2. System x® and
BladeCenter®
2
x86 servers are taking on more demanding roles, including
high-end business critical applications
x86 server segment is the
largest and fastest
growing
60% of customers
deploying business
critical workloads onto
x86, up from 30% in 2008
51% of all x86
workloads virtualized
by end of 2010, growing to
69% by 2013
Virtual workloads getting
larger, up by 3x
between 2006 and 2009
x86 important for database:
52% of relational
database on x86
59% of these databases
are projected to be
larger than 3TB by
2013
3. System x® and
BladeCenter®
3
eX5 is the RIGHT x86 platform for business applications.
Today and tomorrow.
Virtualization
Database
Reliability
eX5 can save you money starting at 30 virtual
machines per server
eX5 maximizes database performance, minimizes cost
and grows with your data
eX5: The most reliable x86 platform to date
5. System x® and
BladeCenter®
5
Server virtualization
and consolidation
of “low hanging fruit”
Virtualization of business
critical applications,
database and desktops
Private cloud
enabling IT as a
Service
The majority of all x86 workloads will be
virtualized by the end of 2010
Benefits
CapEx
OpEx
Challenges
Consolidation
Benefits
Quality of Service
Decrease in Provisioning Time
Challenges
Availability
Application
Performance
Benefits
Zero-touch Operation
Best Quality of Service
Challenges
Compliance
Time to Market
Clients are running more, larger, and mission critical workloads
as virtual machines
6. System x® and
BladeCenter®
6
With increased virtualization and larger workloads, memory
and I/O become the new bottleneck
While cores per CPU keep increasing with each generation,
system memory and I/O flexibility have not kept up.
Now, virtualization requires systems with large
memory footprint and flexible I/O in order to drive
maximum VM agility
7. System x® and
BladeCenter®
7
eX5 is the only x86 platform that delivers the memory technology
and I/O flexibility required for breakthrough virtualization agility
MAX5:
Balanced memory capacity and
Independent memory expansion
5x the memory of leading 2-socket x86
50% more memory of any 4-socket x86
eX5 technology delivers unique customer value
Virtual Fabric:
More and flexible I/O pipes
4x I/O ports over traditional 10G
Fast, scalable, flexible
8. System x® and
BladeCenter®
8
2S, 16 DIMM
Expand memory capacity without additional CPUs
or software licenses
Double the number of memory DIMMs per CPU
2x memory bandwidth
Maintain full memory performance
5x the memory capacity in two-sockets vs. today’s
leading two-socket systems
THE PROBLEM
Industry standard
With embedded memory controllers,
memory capacity is tied to processors
✗ More software licenses to purchase
✗ More systems to manage
THE SOLUTION
MAX5
MAX5 doubles memory without adding more processors
2S, 12 DIMM 2S, 12 DIMM
2S, 40 DIMM
9. System x® and
BladeCenter®
9
Any
Networking technology,
Networking speed
50% Fewer adapters
75% Fewer switch modules
Avoid complex LAN vs. SAN capacity
planning: Virtual Fabric can become
Ethernet, FCoE, and iSCSI
Allocate bandwidth where needed:
Mission critical workloads get most, while
test projects get appropriate share of
bandwidth
Deploy less hardware:
Fewer adapters and switches needed to
support the same workload with IBM
Virtual Fabric technology
Pre-integrated in select models:
Rack servers – x3850 X5, x3690 X5
Blade servers – HX5, HS22, HS22V
Virtual Fabric breaks the I/O bottleneck and simplifies deployment
10. System x® and
BladeCenter®
10
x3690 X5 is 34% less expensive than HP 2-socket for virtualization
Support 2x or larger VMs vs HP with half the servers and half the SW license cost
1x x3690 X5
$22K Server Hardware
$10K VMWare license cost
$1K/yr Datacenter power cost
$5K/yr Maintenance cost
$38K First year TCO
2x HP DL380 G7
$26K Server Hardware
$20K VMWare license cost
$2K/yr Datacenter power cost
$10K/yr Maintenance cost
$58K First year TCO
IBM Advantage
▼
1 x3690 X5 does the work of 2 DL380s
▼
$20K savings
35 VM
environment
x3690 X5:
28 x 8GB
Each DL380 G7:
14 x 8GB
11. System x® and
BladeCenter®
11
IBM Advantage
▼
1 x3690 X5 + MAX5 does the work of 4 DL380s
▼
$57K savings
x3690 X5 + MAX5 is 48% less expensive than HP for virtualization
1x x3690 X5 + MAX5
$45K Server Hardware
$10K VMWare license cost
$1K/yr Datacenter power cost
$5K/yr Maintenance cost
$61K First year TCO
4x HP DL380 G7
$54K Server Hardware
$40K VMWare license cost
$4K/yr Datacenter power cost
$20K/yr Maintenance cost
$118K First year TCO
75 VM
environment
x3690 X5:
56 x 8GB
Each DL380 G7:
14 x 8GB
12. System x® and
BladeCenter®
12
HX5+MAX5 is 22% less expensive than HP DP servers for Virtualization
Support 2x or larger VMs vs HP with half the servers and half the SW license cost
1x HX5+MAX5
2S, 40x8G (320GB)
$34K Server Hardware (1 2S server)
$10K VMWare license cost
$1K/yr Datacenter power cost
$5K/yr Maintenance cost
$50K First year TCO
2x HP BL490c G6
2S, 18x8G (144GB)
$33K Server Hardware (2 2S servers)
$20K VMWare license cost
$1.5K/yr Datacenter power cost
$10K/yr Maintenance cost
$64.5K First year TCO
IBM Advantage
2x memory, more & larger VMs per server,
drastically reduced SW licensing costs
▼
HP requires 2 WSM-EP based blades to match 1 HX5
▼
$14.5K savings60 VM
environment
All systems @
max memory
w/ 8G DIMMs
13. System x® and
BladeCenter®
13
eX5 can save money starting at 30 virtual machines per server
eX5 is the optimal x86
platform for virtualization
starting at 30 VMs
Lower price per virtual machine
eX5 system does the work
of 2 or more HP systems
50% fewer servers 50% lower
software licensing cost
PLUS, save on management &
energy costs
Look beyond base server
price, and consider total
solution cost for your
workload
eX5 systems plus software cost
is less expensive for virtualization
than industry standard
alternatives
x3690 X5
HX5+MAX5
HP BL490c G6
HP DL380 G7
HP DL580 G7
14. System x® and
BladeCenter®
14
Only eX5 can scale to meet your business’s needs – large or
small – today and tomorrow
IBM Scalable 2-socket
HX5
HP 2-socket DP blade
IBM Scalable 4-socket
x3850 X5
HP 4-socket monolithic
rack server
Pay-as-you-grow expansion
Scalable for memory for over
50% greater capacity than HP
Scalable for CPU up to 8S
IBM Scalable 2-socket
x3690 X5
HP 4-socket monolithic
rack server
Pay-as-you-grow expansion
2x memory per CPU vs HP
2x VMs per server at same SW
license cost
HX5 shipping NOW
Over 25% more memory DIMMs
per CPU w/ MAX5 vs HP
More, larger VMs at same SW
license cost
Pay-as-you-grow expansion for
CPU, memory or both!
