Storage Virtualization
110
True or False: It is not easy to move files in a NAS environment without File Level Storage Virtualization, and requires downtime for NAS systems.
True
111
Creates a logical pool of storage and enables users to use a logical path, rather than a physical path, to access files.
File virtualization appliance in the network
112
Facilitates the movement of files between NAS systems without any downtime.
File Level Storage Virtualization
113
True or False: In a File Level Storage Virtualization environment, clients can access their files non-disruptively while the files are being migrated.
True
114
In a File Level Storage Virtualization, what is used to map the logical path of a file to the physical path names?
Global namespace
115
First items to consider when evaluating storage options
The hypervisor and, by default, application requirements
116
Six hypervisor considerations when evaluating storage design options
1. Whether or not to use shared storage
2. Whether or not to use live migration over distance
3. Whether or not to use stretched clusters
4. Which to use: block or file storage
5. Which storage protocols to use
6. What are the BC/DR requirements
117
Refers to expanding a storage device by adding more capacity
Scaling-up
118
Increased storage capacity, in addition to more disk capacity can also be what?
1. Larger or faster drives: increases storage capacity or IOPs.
2. Additional cache: increases IOPs.
3. More storage controllers: increases throughput and IOPs.
4. More or faster front-end (host) ports: increases throughput.
119
Brings additional set of management challenges
Expanding beyond a single storage devices to another storage device
120
When does it become necessary to scale beyond a single device?
When 1) a single storage device cannot scale to the needed capacity (storage, IOPs, throughput, etc.) and 2) when a device comes in a fixed size.
121
Scale-up hardware scaling
Adding more components to existing storage system
122
Scale-up hardware limits
Expand to single storage system limits
123
Scale-up availability, resiliency
Protects against component failures
124
Scale-up storage management complexity
Single storage system to manage
125
Scale-out hardware scaling
Add more storage systems
126
Scale-out hardware limits
Expand to aggregated storage system limits
127
Scale-out availability, resiliency
Protects against storage system failure
128
Scale-out storage management complexity
Multiple storage systems to manage
129
Challenges in large hypervisor deployments
1. Multiple NAS systems required to store data
2. Multiple management systems
3. Multiple file systems
4. Disjointed file namespaces
5. Replicating data across systems to protect against failure
130
Examples of single storage device limits in a large hypervisor deployment
A single storage device may not be able to handle:
1. volume of data (capacity),
2. amount of data being read from or written to storage (throughput),
3. number of transactions (IOPs) that is required
131
True or False: If the storage system is not designed to be a scale-out system, you will be faced with the problem of having multiple disparate file systems, which are only accessed from a single storage system.
True
132
Problems associated with having multiple disparate file systems which are only accessed from a single storage system
Heavily utilized file systems will 1) create bottlenecks on that specific storage system, 2) will eventually lead to having to manually redistribute other file systems to different storage, or 3) fragmenting the offending file system into multiple, smaller files systems to be distributed.
133
Result of file systems dispersed across multiple storage systems
The hypervisors will have to access multiple namespaces to mount file systems across these systems.
134
True or False: There may also be multiple namespaces on a single storage system, depending upon the networking capabilities and how the NAS heads are integrated.
True
135
Required to ensure that data is protected against a storage system failure.
Data must be replicated to another NAS in the environment
136
Required if data is duplicated to another NAS in the environment
Either 1) having a duplicate set of NAS systems that sit idle until needed, or 2) replicating between primary systems (which reduces the overall capacity of each system).
137
Similarity of EMC Isilon to traditional NAS solutions
Provides CIFS and NFS connectivity
138
Scale-out of EMC Isilon
Can grow from 3 to 144 nodes in a single cluster.
139
Scale-up capacity of EMC Isilon system in terms of storage
Up to 15.5 Petabytes (PB) of storage
140
Scale-up capacity of EMC Isilon system in terms of throughput
Up to 85 GB/s throughput
141
Required to scale up an EMC Isilon system
Adding the appropriate number of nodes to the environment
142
Each EMC Isilon node has its own what?
1. Processors
2. Memory
3. Storage
4. Network connectivity
143
True or False: There are different types of EMC Isilon nodes available, depending on your requirements.
True
144
How is EMC Isilon unique?
