Friday, April 29, 2011
IBM Smarter Systems for a Converged Network
Part one is this is my first posting trying out the BlogPress tool for my iPad, making a posting from a cafe.
Second part of the experiment is that the posting is pretty much my unedited, raw notes from the session.
I'd really like the feedback from my readers to know if getting this format is of value, because in future this will get the blog posted almost immediately. Or would you rather wait a few days for me to format more and pretty it up?
Let me know!
On Apr 26, 2011 I attended this IBM session, marketed for Infrastructure managers with an eye towards data centre consolidation and infrastructure convergence. I noted low attendance at 09:35, the Tuesday after Easter might have been an off date? Anyhow below are my notes taken during the session for whatever value you might all find from them.
Need for a converged infrastructure
Matthew Eastwood, group VP, IDC global platform research
The Research was dated Jan 2011, but statistical data was from Jan 2010.
What's happening at the edge that's reshaping the data centre?
Belief is that customers are looking to achieve more and additional business benefits from IT.
Over 60% of all data centre workload is on x86 platforms, but most data centres have a mixture of platforms at about 38% unix and other.
Rack is the fastest growing format, currently holding 63% of the x86 Market share, blade is at 21%, tower at 16.
Server spend on enterprise caliber data centres is down 12 pts, SMB and Mega DC are up.
Trends, efficiency, complexity, elastic scaling, off premise.
Data growth, transactions (8.4 x growth), and 10G ports (12.4x) growth by 2013 from 2005.
Customer surveys show customers feel that convergences means server and storage and virtualisation.
Average virtualisation rate is 30%, expect 50% in 12 months
Numbers of VM hosts not increasing, but number of VMs is increasing quickly year over year, ergo, new systems being deployed are being deployed as virtual, not physical.
Server hardware spending is being optimised as CapEx, but OpEx spending is getting out of control due to lack of Virtual system management. Same message we heard from Microsoft 2 weeks ago.
Consider 25 servers per admin, 35 VMs per admin
increased focus on automation and optimisation
Industry is beginning to move from consolidation focus to assured computing, and not yet ready for the maturity required for private cloud, 80% of customersbare working in these areas, only 5 in private cloud.
CMDB required to move to assured computing
Gaps in tools are, migration, capacity planning, dynamic load optimisation, application matching, power management, chargeback. Should have systems backup, health monitoring mature and in place.
Consolidate, virtualised, automate, provision...
Server workload topology
BPM: ERP, CRM, OLTP, batch
Decision. Support: data warehousing, data mining, analysis
convergence is the reigning idea, and will continue to be until IT shops are able to provide more transparent services to other business units
Scale matters in driving efficiencies
Volume economics count across the entire DC ecosystem, systems, facility, operations... Processing workload per kilowatt
Responsiveness to business drivers is more important to IT than lowering costs.
Last week, AMAZON cloud failures highlight the onus being on those who have outsourced their applications to do so responsibly. You get what you pay for, and you need to understand what you're buying and how it affects your service delivery.
Server hardware $52b, networking $62b, storage $32b
Sweet spot in the Market is convergence points between these, the application space, the SaaS space, converged infrastructure space.
Data centre needs for tomorrow
Evolutionary approach avoiding forklift upgrades
Integrated physical and virtual fabric switch management
DC convergence requiring process and management
Long term, dcs being discrete computing units.
Improved asset uptime
Faster time to Market
Flexible capacity planning
Reduced operating costs
Pay attention to time, money, and people
Workloads are the critical pivot point for cloud and convergence decisions
increase collaboration and device usage versus traditional siloed approaches
Thinking differently about x86 and infrastructure convergence
Harsh Kachhy WW Product mrkting Mgr, IBM Blade Centre
Want to move staff to being value add instead of operational costs
Workload thinking, what's the best way to deploy any workload
Shared pools of resources, improve overall utilisation
Independent scaling of resources, just what we need, when we need it
Automated management of workloads
Virtual and physical convergence leads to increased management and automation, which leads to optimisation with integrated service management
Convergence realities and critical considerations
Interoperability with existing infrastructure.
Scaling without rip and replace or forklift upgrades
FCOE, 10GbE readiness
Single vendor lock in
Best of breed solutions
Secondary effects on my infrastructure
plugged IBM Systems Director and Tivoli as delivery a tightly integrated single pane of glass view to provide the insight into systems needed to get through consolidation and into the cloud.