Rack Servers Blade Servers
15 – 65 VMs per server Large VMs or 40+ VMs 15+ VMs per server
15. System x® and
BladeCenter®
15
eX5 supports over 300 VMs and can save you money starting at
30
Only eX5 platforms enable you to eliminate the virtualization
bottleneck
Delivers the most memory per server
Solves the I/O bottleneck with Virtual Fabric
Saves you money starting at just 30 VMs per server
Adapts to your changing workloads – scales for CPU or memory
&
17. System x® and
BladeCenter®
17
Companies use data to run their businesses. Databases run
on x86 systems
Today: 36% of data
warehouses are over
3TB
Tomorrow: Nearly
60% of data
warehouses projected
to be over 3TB by
2013
Today, x86 systems
host 52% of all
relational
databases.
Database workloads continue to outpace available system
resources. Clients require systems that can reliably handle larger,
more I/O intensive databases
18. System x® and
BladeCenter®
18
Databases tax system resources
Today x86 systems host 52% of all relational databases.
You need solutions that can match increasing requirements.
Tomorrow you will face even larger demands.
Today’s Database Challenges
Increasing number of users and more queries
Database size increasing at a fast rate
More I/O intensive applications
19. System x® and
BladeCenter®
19
Databases run faster at lower cost on eX5 platforms
eX5 technology delivers unique customer value
MAX5: Balanced memory capacity and independent memory expansion
Over 5x the memory of leading 2-socket x86; 50% more memory than any 4-socket x86
eXFlash: Integrated, fast, economical data access
240K IOPS at 97% lower cost and 99% less power vs traditional alternatives
Scalable servers: Expand for CPU or memory to keep up with growing datasets
20. System x® and
BladeCenter®
20
eX5 delivers integrated, high speed data access at 97% lower
cost and 99% less power versus traditional alternatives
THE PROBLEM
Industry standard
✗ 80 JBODS with 800
spinning disks
✗ Each with 300 IOPS
✗ Cost $1153K over 3 years
Only eX5 can deliver the fast access to data required for high
quality of service in multi user database environments
x3850 X5
with eXFlash
1x eXFlash for self contained fast
(240K) IOPS solution
No additional parts required
97% less expensive and
99% less power
THE SOLUTION
eXFlash
Customer requirement: Fast data access with 240K IOPS
21. System x® and
BladeCenter®
21
Your database is growing, your server needs to keep up
Offers pay-as-you grow expansion
Scales up to 8-sockets
Supports independent memory expansion
Get up to 2x system resources in minutes
✗ Requires purchase of additional servers
to keep up
✗ Duplicates resources
✗ Adds delays
✗ Increases cost
THE PROBLEM
Industry standard
THE SOLUTION
eX5
2S 2S 2S 2S
2S4S 4S8S
22. System x® and
BladeCenter®
22
x3690 X5 is 41% less expensive than Dell for 1000 user SQL
environment
IBM Advantage
▼
2-socket high-end system right sized
for a 1,000 user SQL deployment. Clients forced into a
more expensive 4-socket solution with Dell.