EMC Isilon is unique because:
1. It provides a single point of mgmt for the entire system, regardless of the number of nodes installed; and
2. All of the EMC Isilon nodes contribute to a single file system with a single namespace.
145
Advantages of EMC Isilon's single file system with a single namespace
All hypervisors can be configured identically, with a single mount point to access all relevant data.
There is no need to manually rebalance workloads or redistribute file systems across nodes.
147
How does EMC Isilon provide high availability?
Depending upon configuration, data within an Isilon cluster is distributed across disks within a node, and/or across nodes.
The result is a highly available system that can withstand multiple failures.
Data can also be replicated to another Isilon system for additional protection.
148
3 management tools considerations in a multi-vendor scenario
1. Dp any of the vendors have a solution to manage multiple vendors' systems?
2. Is there a third-party option?
3. Are there APIs that can be utilized to build a custom tool?
149
3 management tools considerations in a single-vendor, multiple products scenario
1. Is there a singe management offering for multiple products?
2. Is there a third-party option?
3. Are there APIs that can be utilized to build a custom tool?
150
2 management tools considerations in a scale-out environment
1. Are the "nodes" managed as a single device or multiple devices?
2. Is information shared across nodes?
151
True or False: Multiple products from the same vendor can be just as complex to manage as a multi-vendor environment.
True
152
Designed to allow you to add processing nodes to an existing system
Scale-out environment
153
2 scale-out design options
1. Have the nodes appear as a single system, and configuration changes are made globally across all nodes.
2. Have a group of independent nodes that must be managed as separate entities.
154
2 concerns when deploying a storage virtualization product
1. Scalability
| 2. Management of multiple virtualization solutions
155
Storage Virtualization Performance Limit Questions
1. Can a single device support back-end storage?
| 2. Is performance aggregated for multiple devices?
156
2 Storage Virtualization Management Questions
1. Integrated management with back-end storage?
| 2. If you deploy multiple systems, are they managed as a single device or independent?
157
Can provide mechanisms to improve performance, manageability and several other advantages
Integration between storage and the other layers of the VDC
158
Products with allow a server administrator to provision storage for their hypervisor environment by using pre-defined pools
1. EMC Virtual Storage Integrator (VSI)
| 2. EMC Storage Integrator (ESI)
159
Advantage of using EMC Virtual Storage Integrator or EMC Storage Integrator
Frees the Storage Administrator from having to deal with the day-to-day provisioning tasks, and lets him/her focus on more strategic tasks.
160
Allows you to centralize the write splitters onto intelligent SAN switches (Cisco MDS 9500 series or 9222i, or the Brocade AP7600)
Using EMC RecoverPoint for data replication
161
Switches that have software and hardware that allows them to interact with the EMC RecoverPoint appliances, clone the data being written to storage and send the copy to the RPA to be sent to the destination
Cisco MDS 9500 series
Cisco 9222i
Brocade AP7600
162
In large environments, with numerous hosts or storage arrays or with different types of hosts or storage arrays, what provides the option to not have to manage the splitter across a number of different platforms?
Intelligent Fabric Splitter
163
Where are you likely to encounter a requirement to keep one tenant's information segregated from another?
Multi-tenant environments
164
Four options for isolating tenants or applications, in order of increasing levels of isolation
1. Shared Pool
2. Separate Pools
3. Separate Ports
4. Separate Everything
165
Provides a logical separation of data at all levels.
Shared Pool
166
Provides physical separation of data on the underlying disks
Separate Pools
167
Provides physical separation of data on the underlying disks AND isolates ports to specific tenants.
Separate Ports
168
All components isolated
Separate Everything
169
Allocates a portion of cache for each LUN or group of LUNs, providing a separation of memory resources
Separate Everything
170
What is a downside of resource isolation?
It can significantly reduce the scalability of a storage system.
171
Can typically be isolated to be used by specific servers
Front-end (or host ports (including NICs))
172
Scenario where back-end ports can be more complex to design
Separate Ports
173
True or False: System cache is not shared in a Separate Ports environment
False.
| System cache is still shared.
174
Data is stored on the same physical disks, but within a different LUN or file system
Shared Pool
175
True or False: The storage system components, such as front-end / back-end ports and cache (as well as the physical disks) are shared in a Shared Pool environment.
