Cloud Bridging
3 views
Skip to first unread message
Rick Hebly
Oct 28, 2008, 6:53:51 AM10/28/08
to cloud-c...@googlegroups.com
Hi,
An interesting metaphore for cloud(data)centers are traditional factories, this is where raw resources are transformed into partial or complete value assets for consumers. While a factory can be very successful creating great product quality for its customers, it would be totally valueless if the factory is unable to deliver it to its customers the right way. This brings us to a major challenge for the future of cloud computing, the logistics of computing value - in all of its forms and shapes - from the datacenter to the user.
One thing is clear, the logistic vehicle for Cloud Services and Roaming Workloads- the manufactured product of cloudcenters - is the internet, the cloud itsself. Now the internets is not a static network, though its features are rather dependant on users, locations etc. Features such as bandwidth, latency, reliability, security and so on. Comparible to FedEx in a traditional manufacturer to user model, a fine grained delivery network is required to brigde the Independant Cloud Centers, Enterprise Cloud Centers, Company Branch Offices and finally the Users. All designed to ensure the quality of service for the applications and data - the product - will reach the user in a manner meeting the following criteria: Fast, Consistent, Easy, Secure, Measurable, Compliant, Reliable and Manageble for Any User at Any Time in Any Location.
In other words the highly dynamic datacenters that cloud centers are, should be complemented with an excellent delivery network for the created product to be of any value to the user. Bridging the Cloud. Although I am aware I am taking this subject to a somewhat higher level than a layer 4-7 application networking discussion, I believe Cloud Bridging will be critical for the usefulness and wide adoption of Cloud Computing.
What are the challenges and who offers the solutions?
Kind regards,
Rick Hebly
Tarry Singh
Oct 28, 2008, 7:47:51 PM10/28/08
to cloud-c...@googlegroups.com
Excellent example and cross-domain (adjacent and tangential domain) comparison. When you start doing that you know that the industrialization of the Cloud (internet thus) has just begun. Finished products also have typical quality checks and cloud products from cloud services providers will have to meet rigorous quality. The challenges are obvious (but not limited to):
- When setting standards, we need to go into other industries such a SCM giant Walmart, who have tamed this art through rigorous and varying demand/supply equation ( I am so curious how they are dealing with it all given the spikes in financial markets that is affecting the demand and supply). In any case they're learning big time now. A typical end-2-end SCM from its conception to its burial (A typical Adaptive BAT Framework needs to be developed. This is important as your CSER obligation is still very important and cannot be ignored.
- FGAC (Fine Grained Access Control), I propose a typical 1-0-1 checkpointing where a typical self-servicing model will enable not only transparency but also encourage due diligence.
- Benchmarked Metrics: " We can guarantee you quickboot apps" - Performance standards measured in a typical TPC type council, that will ensure content delivery which further goes through rigorous checkpoints when transporting and preserving data geo-locationally. Obviously this has to be linked with the SCM where something must happen to data/content/data sets (re-preserve, delete domainwide, whatever)
- Geo-locational Intelligent Session Management - How do we chargeback when we move your dataset to a location where availability is not really as per expectation. do we store your data locally (on your own device which you carry, which is CC approved and encrypts locally stored data in case of theft/loss? How do we ensure that your location (if public, a cafe in Guatemala for instance) dependence will carry what the risks are to you accessing and preserving your data?
Since CC involves humans, it will be pretty challenging finding/defining limits/standards as to who cares about their data and who doesn't.
Two companies are doing and gaining huge amounts of expertise: I'd say GOOG and AMZN. Sure there're others as well who are inching towards taming it.--
Kind Regards,
Tarry Singh
______________________________________________________________
Founder, Avastu: Research-Analysis-Ideation
"Do something with your ideas!"
http://www.avastu.com
Business Cell: +31630617633
Private Cell: +31629159400
LinkedIn: http://www.linkedin.com/in/tarrysingh
Blogs: http://tarrysingh.blogspot.com
- When setting standards, we need to go into other industries such a SCM giant Walmart, who have tamed this art through rigorous and varying demand/supply equation ( I am so curious how they are dealing with it all given the spikes in financial markets that is affecting the demand and supply). In any case they're learning big time now. A typical end-2-end SCM from its conception to its burial (A typical Adaptive BAT Framework needs to be developed. This is important as your CSER obligation is still very important and cannot be ignored.
- FGAC (Fine Grained Access Control), I propose a typical 1-0-1 checkpointing where a typical self-servicing model will enable not only transparency but also encourage due diligence.