▼
$47K savings
Dell R810
$9K Server Hardware
$100K 4S SQL license cost
$1K/yr Datacenter power cost
$5K/yr Systems Maintenance cost
$115K total cost
x3690 X5
$12K Server Hardware
$50K 2S SQL license cost
$1K/yr Datacenter power cost
$5K/yr Maintenance cost
$68K total cost
1000 user
database
23. System x® and
BladeCenter®
23
IBM Advantage
Consolidate 16 single-database servers to 1 x3850 X5
▼
$249K savings
x3850 X5 is 49% less expensive than HP for SQL Database
consolidation
SQL performance based
on TPCE projections
x16
16x HP DL380
$125K Server Hardware
$290K SQL license cost
$16K/yr Datacenter power cost
$80K/yr Maintenance cost
$511K First year TCO
1x x3850 X5
$91K Server Hardware
$165K SQL license cost
$1K/yr Datacenter power cost
$5K/yr Maintenance cost
$262K First year TCO
16 SQL
databases
Source: http://www.redbooks.ibm.com/redpapers/pdfs/redp4690.pdf
24. System x® and
BladeCenter®
24
Only HX5 supports top-bin CPUs in 4-socket configuration and delivers 2x
memory bandwidth. Competition requires 17% more servers to do same
work, leading to increased HW, SW and management costs
Note: SQL performance based on TPCE projections
4-socket HX5 is 23% less expensive for equivalent SQL
transactions compared to Dell
6x IBM HX5
(4x X7560, 32x8G)
$305K server hardware
$600K SQL license cost
$6K Power cost
$30K Maintenance cost
$941K total cost
7x Dell M910
(4x L7555, 32x16G)
$474K server hardware
$700K SQL license cost
$7K Power cost
$35K Maintenance cost
$1.22M total cost
IBM Advantage
▼
15% greater performance per blade with HX5
6 HX5 servers do the work of 7 Dell servers
▼
$275K savings
All systems @
max memory
HX5 w/ 32x8G
M910 w/ 32x16G
x7
x6
25. System x® and
BladeCenter®
25
&
eX5 maximizes database performance, minimizes cost and
grows with your data
Cuts cost on licenses, power and hardware with eX5 innovations
Supports more memory for more data with MAX5
Delivers fast access to data with eXFlash
Scales with your data sets
&
26. System x® and
BladeCenter®
26
eX5 is the x86 platform for your needs today and tomorrow
Virtualization
Database
Reliability
eX5 can save you money starting at 30 virtual
machines per server
eX5 maximizes database performance, minimizes cost
and grows with your data
eX5: The most reliable x86 platform to date
27. System x® and
BladeCenter®
27
Traditionally, availability
previously found only in
large-scale systems was
critical. That’s where you
looked for it.
Now, hardware and
software together provide
the best application
availability.
And eX5
delivers.
With mission critical workloads moving to x86,
systems need to keep up and application
availability is critical.
Confidently deploy mission critical applications on eX5
28. System x® and
BladeCenter®
28
System x® and
BladeCenter®
eX5 enhances industry standard reliability
Reliable
Hardware
Predictive
Failure
Smart
Software
Fast
Recovery
29. System x® and
BladeCenter®
29
Memory ProteXionTM
Why it is important
Memory reliability critical, as lots of memory in high-end systems
Benefits
Up to 4x the memory protection over industry standard
Superior memory reliability
MAX5 with 8G DIMMs more reliable than industry standard server
with 16G base server filled with 16G DIMMs
How it works
Integrated into eX5 and works with off-the-shelf DIMMs
Reliable
Hardware
Predictive
Failure
Smart
Software
Fast
Recovery
eX5 with Memory ProteXionTM
continues
to run and detect errors with two completely
bad DRAMS! eX5 (MAX5)
Nehalem EP fails with only
two bad bits from different DRAMs
Intel EP
Nehalem EX supports SDDC+1 on both x8 and
x4 DIMMs. System continues to run with one
bad DRAM chip and another error corrected
Intel EX
30. System x® and
BladeCenter®
30
High availability features such as IBM Predictive
Failure Alerts maximize application uptime
Reliable
Hardware
Predictive
Failure
Smart
Software
Fast
Recovery
Predictive Failure:
Warns ahead of potential hardware failures
Triggers preemptive action
Maintains application availability
Only eX5 monitors vital system components for maximum availability
PFA DELL HP IBM
Hardware Yes Yes Yes
Disk Yes Yes Yes
Memory No Yes Yes
CPU No No Yes
Fan No No Yes
VRM No No Yes
Pwr supply No No Yes
31. System x® and
BladeCenter®
31
IBM VMControl monitors hardware and
automates smart movement of workloads
ahead of potential hardware failure
VMControl delivers:
A single tool to manage both physical and virtual servers
System Director automation to link physical alerts to
VMwareVMotion™
VMControl Express, a free Systems Director plug-in available
for download
Systems
Director
Server
VM Control
Express
PFA
Virtual
Machine
Systems
Director
Agent
VMWare
vCenter
Virtual
Machine
Virtual
Machine
Virtual
Machine
Reliable
Hardware
Predictive
Failure
Smart
Software
Fast
Recovery
32. System x® and
BladeCenter®
32
Automatic error detection
System reboots as a single-node, and brings the eX5
system back online faster
System continues to operate without the failed node
✗ Monolithic multi-socket systems down
until the fatal error resolved
✗ No automatic reboot
THE PROBLEM
Industry standard
THE SOLUTION
FlexNode Failover with eX5
Back up
and running
Down until
help arrives
Industry
Standard
Automatic node failover makes eX5
multi-node systems the most available
Situation: Unrecoverable failure in an MP server
Reliable
Hardware
Predictive
Failure
Smart
Software
Fast
Recovery
33. System x® and
BladeCenter®
33
eX5 provides more reliability with fewer components
IBM Advantage
Consolidating with eX5 lowers cost and increases reliability
▼
Fewer servers to buy and manage
More reliable hardware
Fewer components that can fail
7 less-reliable
HP DL 380 G7
Higher risk with more components
that can fail:
10 more power supplies
40 more HDD, no reliable SSD
36 more fans
2 high-reliability
x3850 X5
Lower risk with more RAS features:
54% more memory reliability features
60% more CPU reliability features
Customer requirement: Reliable hardware platform to deploy 250 VMs
Challenge: Cost, management complexity, datacenter space
34. System x® and
BladeCenter®
34
&
eX5 combines highly reliable hardware with smart software to
deliver the most reliable x86 platform to date
Delivers superior system availability with highly reliable hardware and
unique features like Memory ProteXionTM
Maximizes application availability with features such as IBM Predictive
Failure Alerts
Enables outstanding application availability with VMControl
Delivers the most available MP platform with FlexNode Failover