True
176
Two jobs of QoS mechanisms
1. Enforce SLAs in multi-tenant environments
| 2. Mitigate DoS attacks.
177
In multi-tenant environments, where SLAs are being enforced, what tool can be used to enforce tenant resource limits?
QoS
178
If data has been encrypted at the storage device, how is the data being transmitted?
In clear, unencrypted form
179
If data has been encrypted at the storage device, would is required?
Protection (encryption) at the network level
180
Challenge that EMC Symmetrix VMAX D@RE addresses
1. Data stored in clear format on array.
| 2. Sensitive information at risk if drives are removed from the array or array is replaced.
181
D@RE
Data At Rest Encryption
182
Provides a mechanism to encrypt data BEFORE it is written to disk.
EMC Symmetrix VMAX Data @ Rest Encryption (D@RE)
183
In an EMC Symmetrix VMAX Data @ Rest Encryption (D@RE) environment, what generates the key used for encryption?
RSA Key Manager Appliance
184
Four Attributes of EMC Symmetrix VMAX D@RE Solutioni
1. Data encrypted before being written to storage.
2. Each disk encrypted independently.
3. RSA Key Manager utilized to generate and store key information.
4. Data encrypted at line rate.
185
In EMC Symmetric VMAX D@RE solution, where is encryption process performed?
On the VMAX engines
186
What does the encryption process (performed on the VMAX engines) require?
Requires different back-end ports than the standard ports.
187
What happens when encrypted data is read from the VMAX?
The data is unencrypted by the engine before being sent to the requesting host.
188
How does VMAX encryption differ from solutions where data is encrypted on a per-host or per-LUN basis?
The VMAX encrypts data at the drive level across the entire array and each drive is encrypted with a unique key.
189
What happens to the data on a drive removed from an encrypted Symmetrix VMAX D@RE array?
Renders the data unusable since the decryption keys are stored on the Key Manager.
190
Standard in most block and file storage systems
HA and failover capabilities
191
HA and Failover for Block Storage
Multiple processors
Active / Active or Active / Passive
Multipathing Software
192
HA and Failover for File Storage
Multiple NAS heads
Active / Active or Active / Passive
NIC aggregation
193
Active / Active Array
Active paths from all processors to all LUNs
194
Active / Passive Array
Active paths from one processor to any LUN
195
Active / Active NAS
Active paths from all NAS heads to all file systems
196
Active / Passive NAS
Active paths from one NAS head to any file system
197
How is HA achieved?
By having redundant components such as host ports, cache, back-end ports / disk paths, NICs, and even spare drives.
198
Function of redundant processors in block systems
For block systems, redundant processors on the system allow cache to be mirrored so that no data is lost in a failure scenario.
199
In the event of a processor or port failure, what allows I/Os to be moved to another available path in order to preserve access to the appropriate storage?
Use of multipathing software on the hypervisor.
200
For an active/passive storage system, what does failure of a processor often mean?
Momentary interruption in I/O while access to the LUN is transferred to the passive processor.
201
Typically not noticeable in most active/passive deployments, but applications that are sensitive to latency variations may have an adverse reaction.
Failure of a processor which results in a momentary interruption in I/O while access to the LUN is transferred to the passive processor.
202
What must be determined when planning for outages, regardless of active/active or active/passive processors?
Whether to preserve performance levels during an outage (i.e., in a dual-processor system, no processor should exceed 50% utilization under normal circumstances) or run in a degraded mode during an outage.
203
What are used to provide redundancy in file storage systems instead of multiple processors in block storage systems?
Multiple NAS heads
204
Two forms of file storage processor redundancy
Active / Active
| Active / Passive
205
Environment where multiple NAS heads can read and write to the same file system(s) simultaneously
Active / Active system
206
Advantage of an active / active NAS system
If a NAS head fails, access is still available through the remaining NAS head(s).
207
How does an active / passive NAS system work?
One NAS head owns a particular file system, and in the event of a failure, that file system is transferred to another NAS head.
208
Can vary widely and should be considered when evaluating NAS solutions.
The amount of time it takes to fail the resources over to the passive NAS head
209
Is multipathing software used in a NAS solution?
No - there is no storage-based equivalent in a NAS solution.
210
Can be used to protect against an interface, connection, or switch failure in a NAS environment.
Standard networking features such as link aggregation.
