CROSS-CLOUD MANAGEMENT AND TROUBLESHOOTING
BACKGROUND
[0001] Datacenters provide servers for running large applications. Enterprises often use
datacenters to run core business functions such as sales, marketing, human resources,
billing, product catalogs, and so forth. Datacenters may also run customer-facing
applications, such as web sites, web services, email hosts, databases, and many other
applications. Datacenters are typically built by determining an expected peak load and
providing servers, network infrastructure, cooling, and other resources to handle the peak
load level. Datacenters are known for being very expensive and for being underutilized at
non-peak times. They also involve a relatively high management expense in terms of both
equipment and personnel for monitoring and performing maintenance on the datacenter.
Because almost every enterprise uses a datacenter of some sort, there are many redundant
functions performed by organizations across the world.
[0002] Cloud computing has emerged as one optimization of the traditional datacenter.
A cloud is defined as a set of resources (e.g., processing, storage, or other resources)
available through a network that can serve at least some traditional datacenter functions
for an enterprise. A cloud often involves a layer of abstraction such that the applications
and users of the cloud may not know the specific hardware that the applications are
running on, where the hardware is located, and so forth. This allows the cloud operator
some additional freedom in terms of rotating resources into and out of service,
maintenance, and so on. Clouds may include public clouds, such as MICROSOFT TM
Azure, Amazon Web Services, and others, as well as private clouds, such as those
provided by Eucalyptus Systems, MICROSOFT TM, and others. Companies have begun
offering appliances (e.g., the MICROSOFT TM Azure Appliance) that enterprises can
place in their own datacenters to connect the datacenter with varying levels of cloud
functionality.
[0003] Enterprises with datacenters incur substantial costs building out large
datacenters, even when cloud-based resources are leveraged. Enterprises often still
planned for "worst-case" peak scenarios and thus include an amount of hardware at least
some of which is rarely used or underutilized in terms of extra processing capacity, extra
storage space, and so forth. This extra amount of resources incurs a high cost for little
return. Customers using cloud based computing on premise expect to be able to use
capacity in another compatible cloud (e.g., a second instance of their own in another
location, Microsoft's public cloud, and so forth) for peak capacity times, for disaster
recover scenarios, or just for capacity management. Doing so is much less expensive than
building out for the worst-case scenario and then doubling for redundancy. In addition,
they expect to be able to manage (e.g., troubleshoot, operate) applications split across
multiple clouds. Today, applications, cloud management, and troubleshooting do not
operate across clouds or other datacenters.
SUMMARY
[0004] A cloud management system is described herein that provides the ability for an
application to span two or more clouds (which may be across great distances), while
allowing operation, management, and troubleshooting of the distributed application as a
single application. The system provides infrastructure that communicates across
datacenters for execution and for centralizing knowledge of instances of an application
that are running at different locations. In some cases, the system provides a computing
appliance that an enterprise can place in its own private datacenter that allows an
administrator to distribute at least some application loads to a public cloud or other
separate locations, while providing unified management via the computing appliance. The
infrastructure provided by the system monitors both the application and connections
between the clouds, with intelligence to know if issues are within the application or
because of the connection between the clouds. The system coordinates management
functions across multiple cloud platforms/locations. If an administrator wants to debug
the application, the system allows live debugging at the correct location through a
seamless, unified interface. Thus, the cloud management system creates a single
monitoring and troubleshooting interface and knowledge and execution "fabric" across
multiple clouds so that applications spread across multiple clouds can be monitored,
managed, and debugged more easily.
[0005] This Summary is provided to introduce a selection of concepts in a simplified
form that are further described below in the Detailed Description. This Summary is not
intended to identify key features or essential features of the claimed subject matter, nor is
it intended to be used to limit the scope of the claimed subject matter.
BRIEF DESCRIPTION OF THE DRAWINGS
[0006] Figure 1 illustrates an application running in two clouds with associated
management infrastructure, in one embodiment.