A managed environment can save up to 40% in management costs.
VMControl free plugin to IBM systems Director, supports VMWare, KVM, HyperV, or Xen.
Link physical alerts to Vmotion, automatically migrate VMs from the failing server
Active Energy Manager
Open Fabric Manager, for blade chassis I/O configuration and failover management
Plugged Tivoli for automated provisioning and image management.
Private cloud using Tivoli tools, IBM Bladecentre Foundation for Cloud ( IaaS), or IBM cloudburst which provides self service portal. Curious about if you can select where the data is kept and or processed...
Talk turned to hardware
eX5 technology available in blade or rack format, claims to be the only x86 platform that delivers the memory tech that allows IO boosts, claim is 22% less expensive for virtualisation. Compared 1 HX5 + MAX5 against 2 HP BL490c
IT optimisation assessment benefit, no obligation to purchase anything, but complete audit of all systems, apps, and approach to virtualisation and consolidation.
BNT, an IBM company - importance of System Networking
Shailesh Naik, NA System x Networking, IBM System and Technology Group
Claim is that networking is the key to service simplification, virtualisation and automation, converged and virtual fabrics, high density computing, and reduced costs.
TCO can be lower with 2nd vendor instead of being dependent on just one, up to 20%
IBM is re-entering the network business by acquiring Blade Network technologies
BNT RackSwitch Portfolio in use by Akami, NYSE, and others.
Claims lowest latency, power consumption, compared favourably against Cisco
Claims twice the functionality, at half the price
Virtual fabric; 20Gb per server on a 10Gb connection to the rack
Appears to be a management layer ontop of your existing management. NMotion technology converged fabric management over Ethernet.
Presenting as a FCoE solution.
Future Vision of Blades
Described the purpose of the group Blade.org
Blade.org Panel Discussion
Brent Mosbrook, Emulex
Ken Pierrehumbert, APC
Matt Eastwood, IDC
Discussed trends their respective companies are planning to bring to Market in conjunction with IBM.
10Gb on the MB for blades, moving the fabric closer to the processing.
APC plugged in-row cooling, and scalable power supply UPS, plus complete automation to feed the SPoG in tools like Tivoli
General DC trends from Matt, shifts in workload communication and collaboration, proliferation of more mobile computing p,affirms tethered more tightly to business networks. Forces outside of the dc increasing pressure to optimise, virtualise, consolidate, and increase efficiency of processing. Workload becomes more spiky and difficult to predict. Shift from code optimisation to data access and processing optimisation.
When it comes to server virtualisation, we are ahead of the global curve for densities, but in the top group in Vancouver/Canada. Claim is that few are using VMotion technology, used as plug for "NMotion"
PUE brought up to discuss with the small group. Made the point that as you increase virtualisation but don't scale back cooling or energy supply you are not improving your PUE.
Modular data centre concept brought up by Matt.
- Posted using BlogPress from my iPad
Wednesday, April 27, 2011
To the Cloud! No! Wait! Come Back! Crap!
Microsoft Align IT Tour 2011: Real World Strategies for Infrastructure Managers
First, I noticed they are using Dell laptops, prominently placed. Also, the event was streaming video live. Kind of cool.
IT as a Service, Cloud Services
Ruth Morton, technology advisor; @ruthm
Ruth started by quoting Nicholas Carr about how IT matters, how it is becoming a commodity utility.
Cited as an example of how is our world changing for IT Infrastructure Managers, the convergance of consumer technologies and enterprise applications: An Xbox Knect used in a Toronto hospital, so that prior to surgery scrubbed surgeons can interface with patient records using hand gestures. Kind of cool.
Some stats we've likely all heard before:
50% of business devices are expected to be Smart phones by 2014
84% of organisations today have a remote workforce
85% of data centre capacity on average is idle
Because we build for peak capacity but rarely hit it, but can't miss when we need it to empower the business.
70% of IT budgets is spent maintaining data centre operations. I'd challenge that stat applied here, but open for discussion!
Expectations are changing
IT is needed to be a business and productivity enabler (not new). IT is desired to partner to drive revenue, design new solutions, be a proactive business asset, and focus on best practices. Sure, OK.