&
35. System x® and
BladeCenter®
35
&
eX5 can save you money starting at 30 virtual
machines per server
eX5: The most reliable x86 platform to date
eX5 maximizes database performance, minimizes
cost and grows with your data
37. System x® and
BladeCenter®
37
Virtualization
Positioning
15+ VMs per server
(CPU limited)
15+ VMs per server
(memory limited)
15 - 65 VMs per system
Large VMs or 40+ VMs
per system
Specs
30/60mm, up to 4S, 32 cores
16 – 32 DIMM, 256GB
2-4 SSDs w/ HW RAID
60 mm, 2S, 16 cores
40 DIMM, 320GB
2 SSDs w/ HW RAID
2U, 2 socket, 16 cores
32 DIMMs, 512GB
64 DIMMs, 1TB with MAX5
16x 2.5” HDD / 24x 1.8” SSD
4U, 4 socket, 32 cores
Scalable to 8 proc, 64 core
64 DIMMs, 1TB
96 DIMMs, 1.5TB with MAX5
8x 2.5” HDD / 16x 1.8” SSD
Workloads
Database
Simulations
Business intelligence
Enterprise (ERP, CRM)
Max virtualization performance
Virtualization
Database (using internal or
external storage)
Enterprise Clusters
Database
ERP
Large Scale Virtualization
Key Features
Max compute density
Max IO
Node partitioning
Node failover
Max memory density
Max memory per core
Leadership memory per core
Maximum internal IOPs
Enterprise Reliability
Two socket building block for
software license savings
Leadership x86 performance
Enterprise Reliability
Memory and processor scalable
Large Flexible I/O
configurability
eX5 is the best system platform for application deployment
Virtualization, database and transaction processing
Enterprise database
& simulations 2S-4S
HX5
Enterprise virtualization 2S
HX5 + MAX5 x3690 X5
High End 2-Socket Enterprise 4-Socket
x3850 X5
Notas do Editor
Hello…welcome…and thank you for taking the time to be here with us today. I couldn’t be more excited to share with you the value of eX5, our most significant x86 portfolio to date.
This presentation shows how eX5 is optimized for YOUR business critical workloads. With unique innovations, eX5 goes beyond industry standard to break through today’s IT bottlenecks to enable you to deploy confidently the most demanding workloads such as virtualization and database.
eX5 is about ushering in an era of NEW THINKING in x86 servers
So let’s begin.
Pitch from chart.
Key message: x86 servers are taking on more demanding roles, including high-end and business critical applications. Virtualization growing more than ever. Database workloads growing more than ever.
New thinking is required, especially in the x86 space….since 98 percent of data center is x86…or 53 percent of the overall server market.
-------------
Key points to make: We must win the mindshare at virtualization and database platforms…if successful, then we will have won the most significant opportunities in the datacenter
Pitch from chart.
Key message: eX5 is the optimal x86 platform for virtualization and database. eX5 lowers costs per workload and offers rock solid reliability for your mission critical workloads.
Key points to make: Really hit home the importance of these two applications, and how reliability is critical for each
Let’s begin with virtualization..
Note to presenters: This chart is adapted from VMWare presentation, and tightly aligned to VMWare’s view of virtualization
-------
As clients move along the virtualization journey, clients are running more, larger and mission critical workloads as virtual machines.
According to VMWare, organizations that undertake the virtualization Journey typically go through 3 stages of adoption as shown on this chart.
Let’s spend some time going deeper into each of these stages, and understanding how each stage has its own unique challenges and benefits
The first stage of the journey we call IT Production
This is the Cost-Efficiency phase, where IT organizations start virtualizing the “low risk”, low hanging fruit: File, Print, Web servers, domain controllers, Test and Development servers. The predominant value proposition in this phase is consolidation of IT infrastructure to save on hardware, space and cooling.
The second stage of the journey we call Business Production
This is the Quality of Service phase, tackling business applications (such as Microsoft Exchange, SAP, Oracle Apps, Databases etc.) Here, clients are not so much worried about consolidation and cost efficiency but they are concerned about performance, about meeting SLA, and reducing downtime.
At this stage, value becomes more than cost savings. Here, clients use virtualization to improve the quality of service to application owners by providing better availability without costly clustering technologies, better performance for older applications on new hardware, and faster time to market for new applications due to reduced provisioning time.
The final stage of the journey we call IT as a Service
This is the Business Agility phase, leveraging a private cloud network. At this stage, organizations are looking to scale their virtualization effort so that they can in turn scale the associated benefits that come from process automation, better resiliency, increased quality of service and Business Agility.
As clients move along the virtualization journey, it is clear that they are consolidating more virtual machines per server and these virtual machines are getting larger.
This new world of virtualization places new and increased requirements on memory and network resources. Increasingly, virtualization clients have told us that they are running out of memory and/or I/O lanes far before they run out of CPU capacity – leading to underutilized servers even with virtualization.
This is not surprising, as cores per CPUs keep increasing with each generation, system memory and I/O flexibility has not kept up, leading to new bottlenecks.
Today, virtualization requires systems with large memory footprint and more I/O ports in order to drive maximum VM agility
eX5 is the only x86 platform that delivers the memory technology and I/O flexibility required for breakthrough virtualization agility
MAX5 enables you to expand system memory without additional CPUs or SW licenses. In fact, with MAX5, eX5 2-socket systems can deliver 5x the memory of leading 2-socket x86 servers to enable new levels of utilization for memory constrained applications.
Another unique technology is Virtual Fabric. It allows you to split a 2x 10G networking ports into eight ports with flexible bandwidth – anywhere from 100Mb to 10G. More ports and greater flexibility enables clients to assign quality of service parameters to I/O ports for maximum VM agility.