[0007] Figure 2 is a block diagram that illustrates components of the cloud management
system, in one embodiment.
[0008] Figure 3 is a flow diagram that illustrates processing of the cloud management
system to handle a request from a management tool to access data from distributed
application instances, in one embodiment.
[0009] Figure 4 is a flow diagram that illustrates processing of the cloud management
system to report data back from and handle troubleshooting requests at a location of a
remote application instance, in one embodiment.
DETAILED DESCRIPTION
[0010] A cloud management system is described herein that provides the ability for an
application to span two or more clouds (which may be across great distances), while
allowing operation, management, and troubleshooting of the distributed application as a
single application. The system provides infrastructure that communicates across
datacenters for execution and for centralizing knowledge of instances of an application
that are running at different locations. For example, the system may centralize logging,
performance tracking, and other management functions, regardless of where the
application is running. In some cases, the system provides a computing appliance that an
enterprise can place in its own private datacenter that allows an administrator to distribute
at least some application loads to a public cloud or other separate locations, while
providing unified management via the computing appliance.
[0011] The infrastructure provided by the cloud management system monitors both the
application and connections between the clouds, with intelligence to know if issues are
within the application or because of the connection between the clouds. The system
coordinates management functions across multiple cloud platforms/locations (from one
cloud's infrastructure, tasks are coordinated to execute across two or more clouds). If an
administrator wants to debug the application, the system allows live debugging at the
correct location through a seamless, unified interface. Thus, the cloud management
system creates a single monitoring and troubleshooting interface and knowledge and
execution "fabric" across multiple clouds so that applications spread across multiple
clouds can be monitored, managed, and debugged more easily.
[0012] Figure 1 illustrates an application running in two clouds with associated
management infrastructure, in one embodiment. In some embodiments, the cloud
management system involves the application (and/or the administrator) using
infrastructure in one cloud that has the data/access to the data at all locations to be able to
fully monitor/troubleshoot the application. As an example, consider an application with
instances running in two clouds, cloud 110 and cloud 150 as shown in Figure 1. Cloud
110 includes a MICROSOFT TM Azure appliance instance 120 that includes
infrastructure 130. The appliance instance 1 0 includes application instance 125 that is
running role 140 and role 145. A second cloud 150 includes application instance 155 that
is running role 160and role 170. The second cloud 150 also includes infrastructure 180.
The appliance instance 120 knows about each of the roles and that they are part of the
same application. The infrastructure plumbing at each location allows the appliance
instance 120 to retrieve information about role 160 and role 170 executing at the second
cloud 150. The system can distribute either individual roles, whole applications, or both.
With all of the management data (e.g., logs from the applications, machines, and
infrastructure), the system can assess the health of the application just as if all of the roles
were local by applying pre-defined health rules. The system can also see the infrastructure
health across both locations as well as the connection 190 in between to assess if a
problem is occurring with the application or the infrastructure/network.
[0013] Similarly, when automated or manual troubleshooting or remediation steps are
needed, the infrastructure 130 in cloud 110 can coordinate with the infrastructure 180 in
cloud 150 to provide troubleshooting and debugging support. For example, the system
fabric can reach across locations to execute an application wide update, shutdown, and so
forth. Those of ordinary skill in the art will recognize numerous ways to perform crosslocation
control. For example, infrastructure 130 may directly control infrastructure 180,
infrastructure 130 may request infrastructure 180 to execute on infrastructure 130's behalf,
and so forth. Likewise, with operator/administrator troubleshooting tools (e.g., monitoring
visualization, alerting, log and configuration data viewing, and so on), the location of the
applications and infrastructure is available and logically displayed, but does not involve
separate tools and mental gymnastics from the administrator to put together. For instance,
when troubleshooting and viewing data on all roles, if the administrator 105's next step is
using one or more tools 195 to view the application's logs or start a remote session to the
role instance, the system connects the administrator 105 directly, regardless of at which
location the role is residing.