So now we move the focus to Cloud Computing, the ANSWER TO ALL!! (or maybe not so much, but it makes good marketing hype)
Benefits and concepts behind Cloud Computing
- On demand self service
- Discussed illustration of private cloud for quickly provisioning test systems
- Ubiquitous network access
- Location transparent resource pooling
- Note that MS Azure (their cloud solution) does allow regional specification but not site specific
- Rapid elasticity
- Accommodating peak Usage periods on demand Measured service with pay per use
- Move business infrastructure from CapEx to OpEx
- Hold third party responsible for management of Infrastructure (hmmmm...)
Recent research in Canada by CTRC noted a shortage in specific IT skills in Canada, companies are encouraged to invest in skills base of IT folks to close the gap.
Skills deficits: cloud services architecture, virtualisation, SoA
Cloud Architect skill set: look at a biz app, determine if it makes sense to move it to the cloud, are there architectural app changes needed for cloud compatibility, provide recommendations to the organisation on best biz decisions PHP, .net, and other languages are cloud ready Systems Centre is cloud aware.
Charting a path to a vNext environment
Damir Bersinic Sr. Platform Advisor
Target stores moving 15k virtual machines to MS HyperV SCVMM 2012 announced; it manages VMware, HyperV, and Zen
For the live demo, Damir was RDP'ed into SCVMM console in Toronto.
He showed the conversion utility to move VMs from other hypervisors to HyperV. Converts SCSI boot drive to IDE. Creates a conversion script that you can leverage via PowerShell for bulk migrations taking 25-30 minutes each.
Discussed high availability architecture design, spreading clustered services across multiple virtual and physical hosts.
Executed queries while vm was being moved between nodes to illustrate no downtime during move, but that is dependent on correct architecture to start with.
A recommended best practice is to create a library of templates to speed provisioning of basic machines with standard configuration. But we need to be cautious of VM sprawl; when it is too easy to create hosts, they get created and not organised or retired.
You need process, tools, and education to manage VMs within best practices. Physical host best practices should apply to VMs.
"Virtualisation without management is more dangerous than not using virtualisation in the first place."
Tom Bittman, Gartner researcher
Self service is a new paradigm where Infrastructure Services doesn't control who makes new VMs, they create the templates for systems, and enable the "customers" to create their systems, by setting the rules under which other groups can create VMs, and being the gateway to approve the creation of that system.
Interesting ideas that are incorporated:
- Charge back model built in
- Set start and finish dates
- Launches jobs within VSSC
Question asked, "what's the business case for business users creating their own VMs?"
Answer given: It's a question of definition and semantics. Should really only be technical users.
Private cloud server platform leverages (needs) nearly the whole Systems Centre portfolio to be managed correctly. Start with standardising your directory services Virtualised with a hypervisors Standardise and automate data centre management Enable self service with a portal Evaluate what needs to be on a private cloud versus public cloud
Visual Studio 2010 contains Azure simulation platform (emulator) for developing and testing web apps without actually deploying to the cloud
Platform, database, fabric controller (middleware, service bus, access control) data sync functions in SQL Azure allows data synchronisation between private and public cloud hosts
Security and privacy in the cloud were noted by some (including me!) as a concern, the assumption seems to be made that "because it's on Azure, it's secure." David McLaren president & CEO at VRX Studios didn't actually state this, instead he told us all that his business isn't based on privacy legislation. That I can see...I specifically asked the question about how do they deal with privacy legislation concerns. All three panelists "answered" the question by saying, in effect, "don't put that kind of app or data in the public cloud." I got the feeling that these guys aren't here to answer the tough questions, either technical or business oriented. I was particularly disappointed with David as a panellist because i felt he had little of value to offer the audience and started his answer to each question with a plug for his business. Very off-putting.
Dave Kawula from TriCon consulting was able to sell the concept of private cloud self service, automation, orchestration, and governance. He was very clear on the point that you really need to architect the solution for data latency concerns, privacy, and data protection. I felt Dave had the most value to add to the conversation. He kept the conversation at a high level, but was able to provide realistic direction to the questions without getting bogged down into technical details.
How is back up and data retention managed in the cloud? Bottom line, it is your responsibility to architect a cloud solution that provides that, the public cloud doesn't provide any such service de-facto. Consider latency around replication for near-real time replication so you generally want systems in the same DC.
Question came up of performance for users comparing Office365 and traditional remote desktop technologies. The answer was yes, it will likely be a little slower. :-|