With technologies such as MAX5 and Virtual Fabric, eX5 enables clients to break through the virtualization bottleneck for optimal server utilization.
Let me take a moment to introduce MAX5 and Virtual Fabric innovations in a bit more detail. First, MAX5. This innovation is unique to IBM, and unique to eX5.
Let’s see how MAX5 delivers value for virtualization….
For memory constrained applications such as virtualization, industry standard x86 servers are the problem that results in underutilized servers. As discussed earlier, virtualization clients often out of memory much before they run out of CPU capacity. To support more virtual machines or run more applications, clients are forced to buy additional servers…it means more CPUs…more software licenses and so on, all leading to higher system cost and greater complexity.
MAX5 is the solution to the industry standard. MAX5 enables the eX5 systems to scale system memory without adding CPUs, and deliver up to twice as much memory as competitive NEH-EX systems. MAX5 not only allows you to break through the memory capacity barrier, it also increases memory bandwidth and does so while maintaining full memory performance. No one else in the industry can do this.
With MAX5, customers can run more virtual machines or larger virtual machines. They can run bigger faster in-memory databases – all without additional CPUs or additional SW licenses.
Next innovation I’d like to talk about is Virtual Fabric.
We help clients simplify deployment by extending the virtual fabric that we introduced last year with our BladeCenter to our eX5 portfolio. With virtual fabric, clients can simplify their infrastructure because they can avoid complex LAN versus SAN capacity planning.
Because Virtual Fabric can become Ethernet, Fibre channel Over Ethernet or iSCSI.
Because they can run any network technology at any networking speed.
And it can also do so with 50% fewer adapters, and 75% fewer switch modules.
Transition: Now, let’s look at a few examples of how these eX5 features translates to quantifiable value for deploying virtualization
Key message: One x3690 X5 (2-socket 32 DIMM Neh EX) does the work of two HP DL380 G7 (2-socket 18 DIMM WSM-EP) systems
- Take this 35-VM example – it takes 2 HP DL380s to reach 35 VMs, but you only need one 3690.
- Support over 2x VMs or larger VMs vs HP with half the servers and half the SW license cost
- Also, save on power & cooling and management costs
Here, we see a comparison between 3690 and HP DL380 G7.
3690 delivers a much larger memory footprint compared to mainstream HP 2-socket servers, and support 2x VMs or larger VMs vs HP with half the servers and half the SW license cost. Fewer servers also results in energy and management savings.
In this 35 VM example, 3690 based solution is 32% less expensive than HP for virtualization, as one 3690 with 32 DIMMs does the work of two HP DL 380 (2-socket 18 DIMM WSM-EP) systems.
While the eX5 server may initially cost more, it’s easy to see how it delivers lower cost per VM and is a superior choice for virtualization.
VMs per Server
3690: 35 VMs per server at ~6G / VM
HP DL380: 17 VMs per server at ~6G / VM
Max VMs per server derived from Virtualization Sizing Guide adjusted for 2012 growth case.
Server Cost
US list price current as of September, 2010
IBM: Server (1x x3690 X5 @ $x) + Memory (28 x 8G @ $x each)
HP: Server (2x DL380 G7 @ $x) + Memory (28 x 8G @ $x each)
Neh –EX x7560 used in config
VMWare License Cost
10K per 2-socket server ($7K for Enterprise Plus + $3K for 3-year subscription)
Datacenter power savings
IBM: 800 W per server * 24 hours * 365 days * $0.10 / KWh * 1.5 for cooling overhead $1K per server
HP: 500 W per server * 24 hours * 365 days * $0.10 / KWh * 1.5 for cooling overhead $657 per server
Management savings
Assume $100,000 of server administrator costs to manage 20 servers. For 1 server, yearly cost will be $5K
Key message: Increase consolidation of standard two sockets to 4:1 by adding MAX5
- Support over 2x VMs or larger VMs vs HP with half the servers and half the SW license cost
- Also, save on power & cooling and management costs
Here, we see a comparison between 3690 + MAX5 and HP DL380 G7.
3690 + MAX5 delivers a much larger memory footprint compared to mainstream HP 2-socket servers, and support 4x VMs or larger VMs vs HP with one quarter the servers and one quarter the SW license cost. Fewer servers also results in energy and management savings.
In this 75 VM example, 3690 based solution is 32% less expensive than HP for virtualization, as one 3690 + MAX5 with 64 DIMMs does the work of two HP DL 380 (2-socket 18 DIMM WSM-EP) systems.
While the eX5 server may initially cost more, it’s easy to see how it delivers lower cost per VM and is a superior choice for virtualization.
VMs per Server
3690: 35 VMs per server at ~6G / VM
HP DL380: 17 VMs per server at ~6G / VM
Max VMs per server derived from Virtualization Sizing Guide adjusted for 2012 growth case.
Server Cost
US list price current as of September, 2010
IBM: Server (1x x3690 X5 + MAX5 @ $x) + Memory (28 x 8G @ $x each)
HP: Server (4x DL380 G7 @ $x) + Memory (28 x 8G @ $x each)
Neh –EX x7560 used in config
VMWare License Cost
10K per 2-socket server ($7K for Enterprise Plus + $3K for 3-year subscription)
Datacenter power savings
IBM: 800 W per server * 24 hours * 365 days * $0.10 / KWh * 1.5 for cooling overhead $1K per server
HP: 500 W per server * 24 hours * 365 days * $0.10 / KWh * 1.5 for cooling overhead $657 per server
Management savings
Assume $100,000 of server administrator costs to manage 20 servers. For 1 server, yearly cost will be $5K
You may think that a mainstream 2-socket 18-DIMM WSM-EX server may be a natural choice for virtualization. Well, as you will see on this page, that is not always true. Here, we see a comparison between HX5+MAX5 and HP BL490.