[0014] The design of the cloud management system provides simplified and consistent
running of a service across multiple clouds/location. The system moves the definition of
"a computing resource" from a server, beyond a datacenter to a portion of the internet (the
datacenters and the connection between them). This allows service level agreements
(SLAs) to be defined, monitored, and managed at the service level - which is what service
owners often care most about.
[0015] In some embodiments, the cloud management system works in cooperation with
a cloud migration system that seamlessly migrates applications from one location to
another as needed, called bursting. The cloud migration system provides capacity
management and disaster recovery by detecting peak load conditions and automatically
moving computing to another source (and back) and by providing computing across two or
more clouds and moving completely to one in the case of a disaster at one site. This
allows enterprises to plan for local resources for a sustained level of load and to leverage
cloud-based resources for peak or other unusual loads. In many cases, an enterprise's
business is such that a particular time of year is busier, and extra resources may only be
needed during those times. For example, tax-planning enterprises are particular busy in
mid-April, e-commerce sites experience holiday rushes around Thanksgiving and
Christmas, and so on. The cloud migration system monitors loads within a datacenter and
detects a threshold that indicates that the current load is nearing the datacenter' s capacity.
For example, the system may monitor central processing unit (CPU) usage, memory
usage, storage usage, network bandwidth, and other metrics to determine how well the
datacenter is handling the current load. The system may also observe trends (e.g., a rate of
acceleration of resource usage) to determine whether the threshold has been or will soon
be reached.
[0016] Upon detecting that the threshold will be reached, the cloud migration system
facilitates an orderly move of at least some datacenter load to another datacenter or cloudbased
resources. For example, the system may migrate some peak load to a public cloud.
Because cloud pricing models may vary, the system may factor cost into the decision. For
example, the system may prefer to host as much load as possible at the enterprise
datacenter to reduce cost, while leveraging cloud resources only to the extent needed to
satisfy client requests. The system may also provide management and monitoring tools
that provide a consistent experience for information technology (IT) personnel regardless
of where particular loads are run (e.g., locally within the enterprise or publicly using a
cloud). The system may also provide planning tools to help decide appropriate workloads
or applications for moving to other resources during high loads. For example, applications
may have various compliance/regulatory or networking/design limitations that make them
more or less suitable for migration. The system can also be used as a disaster recovery
architecture at a datacenter/network level to manage fast workload transition in case of
disaster. If a datacenter resource permanently fails, the system can quickly and efficiently
migrate additional load to the cloud or other resources so that clients of the datacenter are
unaffected or are less affected by the failure. Thus, the cloud migration system allows
enterprises to build smaller and more efficient datacenters that leverage other resources for
rare extra loads.
[0017] The cloud management system works with the cloud migration system to provide
seamless management and troubleshooting as applications are migrated from one location
to another. As described above, the cloud migration system may move resources between
a datacenter and the cloud on a temporary (i.e., bursting) or permanent (i.e., disaster
recovery) basis. Temporary movements include bursting an application or other load for a
short time period to handle a peak or other high load that exceeds the datacenter' s
capacity. A temporary movement may include bursting an entire application or splitting
the application's load across two or more locations. Permanent movements include
longer-term migration of loads due to a failure of hardware in the datacenter, a more
sustained increase in capacity needs, a desire to globally distribute an application with
dynamic load balancing, and so forth. Following are several example scenarios in which
the system may be used by an enterprise.
[0018] In the first example, an enterprise bursts application load to a public cloud to
manage capacity. The business decision maker (i.e., CEO, CFO, or VP Marketing/sales)
and datacenter systems administrator decide it would be more cost effective and provide a
better customer experience to burst some work to the public cloud at their top three peak
day level of use/traffic per year, and maintain their own datacenter (potentially with a
cloud appliance) at their peak monthly usage level. They sign business agreements with
the cloud provider to burst work into the cloud and project estimates of when and how
much work that would be. Their account is set up and the information entered into the
cloud appliance. During a planning phase, the administrator runs a test with a test
application from the cloud provider that ensures the connection is working properly. The
administrator then sets the capacity values (e.g., threshold) for initiating bursting of
applications that keep capacity at the specified level in a capacity management tool. The
administrator goes into the tool to further specify the applications that are eligible to move
in this situation (e.g., no regulatory issues with temporary movement, good technical fit).