- Benchmarked Metrics: " We can guarantee you quickboot apps" - Performance standards measured in a typical TPC type council, that will ensure content delivery which further goes through rigorous checkpoints when transporting and preserving data geo-locationally. Obviously this has to be linked with the SCM where something must happen to data/content/data sets (re-preserve, delete domainwide, whatever)
- Geo-locational Intelligent Session Management - How do we chargeback when we move your dataset to a location where availability is not really as per expectation. do we store your data locally (on your own device which you carry, which is CC approved and encrypts locally stored data in case of theft/loss? How do we ensure that your location (if public, a cafe in Guatemala for instance) dependence will carry what the risks are to you accessing and preserving your data?
Since CC involves humans, it will be pretty challenging finding/defining limits/standards as to who cares about their data and who doesn't.
Two companies are doing and gaining huge amounts of expertise: I'd say GOOG and AMZN. Sure there're others as well who are inching towards taming it.
Kind Regards,
Tarry Singh
______________________________________________________________
Founder, Avastu: Research-Analysis-Ideation
"Do something with your ideas!"
http://www.avastu.com
Business Cell: +31630617633
Private Cell: +31629159400
LinkedIn: http://www.linkedin.com/in/tarrysingh
Blogs: http://tarrysingh.blogspot.com
Stacey
Oct 29, 2008, 3:53:24 PM10/29/08
to Cloud Computing
Hi Rick,
There are several dimensions to how to bridge the datacenter to the
cloud. And it starts with your application. Just as not all
applications were meant for virtualization, or all infrastructure
components for that matter, for companies looking to do a mix
datacenter/cloud environment, they are likely going to do very similar
evaluations. Amazon's cloud is the leader right now, but in truth many
businesses have been using a variety of services from MSPs to do the
same thing for ages. Amazon has just been successful in making it very
commoditized and self-service.
Here are my perspectives on what needs to happen for the application
(peeled from one of my responses on the Larry Ellison post earlier
this month):
For scalable cloud applications you are going to need very specific
applications - not everyone is going to be Facebook. These type of
applications will share/evolve some basic principles:
- Loosely coupled components (ok, call them web services!) for
horizontal scale
- Dynamic language frameworks that can hot deploy without downtime
- Fault tolerance is applied as application logic, not hardware logic.
This goes for the data as well, but does not mean that the application
is responsible for monitoring or traffic direction (full disclosure:
my company focuses on the monitoring part of this)
- State is managed in the content and data - while the application can
be CPU intensive, it has transitory states in content to seamlessly
distribute work across other virtualized components.
For the cloud computing infrastructure tools you are going to need to
handle this elastic infrastructure - you are going to want to have
good provisioning capabilities, and application monitoring visibility
that spans the data center, virtualization guests and cloud instances.
For provisioning, RightScale is leading the pack right now with some
others like 3Tera and Scalant around (IMHO). Most of what they are
trying to tackle is the multi-cloud capabilities - as each of these
cloud providers are still emerging somewhat on those attach points.
That would be my guess anyway - their race is to be all encompassing
of the various clouds.
For the monitoring, my company Hyperic has been working on this
directly - and in particular for Amazon for a while now. So I have
much deeper perspectives here. Problems exist in a number of areas
that require a new approach to monitoring and management software that
all basically bend around the idea that an administrator is familiar
with managing on average 15-50 servers/nodes that change rarely. Once
you introduce the cloud, whether internal, external or hybrid, this
challenge now becomes more like 250-600 easily. How do I improve my
server to admin ratio, while accelerating and decelerating scale to
meet SLAs and cost-effectiveness?
1. Deployment - racking new boxes in the cloud takes an average of
63-65 seconds (per cloudstatus's observations). Placing these boxes
under management should be the same. It requires auto-discovery, auto-
inventory, platform cloning for management setup (think exposing log
files, setting service checks, and security permissions), auto-
grouping like resources to key health, and setting alert policies for
like resources.
2. Security - dealing with resources outside your firewall can require
very specific security configurations. The managed resource may talk
out, but nothing should be able to talk to it. So how do you collect
metric data, automatically update management profiles, run diagnostics
and issue corrective control actions when you can talk directly to it.
3. Application management - with lots of moving parts, and changing
numbers of them, assessing health, and drilling down into
troublespots, running diagnostics and deciding what to do seems like a
unrealistic task to stay on top of. To boot, the businesses looking to
create these environments are typically using this IT as a strategic
advantage for their business, or in fact it is their business itself.
In this regard, while many may use say an Apache server, they don't
all use it the same. Customizing the application management views for
the specific deployment, and the user role in that deployment is
critical - and needs to be easy.