HX5 delivers much larger memory footprint compared to mainstream HP 2-socket servers, and support 2x VMs or larger VMs vs HP with half the servers and half the SW license cost. Fewer servers also results in energy and management savings.
In this 60 VM example, HX5 based solution is 22% less expensive than HP for virtualization, as one HX5+MAX5 (2-socket 40 DIMM Neh EX) does the work of two HP BL 490 (2-socket 18 DIMM WSM-EP) systems.
HX5+MAX5 delivers exactly what the customer needs for virtualization. More memory for virtualization results in lower cost per VM.
MAX VMs per Server
HX5 + MAX5: 60 VMs per server at ~4G / VM
HP BL490: 30 VMs per server at ~4G / VM
Max VMs per server derived from Virtualization Sizing Guide adjusted for 2012 growth case. Exact Number per VM represents mid-point of “moderate” zone.
Server Cost
US list price current as of September, 2010
IBM: Server (1x HX5+MAX5 @ $13,680) + Memory (40 x 8G @ $479 each) + adapters (@ $1500 per blade) = $34K
HP: Server (2x BL490c @ $6500 each) + Memory (18 x 8G @ $509 each) + Networking (@ $1500 per blade) = $33K
VMWare License Cost
10K per 2-socket server ($7K for Enterprise Plus + $3K for 3-year subscription)
Datacenter power savings
IBM: 800 W per server * 24 hours * 365 days * $0.10 / KWh * 1.5 for cooling overhead $1K per server
HP: 500 W per server * 24 hours * 365 days * $0.10 / KWh * 1.5 for cooling overhead $657 per server
Management savings
Assume $100,000 of server administrator costs to manage 20 servers. For 1 server, yearly cost will be $5K
As we have just seen from examples on previous pages, it is imperative to evaluate cost per workload at a system level – not just at server level. Servers by themselves do nothing. You also need to look at cost of software, networking, power and cooling and so on.
When you put this all together, it is evident that eX5 systems can save you money starting at just 30 virtual machines per server as one eX5 system does the work of two or more industry standard DP servers. That saves you money.
eX5 can do the same work with fewer servers, lower software license costs, lower energy bills and simplified management.
Of course, this number assumes a particular virtual machine profile and fully configured systems with 8G DIMMs. You may actually find the crossover point to be lower based on your workloads or system configuration.
Clients have environments that need to support a wide range of workloads. As we did this positioning, one thing we kept in mind is that client workloads are only going to grow. VMs are getting larger, and you are virtualizing more and more workloads. For today’s workloads, eX5 is the best choice..for tomorrow’s workloads, only scalable eX5 systems can grow with your business.
Pitch From Chart…
Key message:
x3690 X5 the right choice if you are looking to deploy 15 – 65 VMs per server. On prev page, you already saw how x3690 is a superior choice at 30 VMs. If you are looking to deploy larger VMs or at less than max system configuration, then the cross over point comes down to about 15 VMs per server. For less than 15, you could deploy a DP server like x3550 M3. Note the advantages as shown on the slide
x3850 X5 remains the ideal choice for maximum system consolidation, or when deploying large virtual machines such as database or ERP. Note the advantages as shown on the slide
Lastly, HX5+MAX5 is the optimal blade platform when deploying over 15 VMs per server. Note the advantages as shown on the slide
In summary, eX5 is the only x86 platform that delivers the portfolio breath, memory technology and I/O flexibility required for break through virtualization agility.
MAX5 delivers balanced memory capacity plus independent memory expansion, as well as the most memory per server that is vital for memory constrained apps such as virtualization.
Virtual Fabric delivers more and flexible I/O pipes required to deliver high quality of service for virtualized workloads
Also, only eX5 can scale with your changing workloads for continued investment protection.
Bottom line is that eX5 not only helps you break through the virtualization bottleneck, but also significantly drives down the cost to deploy virtualization.
x86-based environments are increasingly important for database workloads. Today x86 systems host 52% of all relational databases. These database workloads increasingly require more resources, as they are getting larger and # of users that need to be supported is increasing.
Similar to the scenario with virtualization, database workloads continue to outpace available system resources. Clients require systems that can reliably handle larger, more I/O intensive databases
This new world of databases places new and increased requirements on system resources as Clients increasingly want systems that can reliably handle larger, more I/O intensive databases.
Once again, industry standard becomes the bottleneck, as mainstream systems just can’t keep up with the demands of these large, fast growing database environment. As clients deploy larger datasets, memory and fast access to data have become the bottleneck, causing longer run times and slower response times.
Your database can run faster at lower cost on eX5 platform, as eX5 is the only x86 platform that delivers the memory and I/O technology and scalability required for your demanding database environment.
With MAX5, only eX5 can deliver memory required for large in-memory database or large scale consolidation resulting in lower cost per workload from fewer servers and reduced SW license costs
With eXFlash, only eX5 can deliver the fast access to data required for high quality of service in multi user database environment
Lastly, your workloads are not going to stay constant. Only scalable eX5 systems can adapt.
Punch line: eX5 enables you to breakthrough the database bottleneck and delivers new levels of performance and consolidation to reduce cost and simplify deployment for database workloads.
Transition: Now, let’s look at some of the unique eX5 innovations discussed on this page in a bit more detail. We have already talked about MAX5. Let’s begin with eXFlash.
Clients in a multi-user database environment often require fast access to data. Per example shown on this page, say you are looking for a to handle multiple simultaneous user queries and a for a solution to deliver 240K IOPS (IO per second) – pronounced “I OPS”
The industry standard answer would be to deploy enough HDD spindles to meet the required performance target. In this example, it would mean 800 spinning drives at 300 IOPS each. As you see on the left side, that would take five fully configured racks of storage. For an IT manager, that’s cost, datacenter space, power and cooling overhead, configuration and cable nightmare, time and complexity.
eXFlash is the solution. exFlash delivers integrated solid state disk technology with a high speed backplane, high speed drives and a high speed raid controller.