211
Used for redundancy in a block environment
Multipathing software
212
Two main categories from a disaster recovery and restart perspective
1. Replication
| 2. Backups
213
Supports disaster restart and provides the capability to have data online quickly at the recovery site.
Replication
214
Will impact the networking design and may also influence your choice of protocols to use for host to storage communications.
The protocols supported by a replication product, whether integrated with an array or external
215
Replication solution considerations
Protocols supported.
Integrated with array or external.
Support for different arrays / vendors.
Integration with Hypervisor (failover).
216
Offer disaster recovery, meaning that if there is a failure, data can be recovered from disk or tape using a backup infrastructure.
Backups
217
Backup Deduplication Considerations
Source or Target.
Scope (within client, across clients, across nodes).
Retention.
218
Backup Replication Consideration
Copy backups at primary site or recovery site
219
Can be a huge benefit in reducing redundant data, when evaluating backup solutions
Deduplication
220
Example of where data is obviously redundant
Just within the operating system of the VMs in an environment.
221
Essential in a VDC
Deduplication in any form
222
Scope of deduplication examples
1. Within a single client (within the files of a single VM).
2. Across all clients (VMs) managed by a single backup node / system.
3. Across all backup nodes / systems.
223
True or False: The longer you keep backup data, the more likely you are to deduplicate new data.
True
224
Will generally take longer as the network bandwidth is not as robust as within a site.
Backing up data directly to a remote site
225
Will require that data be recovered to the primary site in the event of a simple file / data restore.
Backing up data directly to a remote site
226
Faster, but does not good if the primary data center is offline and the data is unavailable to restore to the recovery site.
Backing up data to a local system
227
Allows you to back up data locally and then replicate the necessary portions to the recovery site, facilitating both every day data recovery and disaster recovery.
Replication mechanism
228
Five network requirements / considerations
1. Protocols
2. VLANs / VSANs
3. Bandwidth
4. Latency
5, Deduplication (source or target)
229
Protocols which will directly impact the LAN, and possibly the WAN, infrastructure
iSCSI, FCoE, NFS and CIFS
230
Protocol which requires the deployment of independent storage networks
Fibre Channel
231
Not only important for hosts to access storage systems, but for replication purposes - both within a site and between sites.
Bandwidth and latency
232
Can have a tremendous impact on bandwidth
Deduplication, specifically for backups
233
Critical for network design
Where the deduplication occurs
234
Will conserve storage capacity by removing redundant data
Any form of deduplication
235
Can significantly reduce the amount of data that traverses the storage network
Data deduplication at the host, or source
236
Scenario where only the unique data is sent across the network to the storage device
Data deduplication at the host, or source
237
Can also be performed for disk storage, but is typically performed at the storage device, so all data is sent across the network.
Deduplication
238
Block storage protocols
1. Fibre Channel
2. iSCSI
3. FCoE
239
Advantages of Fibre Channel
1. Established protocol
2. High performance
3. Requires separate, redundant infrastructure
240
Disadvantages of Fibre Channel
1. Cost of infrastructure
| 2. Different network management model
241
Advantages of iSCSI
1. Established protocol
2. Lower cost
3. Can use existing Ethernet networks
242
Disadvantages of iSCSI
1. Performance limited on 1 Gb Ethernet
| 2. Software initiator can place load on CPU
243
Advantages of FCoE
1. High performance
2. Combines FC and Ethernet management models
3. Reduces infrastructure components
244
Disadvantages of FCoE
1. Newer protocol
2. Requires 10 Gigabit CEE infrastructure
3. Distance limitations
245
An established block protocol that provides high performance, and typically runs at speeds of 4, 8, or 16 Gbps.
Fibre Channel
246
Provides a solid foundation for storage access, but requires a separate network infrastructure, which is typically deployed in a redundant configuration.
Fibre Channel
247
Adds cost and complexity to the overall management of a data center
Implementing two Fibre Channel networks
248
Established block protocol that can be implemented using existing network infrastructure, making it an inexpensive option for scalable storage.
iSCSI
249
Issues with iSCSI
1. Performance over 1 Gbps Ethernet is limited.
| 2. Using standard NICs can put a significant amount of overhead on the host CPU.
250
Improves iSCSI performance but increases cost of deployment, especially if a large number of hypervisors are involved.