[0019] The day comes when use exceeds the limits and the system automatically moves
applications to the public cloud. Alerts are thrown in the monitoring/usage systems when
capacity is within 5% of the bursting being initiated, when the system does burst, what the
system bursts, and when the system brings the applications back. An explicit log is kept
of all compute resources and/or storage moved and the administrator is alerted to go to
their public cloud account for billing. A review of the bursting parameters and
applications tagged as movable is reviewed in the regular capacity planning meetings with
in the enterprise datacenter group and management.
[0020] In a second example, an enterprise splits applications across clouds to manage
capacity. This scenario is similar to the scenario above except for the type of application
moved is more complex, so is split out to prioritize differently. The company decides to
have a relationship with the cloud provider for splitting applications into the cloud (a form
of bursting). In this large application was pre-identified as a bursting candidate.
When the capacity reaches the threshold, 50 of the 100 worker instances are automatically
moved to the public cloud. The application is now split across two appliance instances or
cloud instances, with all monitoring and billing data being sent to the starting instance so it
can be centrally managed. A cloud appliance in the enterprise's own datacenter has
troubleshooting tools to help debug possible issues of split applications (e.g., networking
issues, network bandwidth/latency issues, fabric communication, and so on). When the
capacity situation has subsided on the appliance, the 50 worker instances are moved back
to the appliance and it is a normally functioning application again.
[0021] In another example, a cloud provider decides to burst from one cluster to another.
The public cloud capacity planning team decides that a cluster in the Chicago datacenter is
critically full but wants to maintain high utilization. They set up bursting to an
underutilized cluster in a West coast datacenter when utilization gets to 90%. The
administrator goes into the capacity management tool and chooses appropriate
customers/applications (e.g., with low data usage) to be move candidates. The day comes
when the usage of the Chicago cluster reaches the threshold and the system automatically
moves the selected applications (e.g., 10% of the cluster's apps) to the West coast
datacenter for one day. As usage returns below the threshold, the system moves the
applications back to Chicago. The system notifies a designated monitoring team
proactively of the burst to be able to answer customer questions.
[0022] In another example, the system is used for cross-cloud portfolio management.
An enterprise decides that to manage capacity efficiently on their cloud appliance, they
want to put all variable demand apps in a public cloud and their constant demand
applications in the appliance or local datacenter resources (and thus be able to run the
appliance at higher utilization). While they want their computing resources split, they still
want a global view across all of their application's health, to have their application
developers manage applications the same way, and to maintain a single view of
departmental billing across both (e.g., what costs to allocate to the consumer sales groups,
internal IT, B2B sales, and so forth). The enterprise is able to set up the aggregation
accounts with the public cloud with the same groups as the appliance and get billing data
to integrate on their side. Similarly, they are able to get application-programming
interface (API) access to the public cloud monitoring data for the platform their apps are
running on as well as application level monitoring, so their network operation center
(NOC) has a complete and consistent view of the state of the computing activity of the
enterprise.
[0023] In another example, an enterprise sets up a globally distributed application with
dynamic load balancing. An enterprise customer wants to manage capacity across two or
more cloud instances and has a significant amount of their load in independent but
geographically distributed instances (e.g., Bing search with a US and UK datacenter that
both serve German queries). Under normal circumstances, a global traffic manager sends
50% traffic to each location. When load gets high at the primary location, the system
instructs the load balancer to send 75% of the traffic to the UK system, thus freeing up
capacity from the US cloud instance, bringing it to acceptable levels. When capacity
returns to normal, the system tells the load balancer to return to the 50/50 split. A
variation of this is for the public cloud to be used as a secondary datacenter (with say 1%
of load, the customer's site with the appliance to be the other 99%). In the case of a
disaster or other reason to move load from the customer's site, 100% of traffic is shifted to
the public cloud.