What clients get for that is the same performance as 800 spinning disks in one small enclosure.
They get 97% cost reduction, and power is less than 1%.
-----------------------------------------
This scenario shows a client deploying and operating a 240,000 IOPs database and compares the cost benefit between using spinning disks / JBODs vs. IBM eX5 eXFlash.
To operate a 240,000 IOPs database, a client would need 800 spinning disks (300 IOPs / disk). These would be deployed in 80 JBODs (12 disks each). These would require 10 entry servers (8 JBODs per server using a ServeRAID adapter with cascading feature). RAID 5 is assumed in this example, however, since the client is maximizing on performance rather than capacity, the 33.4% degradation of capacity is ignored on both spinning disk and EXFlash calculations.
Server acquisition cost savings
10 servers * $5,000 list for each server
= $50,000 (a) vs. $0 for EXFlash since no additional servers are required
Storage acquisition cost savings
70 JBODs * $3,100 list for each JBOD
= $208,000 (b) vs. $0 for EXFlash since no JBODs are required
800 disks * $150 list for each disk
= $120,000 (c) vs. $0 for EXFlash since no disks are required
Total storage acquisition cost savings is (b) + (c) = $328,000 (d)
Datacenter real-estate cost savings
To house the servers and JBODs, the client will need 5 42U racks
5 racks * 8 square feet per rack * $115 per square foot rent (in NY) * 12 months
= $55,000 per year (e) vs. $0 for EXFlash since it is built into the database server
Datacenter power savings
(70 JBODs * 500 watts) + (10 servers * 600 watts) * 24 hours * 365 days * $0.12 / KWh
= $42,000 per year (f) vs. $0 for EXFlash database server PS powers EXFlash
Maintenance savings
Assuming 1 spinning disk failure / year
$150 list for each disk * 52 weeks * 1 failure per week
=$7,500 per year vs. $2,500 per year assuming a single SSD failure / year in EXFlash
= $5,000 per year benefit (g)
Management savings
Assuming $100,000 of server administrator costs to manage 20 servers. For 10 servers, yearly cost will be $50,000 (h)
Total savings: $823,000 over 3 years to deploy and operate the database on spinning disks vs. $20,000 on EXFlash.
Total Hardware savings (CAPEX) of $378,000
Total (OPEX) of $445,000
Database clients have told us that their database servers are fully utilized as soon as they are deployed and they just can’t keep up with growing database demands.
With traditional industry standard approach, that’s understandable. As you can see from top of this chart, when server capacity is exceeded, you must purchase, configure and deploy additional individual servers into the pool. This creates both logistical as well as technical challenges. Logistically, you need to open new purchase orders and so on. Technically, this approach duplicates certain resources and adds unnecessary hops for database access.
Again, eX5 provides a unique alternative. Only eX5 offers pay as you grow expansion for CPU and or memory. As your database grows, you can scale an eX5 system from 2-socket to 4-socket OR 4-socket to 8-socket. Alternatively, you can add MAX5 and expand just the memory without additional CPUs or software licenses.
Databases, such as SQL, will also scale with eX5 to leverage additional resources and deliver performance benefits from the scale system in just minutes.
Here is another example of the database software savings that 3690 can provide. Both the Dell R810 and R910 contain four sockets, so even with 2 of them populated in many cases the clients will have to pay the full 4-socket license, which in the Oracle and S Q L examples here means 50-60% savings with 3690. And don’t forget that with MAX5 the 3690 has twice the memory capacity of the R810 with two AND four procs as well as the R910 with only two processors.
Server Cost
US list price current as of September, 2010
Neh –EX x7560 used in config
License Cost
Microsoft SQL List Pricing as of Sept. 2010 per http://www.microsoft.com/sqlserver/2008/en/us/pricing.aspx
* Enterprise edition per processor license: $24,999
* Four processor license: $99,996
& Projected for SQL
Datacenter power costs and Systems management costs determined using IBM Server Consolidation TCO Calculator for System x and BladeCenter https://roianalyst.alinean.com/ibm_stg
You can choose to run this workload on mult systems…or can consolidate. Results in lower license cost…etc..
A single x3850 X5 is capable of consolidating 16 individual SQL databases onto one server using virtualization. For super-high availability, and for clients who are hesitant to put all of their eggs in one basket,consolidate onto two 3850 systems and have the ability to move the databases between system in the event of a PFA.
http://www.redbooks.ibm.com/abstracts/redp4661.html?Open
Neh –EX x7560 used in config
License Cost
All pricing is Microsoft SQL List Pricing as of Feb 2010 per http://www.microsoft.com/sqlserver/2008/en/us/pricing.aspx
* Enterprise edition per processor license: $24,999
* Four processor license: $99,996
& Projected for SQL
Datacenter power costs and Systems management costs determined using IBM Server Consolidation TCO Calculator for System x and BladeCenter https://roianalyst.alinean.com/ibm_stg
Clients deploy database applications on MP servers, and 4-socket HX5 is the ideal blade platform for database transactions.
4-socket HX5 is 23% less expensive for equivalent SQL transactions over Dell M910. HX5 delivers 15% greater performance per server as only HX5 supports higher performance 130W CPUs in 4-socket configuration, and delivers 2x the memory bandwidth (for more details, see FlexMem discussion in HX5 competitive presentation on SSI / PW).
Net is that it takes at least 7 Dell servers to match 6 HX5 servers. Result is significantly lower cost per database transaction – almost $300K savings in this relatively small 6 server example.