Specialized hardware to offload processing from the CPU
251
Relatively new block protocol that brings together the flexibility of Ethernet and the reliability of Fibre Channel into a single network
Fibre Channel over Ethernet (FCoE)
252
Can reduce the number of I/O cards, cables and switches in a data center by up to 50%.
Fibre Channel over Ethernet (FCoE)
253
Requires non-standard Ethernet switches, negating the ability to reuse existing infrastructure
Fibre Channel over Ethernet (FCoE)
254
FCoE twinax copper distance limitation
10 meters
255
FCoE multimode fiber distance limitation
300 meters
256
FCoE single mode fiber distance limitation
40 kilometers
257
Three file storage protocols
1. NFS
2. CIFS
3. pNFS
258
Established file protocol that has been used in Linux and UNIX environments
NFS
259
Can be accessed by Linux and UNIX systems, allowing flexibility for data migration, recovery, etc.
NFS Shares
260
Can use existing Ethernet networks, and is routable, allowing it to be access from any location
NFS
261
Issue with NFS on a 1 Gigabit Ethernet network
Performance
262
Why might NFS not support all of the functionality of a hypervisor?
Because NFS is a file-level protocol
263
Features which may not be available when using NFS as a storage protocol
Virtual provisioning
Snapshhots
Etc.
264
pNFS
Parallel NFS
265
Same advantages and disadvantages as NFS, except that it is interoperable with Windows based systems
CIFS
266
Less widely adopted as a file level storage option than NFS
CIFS
267
Relatively new file-level protocol that has emerged to address some of the shortcomings with traditional NFS.
Parallel NFS (pNFS)
268
Advantages of pNFS
1. High performance
2. Uses file only, file and block or file, or file and object storage models
3. Creates virtual distributed file systems
269
Provides significant performance improvement over NFS, and can combine multiple storage types but has not been widely implemented on hypervisor platforms as of 2012.
pNFS
270
Three advantages of NFS
1. Established protocol
2. Interoperable with Linux / UNIX
3. Can use existing Ethernet networks
271
Two disadvantages of NFS
1. Performance limited on 1 Gb Ethernet
272
Three advantages of CIFS
1. Established protocol
2. Can use existing Ethernet networks
3. Interoperable with Windows
273
Three disadvantages of NFS
1. Performance limited on 1 Gb Ethernet
2. Limited hypervisor support
3. May limit hypervisor functionality
274
Three advantages of pNFS
1. High performance
2. Uses file only, file and block, or file and object storage models
3. Creates virtual distributed file systems
275
Two disadvantages of pNFS
1. Newer protocol
| 2. Not widely adopted.
276
3 questions to address when planning storage pools
1. What are the tradeoffs between single and multiple pools?
2. If using multiple pools, what criteria is used to define pools?
3. How does Storage Tiering impact this decision?
277
3 Advantages of a Single Storage Pool
1. Simplified management.
2. Simplified capacity planning.
3. Allows LUNs or file systems to be striped across more physical drives, improving performance.
278
2 Disadvantages of a Single Storage Pool
1. Greater risk of resource contention between high I/O systems.
2. Potential for non-optimal RAID types to be uses.
279
In a single pool environment, why is there a greater risk of resource contention between high I/O systems?
Because a larger number if hypervisors share the same underlying drives
280
Why is RAID type a factor in a single Storage Pool environment?
Makes it possible to have specific applications or systems using a non-optimal RAID type, thereby degrading performance
281
2 Advantages of Multiple Storage Pools
1. Can isolate workloads to specific disks to reduce contention.
2. Can optimize storage profile for application needs.
282
2 Disadvantages of Multiple Storage Pools
1. More complex for planning and management.
| 2. Fewer spindles to distribute workload.
283
How is the risk of resource contention minimized in a multiple storage pool environment?
The administrator isolates specific workloads to specific drives, reducing contention.
284
Why would general systems potentially experience worse performance in a multiple storage pool environment?
Because the pools are smaller in a multiple storage pool environment, general systems may experience worse performance than with a larger pool.