[0024] In another example, an enterprise has reached its datacenter' s capacity and needs
extra computing resources, but does not yet have the available capital to expend to expand
the datacenter. In this case, the company can use a public cloud for spillover until they
can get the hardware purchase completed.
[0025] Figure 2 is a block diagram that illustrates components of the cloud management
system, in one embodiment. The system 200 includes a location management component
210, location data store 220, tool interface component 230, one or more management tools
240, a data migration component 250, a troubleshooting component 260, and a billing
component 270. Each of these components is described in further detail herein.
[0026] The location management component 210 manages information about multiple
datacenter locations at which instances of an application are running. The component 210
includes information describing how to reach each location, connections available for
retrieving management information, user accounts to use for each location with associated
security credentials, application and datacenter components from which to gather
troubleshooting information and send troubleshooting commands, and so forth. The
location management component 210 receives information describing any migration of
application loads or bursting from one datacenter/cloud to another and updates the
managed information so that the component 210 has a complete picture of all of the
locations where the application is running. This allows the system 200 to present the
complete picture and to make management of applications uniform, no matter where or at
how many locations the applications are running. As conditions change and applications
are distributed, the location management component 210 can present management tools
with a comprehensive set of management data.
[0027] The location data store 220 stores information that describes locations at which
instances of the application are running. The data store 220 may include one or more files,
file system, hard drives, databases, cloud-based storage services, or other facilities for
persisting information between sessions with the system 200. The stored information may
include connection information, user roles, sources of management data, available log
files, and any other information related to management or troubleshooting of applications
distributed to multiple locations.
[0028] The tool interface component 230 provides an interface to the system 200
through which one or more tools can access management and troubleshooting information
for the application. The interface may include one or more web pages, web services,
application-programming interfaces (APIs), or other interfaces through which an
administrator or tools can directly or programmatically access management and
troubleshooting information of the system 200. In some embodiments, the tool interface
component 230 provides an initial connection point for tools to access information related
to the application at a cloud-computing appliance located within an enterprise's private
datacenter. The appliance may manage migration and distribution of application loads to a
public cloud or other datacenter, and provides a central point of contact for tools that
gather management information or provide application troubleshooting.
[0029] One or more management tools 240 connect to the tool interface component 230
to access management information or perform application troubleshooting. The tools may
include log viewers, reporting tools, debugging tools, or other tools that display
information about or assist in resolving problems with a running application. The
management tools 240 may include tools designed to work with a local application, and
the system 200 provides the tools with information describing a distributed application
running at multiple locations without the tool's knowledge. This allows existing tools that
administrators rely upon to be used even as automatic application load migration is
introduced into a datacenter or cloud. In other cases, tools may be specifically written to
understand distributed applications and to provide specific management information or
troubleshooting related to multiple locations. The tool interface component 230 may
provide multiple interfaces through which management tools 240 connect to the system
200 using paradigms that are understood by each tool.
[0030] The data migration component 250 migrates management information at one or
more remote locations where the application is running back to a home location of the
application. The home location may include a private datacenter, location or a cloudcomputing
appliance, or other location where the application normally runs under steady
conditions. Upon hitting a certain level of load (e.g., peak or periodic bursts), the
application may migrate some load to one or more other datacenters or public clouds to
help satisfy client requests. These other locations generate management data, such as log
files, transaction data, and so on, just like the home location, and the data migration
component 250 migrates this data back to the home location or provides access to the data
from the home location, so that the management tools 240 can provide a comprehensive
picture of the application's activity to administrators.
[0031] The troubleshooting component 260 performs troubleshooting tasks on the
application at one or more locations. Troubleshooting may include debugging, processing
test data, or other forms of determining whether an application is operating correctly.