4. Performance management - this runs the gamut from diagnostics to
reporting and charting to predictive analysis/capacity planning to all
the custom scripts that match the runbooks of the custom applications.
Bottomline - the main issue is consolidating that visibility of the
entire infrastructure into a single view. Breaking apart views and
using multiple tools leaves entirely too much room for manual error,
and requires additional overhead to maintain the tools that are
supposed to be helping you maintain your application. This can be the
difference from an ops team that is perpetually in firefighting mode
because their infrastructure and apps are always surprising them when
change occurs to a more productive, pro-active operations team that is
helping the business move forward.
FWIW, I can talk for hours about this - we've been working on solving
the application performance monitoring challenge across the datacenter
and Amazon's cloud for almost a full year now, and our latest release
Hyperic HQ 4.0 (due in early November) is all about solving this (feel
free to check it out - its in public beta now). We were able to
accomplish this in just a year at enterprise scale because of our
history of working with the biggest web companies on the net today and
their usage of virtualization - which powers the internal clouds. From
CNet (now CBS), Microsoft, StubHub, Ask.com, Mosso and more - we're in
there helping them tackle it.
Cloud is all about commoditizing it so it can be sold in bite sized
chunks and work in the specific environment for cloud infrastructure -
in this case Amazon.
Cheers,
-Stacey
http://www.hyperic.com
http://www.cloudstatus.com
> theusefulnessand wide adoption of Cloud Computing.
There are several dimensions to how to bridge the datacenter to the
cloud. And it starts with your application. Just as not all
applications were meant for virtualization, or all infrastructure
components for that matter, for companies looking to do a mix
datacenter/cloud environment, they are likely going to do very similar
evaluations. Amazon's cloud is the leader right now, but in truth many
businesses have been using a variety of services from MSPs to do the
same thing for ages. Amazon has just been successful in making it very
commoditized and self-service.
Here are my perspectives on what needs to happen for the application
(peeled from one of my responses on the Larry Ellison post earlier
this month):
For scalable cloud applications you are going to need very specific
applications - not everyone is going to be Facebook. These type of
applications will share/evolve some basic principles:
- Loosely coupled components (ok, call them web services!) for
horizontal scale
- Dynamic language frameworks that can hot deploy without downtime
- Fault tolerance is applied as application logic, not hardware logic.
This goes for the data as well, but does not mean that the application
is responsible for monitoring or traffic direction (full disclosure:
my company focuses on the monitoring part of this)
- State is managed in the content and data - while the application can
be CPU intensive, it has transitory states in content to seamlessly
distribute work across other virtualized components.
For the cloud computing infrastructure tools you are going to need to
handle this elastic infrastructure - you are going to want to have
good provisioning capabilities, and application monitoring visibility
that spans the data center, virtualization guests and cloud instances.
For provisioning, RightScale is leading the pack right now with some
others like 3Tera and Scalant around (IMHO). Most of what they are
trying to tackle is the multi-cloud capabilities - as each of these
cloud providers are still emerging somewhat on those attach points.
That would be my guess anyway - their race is to be all encompassing
of the various clouds.
For the monitoring, my company Hyperic has been working on this
directly - and in particular for Amazon for a while now. So I have
much deeper perspectives here. Problems exist in a number of areas
that require a new approach to monitoring and management software that
all basically bend around the idea that an administrator is familiar
with managing on average 15-50 servers/nodes that change rarely. Once
you introduce the cloud, whether internal, external or hybrid, this
challenge now becomes more like 250-600 easily. How do I improve my
server to admin ratio, while accelerating and decelerating scale to
meet SLAs and cost-effectiveness?
1. Deployment - racking new boxes in the cloud takes an average of
63-65 seconds (per cloudstatus's observations). Placing these boxes
under management should be the same. It requires auto-discovery, auto-
inventory, platform cloning for management setup (think exposing log
files, setting service checks, and security permissions), auto-
grouping like resources to key health, and setting alert policies for
like resources.
2. Security - dealing with resources outside your firewall can require
very specific security configurations. The managed resource may talk
out, but nothing should be able to talk to it. So how do you collect
metric data, automatically update management profiles, run diagnostics
and issue corrective control actions when you can talk directly to it.
3. Application management - with lots of moving parts, and changing
numbers of them, assessing health, and drilling down into
troublespots, running diagnostics and deciding what to do seems like a
unrealistic task to stay on top of. To boot, the businesses looking to
create these environments are typically using this IT as a strategic
advantage for their business, or in fact it is their business itself.
In this regard, while many may use say an Apache server, they don't
all use it the same. Customizing the application management views for
the specific deployment, and the user role in that deployment is
critical - and needs to be easy.