285
Criteria for Defining Multiple Pools
1. Drive Technology (Flash, SAS, Near Line SAS, Fibre Channel, SATA, etc.)
2. RAID Type (1,5,6,10,etc.)
3. Drive Technology and RAID Type
4. Tenant
5. Application (database, email, etc.)
6. Function (QA, Test, Dev, etc.)
286
Name 5 types of drive technology
1. Flash
2. SAS
3. Near Line SAS
4. Fibre Channel
5. SATA
287
Can alleviate resource contention in a single storage pool environment
Tiering
288
How is resource contention alleviated in a single storage pool environment?
By using the tiering solution to move frequently access data to a faster drive technology, or a different RAID type, or both.
289
Allows you to group applications or systems based on the application's I/O characteristics (reads vs. writes, random vs. sequential), but does not address the application's I/O intensity
Using RAID type (1, 5, 6, 10, etc.) as the criteria for pool definition in a multiple pool environment.
290
Multiple storage pool criteria that provides more granular control, but can significantly increase the number of pools, and create a very complex environment to manage
Combination of drive technology and RAID Type (e.g., RAID 10 Flash, RAID 1 SA, RAID 5 SAS, etc.)
291
Multiple storage pool criteria that allows you to place the data on the storage that most appropriately matches the I/O intensity of the application or system, but does not address the I/O characteristics of the application or system
Drive technology (Flash, SAS, Near Line SAS, Fibre Channel, SATA, etc.
292
Advantages of Tiering in a Multiple Storage Pool environment
1. Reduces the number of pools needed.
| 2. Eliminates the need to create pools based on disk type.
293
In a multiple storage environment, what allows you to mix different drive technologies and/or RAID types within a pool, reducing the number of pools required to support some of the configurations?
Tiering
294
3 factors to consider for virtual provisioning
1. What are the considerations for using Virtual Provisioning?
2. Do you deploy Virtual Provisioning at the hypervisor, on the storage, or both?
3. What are the concerns with deploying it on both?
295
3 Characteristics of Traditional Provisioning
1. Best / most predictable performance
2. Precise data placement
3. Less concerned about space efficiency
296
4 Considerations of Virtual Provisioning
1. Space efficiency is needed.
2. Minimal host impact.
3. Best energy and capital savings.
4. For applications where space consumption is difficult to forecast.
297
Type of LUN best suited for situations where space efficiency is paramount
Thin LUNs
298
Type of LUN best suited for situations when host disruption cannot be tolerated
Thin LUNs
299
Type of LUN best suited for applications that cannot tolerate performance variations
Traditional LUNs
300
Type of LUN best suited for situations that require the highest level of performance
Traditional LUNs
301
With this type of environment, space is typically not a concern, as capacity will likely sit idle because you have allocated more capacity than is needed initially
Traditional LUNs
302
Array capacity - Thick VM on Thick Storage
Fully allocated at creation
303
Array capacity - Thin VM on Thick Storage
Fully allocated at creation
304
Array capacity - Thick VM on Thin Storage
Partially allocated at creation
305
Array capacity - Thin VM on Thin Storage
Partially allocated at creation
306
Utilization - Thick VM on Thick Storage
Full amount for each VM
307
Utilization - Thin VM on Thick Storage
Needed capacity for each VM
308
Utilization - Thick VM on Thin Storage
Full amount for each VM
309
Thin VM on Thin Storage
Needed capacity for each VM
310
Overprovisioning - Thick VM on Thick Storage
None
311
Overprovisioning - Thin VM on Thick Storage
Hypervisor
312
Overprovisioning - Thick VM on Thin Storage
Array
313
Overprovisioning - Thin VM on Thin Storage
Hypervisor and Array
314
Latency Cause - Thick VM on Thick Storage
None
315
Latency Cause - Thin VM on Thick Storage
Initializing Blocks
316
Latency Cause - Thick VM on Thin Storage
Array allocates storage
317
Latency Cause - Thin VM on Thin Storage
Initializing blocks and array allocates storage
318
True or False: Some hypervisors can emulate the capabilities of a Thin LUN
True
319
In an environment that uses Thin provisioning on the storage, you will need to know what?
How the hypervisor is allocating the blocks for a Thick VM, to avoid pool exhaustion
320
Zero reclamation on Thin pools
Arrays will examine blocks within the thin pool and if they contain no data, will return them to the free capacity for the pool
321
Drawbacks to zero reclamation by arrays
Can lead to performance variability, since the storage capacity will have to again be allocated from the pool when the hypervisor attempts to use it.
