Super-Size Your VDI with Active Flash. Active flash is continually working to automatically optimize performance in your VDI environment by promoting pages requested in real time. VMworld 2015
2. VDI Needs
PERFORMANCE SCALABILITY
VDI systems
must quickly
add new users
AVAILABILITY
The more VDI
users, the more
downtime will cost
Peak time and
steady state
(Boot and log-in
storms)
500 800
1500 2000
SAVE
$$$$$$$$$$$$
3. Very Large VDI Environments
SCALE @ About 2K seats you need super powers
STORMS More seats = more latency
5. Super VDI with Active Flash
Thousands of Seats
SUPER SIZE
SUPER PERFORMANCE
SUPER AVAILABLE
6. Performance and Efficiency Evolution
Dynamic provisioning
• Thin provisioning for efficiency
• Wide striping for performance
Flash POOL
SAS POOL
NL-SAS POOL
DP-VOL
3
DP-
VOL 2
DP-VOL
1
Dynamic tiering
• Automatically moves the most
active data to the fastest tier for
maximum performance; least active
to lower-cost tiers
• 30-minute to 24-hour cycle time
• Inherits HDP* advantages
Flash
SAS
NL-SAS
HDT** Pool
1
DP-VOL 1
Tier 0
Tier 1
Tier 2
Active flash
Real-time −
seconds to sub-seconds cycle time
• HDT with LOWER RESPONSE TIME
capabilities
• Inherits HDT, HDP advantages
Flash
SAS
NL-SAS
HDT Pool 2 DP-VOL 2
Tier 0
Tier 1
Tier 2
Active Flash: Suddenly Active Data
* HDP = Hitachi Dynamic Provisioning ** HDT = Hitachi Dynamic Tiering
7. Time (t)
10:00 10:309:30
Changing I/O load to a page
Average I/O load in a recent period (latest
few seconds) Increased suddenly
Tier0 Tier1Tier1
Relocate to Tier0
I/OLoadofPage
(IOPS)
HDT active flash feature
What Is the Hitachi Dynamic Tiering (HDT)
Active Flash Feature?
Average I/O load in a long period
(latest few days) Stable/continue
8. Dynamic Tiering Active Flash Feature for VDI
Tier 0
SSD
Flash
Tier 1
SAS 10K
Tier 2
NL-SAS
CostPerformance
Active Flash
Boot and Log In
Inactive Seat
Steady State
500 500 500 500
2000 Seat VDI
Point-in-time
VDI scenario
9. Dynamic Tiering Active Flash Feature
Aligns to Business Timing
No special configurations or dedicated flash hardware required
Time to log in
Data that becomes highly active gets a performance boost in real time
High-Activity Boot StormInactive VDI Data 8:01 a.m.
Hey, this isn’t my typical
workload. This is very,
very active!
Boot
Storm!
VALUE Automate performance and maximize flash ROI
Boot Storm!
Performance. Performance requirements can vary greatly between peak time (boot and log-on storm) and steady state. Proper storage, networks, and servers products and design are critical for satisfactory VDI performance.
Scalability. Most VDI systems must be capable of quickly adding new users, incorporating employees from acquired companies, new organizations and the like. Performance, complexity and cost per incremental desktop are key considerations when designing VDI systems for scalability.
Availability. Because VDI systems can support upwards of a thousand of users, downtime can be costly. Therefore, system availability is a critical consideration when designing a VDI implementation.
Source: The New Business Economics of Desktop Virtualization By Tom Trela, Director of Solutions and Cloud Product Management at Hitachi Data Systems, and Scott Shimomura, Director of Product Marketing for Data Center SAN at Brocade
At about 2k seats a “regular” VDI solution loses it’s ability to scale and meet performance requirements. As more seats are added latency is introduced.
Super powers are needed to keep growing your VDI environment
At Hitachi we have regular VDI and Super-Size VDI
Super Size VDI includes UCP 4000 with native integration with VMware vCenter. Choose your storage size from the VSP G series: from entry to ultimate G200 through G1000 (from 1PB to 8 PB)
Choose your Super Size
Super size
UCP 4000 scales from 2 to 128 high-density Hitachi Intel-based blade servers.
VSP G series (VSP G series: G1000 to G200) storage scales up to 8 PB (starts at less than 1PB)
Super performance
G1000
3.3 Million IOPS
896 GB/sec aggregate bandwidth
Super availability
Zero time to recovery (RTO) using stretch cluster
Hitachi Dynamic Tiering has a new feature called active flash. Hot pages are immediately tiered to the fastest media for close to real time workload responsiveness for enhanced performance.
With the new active flash mode, monitoring and relocation occur synchronously so detecting a sudden increase in load results in immediate tiering to fast media. Settings for monitoring collection, relocation and manual relocation are not needed. Active flash is activated at the HDT pool level via a radio button.
Active flash will provide synchronous promotion and high speed demotion. Synchronous promotion performs relocation within seconds to sub-seconds based on the most recent IOPH data concurrently with I/O processing. High speed demotion moves pages from the top tier with new cycle processing using a page list for low use pages. High speed demotion processing is periodically performed to reserve free area in tier 1 to maintain performance. This processing is not performed for tier 2 and tier 3.
HDT continues to work as always in concert with active flash. By way of comparison, with active flash off, the tier range for HDT is determined in each cycle from frequency distribution and allocation is performed asynchronously. Active flash is expected to improve performance by catching sudden workload activity, while the increased tier 1 activity will be offset by high speed demotion to reserve tier 1 capacity. I/O can be balanced using the HDT tier allocation policy to avoid exceeding the performance potential in all tiers. Setting the tier performance % lower than its performance potential will maintain response times under heavy workloads. Processor load for monitoring and relocation with or without active flash turned on is expected to be the same. The pools, sizes, tiers, tier capacity and supported media is expected to be the same as HDP. Capacity per tier, number of volumes will vary by model.
Active flash is continually working to automatically optimize performance in your VDI environment by promoting pages requested in real time. This graph show two data sets (pages, one getting suddenly busy while the other is in steady state. The most active page is promoted to Tier 0 in real time so that it can benefit from Tier 0 flash performance automatically. All this happened transparent to the seat.
Testing has shown that a hybrid environment using active flash can perform at the same level as an all flash environment for less cost; because it uses lower cost hard drives for steady state
Active flash let’s you align your VDI environment to business timing. Your business gets a solution that automatically responds to varying levels of end user data activity automatically.
The super size VDI solution doesn’t require specialized hardware or dedicated all flash solutions.
We have test results that show a hybrid solution using active flash performs equal to a similar sized all flash solution.
UCP4000, VSP G1000 and the entire G series have the highest availability and resilience using active/active storage capabilities that allow these platforms to meet zero recovery point (RPO) and zero recovery time objectives (RTO).
Global-active device is a storage operating system (SVOS) feature that simplifies high availability with data volumes that have the ability to store read/write copies of the same data in 2 places at the same time. Active-active design enables production workloads on all systems while maintaining full data consistency and protection.