4. Performance management - this runs the gamut from diagnostics to
reporting and charting to predictive analysis/capacity planning to all
the custom scripts that match the runbooks of the custom applications.
Bottomline - the main issue is consolidating that visibility of the
entire infrastructure into a single view. Breaking apart views and
using multiple tools leaves entirely too much room for manual error,
and requires additional overhead to maintain the tools that are
supposed to be helping you maintain your application. This can be the
difference from an ops team that is perpetually in firefighting mode
because their infrastructure and apps are always surprising them when
change occurs to a more productive, pro-active operations team that is
helping the business move forward.
FWIW, I can talk for hours about this - we've been working on solving
the application performance monitoring challenge across the datacenter
and Amazon's cloud for almost a full year now, and our latest release
Hyperic HQ 4.0 (due in early November) is all about solving this (feel
free to check it out - its in public beta now). We were able to
accomplish this in just a year at enterprise scale because of our
history of working with the biggest web companies on the net today and
their usage of virtualization - which powers the internal clouds. From
CNet (now CBS), Microsoft, StubHub, Ask.com, Mosso and more - we're in
there helping them tackle it.
Cloud is all about commoditizing it so it can be sold in bite sized
chunks and work in the specific environment for cloud infrastructure -
in this case Amazon.
Cheers,
-Stacey
http://www.hyperic.com
http://www.cloudstatus.com
Alexis Richardson
Oct 29, 2008, 4:30:46 PM10/29/08
to cloud-c...@googlegroups.com
Hi Rick, all,
On the security point that Stacey described so clearly, we at
CohesiveFT have announced a new product this week. It is called
VPN-Cubed: www.cohesiveft.com/vpncubed
This enables you to control security for cloud environments especially
bridging from the enterprise to the cloud.
FYI - here is some industry comment:
* http://www.cloudave.com/link/vpn-cubed-%25e2%2580%2593-cloud-is-ready-for-the-enterprise
* http://www.infoq.com/news/2008/10/cloud-vpn
* http://blogs.zdnet.com/SAAS/?p=601
Cheers
alexis
William Louth (JINSPIRED.COM)
Oct 29, 2008, 5:45:23 PM10/29/08
to cloud-c...@googlegroups.com
Sounds like your putting lipstick on a pig .... taking
the same old IT management approach to deployment and provisioning
calling it cloud computing management ( "cloud are you there?" "let me
ping the cloud") if it can be done quicker than today like in
"63-65 seconds".
Maybe we need to have another rule in defining clouding computing. If you already know or can discover the number of @stacey "boxes, moving parts" running your application then you are not in the clouds - BECAUSE YOU CAN SEE!!!! You are up in the sky, high as a kite and pumping gases out like there was no acquisition tomorrow.
I would like to lay claim to "atmosphere computing" ...delivering the gas to keep your applications running (and employees).
Cloud computing customers do not want to manage the cloud this is secondary to their goals or a means to an ends with the current state of technology. They need a complete different management model and set of tools and approaches (if at all) than those running or reselling some cloud computing platform or service.
William
Maybe we need to have another rule in defining clouding computing. If you already know or can discover the number of @stacey "boxes, moving parts" running your application then you are not in the clouds - BECAUSE YOU CAN SEE!!!! You are up in the sky, high as a kite and pumping gases out like there was no acquisition tomorrow.
I would like to lay claim to "atmosphere computing" ...delivering the gas to keep your applications running (and employees).
Cloud computing customers do not want to manage the cloud this is secondary to their goals or a means to an ends with the current state of technology. They need a complete different management model and set of tools and approaches (if at all) than those running or reselling some cloud computing platform or service.
William
Stacey
Oct 31, 2008, 2:46:16 PM10/31/08
to Cloud Computing
Hi all,
Many apologies but in rereading my post, I realize I made a mistake in
spelling one of the vendor's names I mentioned below. Scalant should
really be Scalent, as in http://www.scalent.com. In an emerging market
with a lot of vendors - they definitely appreciate having their name
spelled correctly so they can be found later! Sorry about that
Scalent!
On Oct 29, 12:53 pm, Stacey <staceyeschnei...@yahoo.com> wrote
Many apologies but in rereading my post, I realize I made a mistake in
spelling one of the vendor's names I mentioned below. Scalant should
really be Scalent, as in http://www.scalent.com. In an emerging market
with a lot of vendors - they definitely appreciate having their name
spelled correctly so they can be found later! Sorry about that
Scalent!
On Oct 29, 12:53 pm, Stacey <staceyeschnei...@yahoo.com> wrote
Reply all
Reply to author
Forward
0 new messages