Can also lead to capacity issues, since the hypervisor believes that is has allocated the full amount for the VM, but some of that space has been returned to the pool, and may be consumed by other hosts.
322
What happens if the hypervisor zeroes all of the blocks immediately upon creation?
You will use the full amount of the capacity from the Thin storage.
As multiple VMs are created on that same storage, the consumed storage will approach the allocated storage.
If the Thin storage is heavily oversubscribed, you may exhaust the pool.
323
3 Storage Tiering Questions
1. What are the considerations for using Intra-Array Tiering?
2. What about Inter-Array Tiering?
3. When should you use a Caching Tier?
324
Questions for Intra-Array Tiering
1. Tier everything on the array or certain applications?
2. Performance considerations?
3. Management overhead? Is it automated?
325
Questions for Inter-Array Tiering
1. What do you tier?
2. Performance considerations?
3. Single management interface or multiple?
4. How is data transmitted between systems?
5. How does this work with a multi-vendor environment?
326
Used to accelerate reads and writes for entire array
Caching Tier
327
Accommodates applications that have occasional bursts
Caching Tier
328
No real benefit for pools / LUNs that are using SSD as a dedicated tier
Caching Tier
329
May not be good candidates for tiering as data will always be active
Applications that are highly transactional, or that do not store long-term data
330
How is tiering automated?
Automated tiering is performed automatically on the array based on policies
331
Does not force you to allocate SSDs for a specific pool or LUN
Caching Tier
332
Uses a general pools of SSDs to transfer data into or out of the array cache more efficiently.
Caching Tier
333
3 Challenges Answered by EMC VFCache
1. Need for improved performance on applications - web and OLTP
2. Need to protect data from device failure
3. Need to synchronize data with SAN storage system
334
Application needs that may require the infrastructure to be scaled to meet the requirements
Higher IOPS or Reduced Latency
335
Designed to accelerate read requests
VFCache
336
How does VFCache improve write performance?
By offloading a number of read requests from the array, VFCache indirectly improves write performance
337
What does VFCache essentially do?
Essentially, VFCache extends FAST (Fully Automated Storage Tiering) to the server
338
Where is a VFCache caching card installed?
In the relevant vSphere servers
339
How does EMC VFCache work?
VFCache will fulfill read requests locally.
Data read from SAN storage is cached locally.
Data written to SAN storage is cached locally.
Caching can be enabled on a per-VM basis.
340
What happens when a read request is sent from a VFCache-enabled VM?
It passes to the hypervisor, which attempts to fulfill the request from the local VFCache.
If data is available, it is retrieved locally & returned to the VM.
If data is not in cache, read request is forward to the storage system (VMAX) and fulfilled from array cache or disk.
When data is returned to the vSphere server, it is passed to the VM and asynchronously written to the VFCache, where is can be used to fulfill a later request.
341
How are write operations handled in a VFCache-enabled environment?
Write operations are not cached but instead are written directly to the storage array, just as if the VFCache was not present.
However, the data that is written to the array is also written asynchronously to the VFCache and can be later used to fulfill a read request.
342
Advantage of VFCache write model
Data is always written to storage, and therefore is protected by the normal redundant components and RAID.
343
In a classic environment, provides a mechanism to very quickly (nearly instantaneously) restore data to a previous point in time configuration.
Using snapshots and clones
344
Provides an effective solution to counter data loss or corruption (but not for disaster recovery).
Snapshots and clones
345
For block level storage, snapshots and clones are taken at what level?
LUN level
346
For block level storage, what is an important implication of snapshots and clones?
Snapshots and clones are taken at a LUN level, and therefore are REVERTED at the LUN level.
Multiple VMs are often storing their virtual disks on the same LUN, so it is possible to inadvertently rollback a number of VMs during a recovery event, possibly causing other data loss.
347
Ultimate consideration if you plan to use snapshots or clones
Need to ensure that you do not inadvertently destroy valid data;
either by having a process in place or configuring the storage to operate with these restrictions.
348
BC/DR: Replication & Failover Considerations for Storage Design
1. Replicating all VMs/data to remote site or just some?
2. Performance at remote site identical to primary or degraded?
3. Same vendor's storage or different vendor?
4. Is there integration with the hypervisor?