SQL Pricing
All pricing is Microsoft SQL List Pricing as of Feb 2010 per http://www.microsoft.com/sqlserver/2008/en/us/pricing.aspx
* Enterprise edition per processor license: $25K
* Four processor license: $100K
& Projected for SQL
SQL Performance
Projected TPCE performance per IBM performance group
Datacenter power savings
800W per blade * 24 hours * 365 days * $0.10 / KWh * 1.5x cooling overhead = $1000/year
Management savings
Assume $100,000 of server administrator costs to manage 20 servers. For 1 server, yearly cost will be $5K
Key Points: Databases value memory, IOPS and performance. eX5 systems lead the industry in memory per core, lead in IOPS with eXFlash and lead in overall performance with our scalability offerings.
In summary, eX5 is the only x86 platform that delivers the portfolio breath, memory technology and dense, high speed I/O technology required to break through database bottleneck, maximize database performance and minimize cost.
With MAX5, only eX5 can deliver memory required for large scale consolidation resulting in lower cost per workload from fewer servers and reduced SW license costs
With eXFlash, only eX5 can deliver the fast access to data required for high quality of service in multi user database environment
Lastly, only eX5 delivers pay-as-you-grow expansion and scales with your data sets
With mission critical workloads moving to x86, systems need to keep up. Application availability critical to clients.
Reliability and availability are important to clients because fast access to data and business processes that stay up and running make clients more efficient and profitable. The data hosted on the server needs to be accessible at any time. Business processes hosted by a server need to be up and running at all times.
Traditionally, clients looked at large-scale systems for high level of system and application availability. Now, clients expect high level of availability from their x86 systems. However, not all x86 systems are created equal. Let’s see why.
With eX5, IBM has leveraged its knowledge of designing rock-solid large-scale Mainframe systems to x86 and gone beyond industry standard to deliver the most reliable x86 platform to date.
eX5 is designed from ground up to have hardware and software work together to provide the best application availability.
Let’s begin with hardware reliability.
While there are many features in eX5 that make it highly reliable, I will specifically call out Memory ProteXion that highlights IBM’s commitment to go beyond industry standard to deliver unique customer value.
We’ve seen clients deploy 2x the memory as just few years ago…So, memory availability is critical to system availability and application availability. With features like Memory ProteXion, only IBM delivers the memory availability to ensure that an IBM system delivers the best application availability.
Memory ProteXion represents that new thinking. This is integrated into eX5, requires no additional actions from you, and works with off-the-shelf DIMMs.
<Let’s look at a short video from one of our experts to understand how Memory ProteXion works and it’s benefits>
I think of Memory ProteXion as having “run flat” tires on your car. Even when you a few flat tires on your car, the car keeps running and you get to your destination without being stranded.
Bottom line: Memory ProteXion is just one example of built-in features that makes eX5 hardware highly reliable
We have significant RAS innovations borrowed from the IBM System z product line
Predictive failure alert is used to identify problems before they become real downtime issues.
We can use this for Hard Disk Drives, memory, cpu, fan, voltage regular modules and power supplies.
This allows administrators the ability to replace hot swap components while the system is still running, which gives greater application availability
Other features you may want to mention in this discussion
QPI Light Path Diagnostic
Auto re-enable of DIMMs after replacement
Thermal monitoring
Machine Check Architecture (MCA) recovery
eX5 not only actively monitors hardware with predictive failure alerts, it can also automatically move workloads away from flagged server to with VMControl.
VMCOntrol Express is a free Systems director plugin, that actively monitors your systems. In an event of a PFA, it leverages your virtualization software’s mobility features to move virtual machines to another server. This combination of smart hardware and smart software means that even if your server were to experience any issues, you maintain maximum application availability.
Another built-in eX5 availability feature is FlexNode Failover, where if one node in a multi-node system experiences a fatal error and a shutdown, the system reboots as a single-node. After reboot the remaining system continues to operate without the failed node.
Consider a situation where your server experiences an unrecoverable failure. Here, a monolithic industry standard server will go down and stay down until manual intervention which could take hours if not days.
Again, an eX5 system has a unique built-in solution with FlexNode Failover. An eX5 system can detect a fatal error and automatically reboot with the healthy node – bringing back available resources back online in just minutes. System now continues to operate without the failed node.
The advanced RAS features of our eX5 systems gives you the piece of mind to consolidate many 2-socket servers down to a high-availability solution of two x3850 X5s.
In addition to 54% more memory reliability features and 60% more CPU reliability features, consider how many more points of failure there are to contend with when you spread your workload across these Dl 380s.The IBM advantage lets you reduce the number of systems you need, provides you more reliable hardware and gives you peace of mind with fewer components that can fail.
eX5 is a robust platform for companies to deploy business critical applications with confidence.
The combination of IBM unique eX5 hardware and smart software means that eX5 systems are less likely to go down. Plus there is built in insurance of Predictive Failure Alerts, VMControl and FlexNode Failover result in unmatched application availability.
With the combination of smart hardware and smart software, eX5 delivers the most reliable x86 platform to date.
Take the most salient points from speaker notes…
We have already seen that running more, cheaper systems is the high cost way to run a datacenter. We’ve seen that licensing costs is critical to the calculation of your value. We have also seen that performance and reliability that deliver maximum application agility and availability are essential to get most for your IT $$.
Only eX5 delivers the max memory you need for optimal CPU utilization and max database performance. Only eX5 delivers the I/O flexibility and performance you need for VM agility and fast data access. Only eX5 delivers tuned systems where hardware and software work together to deliver maximum application availability.
eX5 brings x86 systems innovation to an entirely new level of capability, one that can solve true enterprise-class business challenges.