Troubleshooting is generally well understood at the home location, but becomes more
complex as an application begins to span multiple datacenters or clouds. The cloud
management system 200 isolates management tools 240 and administrators from this
complexity by providing a uniform interface through which tools and administrators
access management information and perform troubleshooting at multiple locations. Thus,
if a management tool allows an administrator to place a breakpoint at or receive trace
information from a particular piece of application code at the home location, then the
troubleshooting component 260 makes it just as easy to do so at a remote cloud-based
instance of the application. The tools and administrator may even be unaware of all of the
locations where the application is running, but can still perform management tasks as if the
application were only executing at the home location.
[0032] The billing component 270 reports billing information related to one or more
locations where an application is running. One common management task is to manage
computing costs, and public clouds often charge based on metrics related to workload
(e.g., computing time, storage space used, and so forth). It may be useful for an
administrator to gather a picture of the costs that application instances are incurring at
various locations, and the cloud management system 200 can optionally provide the billing
component 270 to gather this type of information so that the information can be reported
through management tools and reports.
[0033] The computing device on which the cloud management system is implemented
may include a central processing unit, memory, input devices (e.g., keyboard and pointing
devices), output devices (e.g., display devices), and storage devices (e.g., disk drives or
other non-volatile storage media). The memory and storage devices are computerreadable
storage media that may be encoded with computer-executable instructions (e.g.,
software) that implement or enable the system. In addition, the data structures and
message structures may be stored or transmitted via a data transmission medium, such as a
signal on a communication link. Various communication links may be used, such as the
Internet, a local area network, a wide area network, a point-to-point dial-up connection, a
cell phone network, and so on.
[0034] Embodiments of the system may be implemented in various operating
environments that include personal computers, server computers, handheld or laptop
devices, multiprocessor systems, microprocessor-based systems, programmable consumer
electronics, digital cameras, network PCs, minicomputers, mainframe computers,
distributed computing environments that include any of the above systems or devices, set
top boxes, systems on a chip (SOCs), and so on. The computer systems may be cell
phones, personal digital assistants, smart phones, personal computers, programmable
consumer electronics, digital cameras, and so on.
[0035] The system may be described in the general context of computer-executable
instructions, such as program modules, executed by one or more computers or other
devices. Generally, program modules include routines, programs, objects, components,
data structures, and so on that perform particular tasks or implement particular abstract
data types. Typically, the functionality of the program modules may be combined or
distributed as desired in various embodiments.
[0036] Figure 3 is a flow diagram that illustrates processing of the cloud management
system to handle a request from a management tool to access data from distributed
application instances, in one embodiment. Beginning in block 310, the system receives
from a management tool a request to access management data related to an application
running instances in one or more datacenters. For example, a performance-monitoring
tool may request status information describing how many client requests the application is
handling, resource usage of the application, or other information from the application.
The system may receive the tool request through an API the system exposes to tools for
requesting management data. The API may comprise a uniform interface for accessing
management data irrespective of where or at how many locations application instances are
running.
[0037] Continuing in block 320, the system identifies one or more types of management
data that satisfy the received request. For example, the system may determine that the
request asks for log information that is produced by each instance of the application.
Identifying the requested data allows the system to determine which information to gather
from each application instance or whether the data is already gathered locally from data
pushed to a central location by each application instance.
[0038] Continuing in block 330, the system determines a distribution of the application
that includes two or more instances of the application. The distribution determines where
the application is running and where the system will find management data to satisfy the
request. The system may include a data store that tracks information describing each burst
or other migration of application load to and from other datacenters, so that the system is
aware of each location where application instances are running. Upon receiving the
management tool request, this information allows the system to determine from where to
gather management data.
[0039] Continuing in block 340, the system gathers management data to satisfy the
request from each distributed application instance. The instances may include an instance
in a local private datacenter, a remote private datacenter, a private cloud computing
facility, a public cloud computing facility, spare resources offered by other private
datacenters, and so on. The system contacts each instance of the application or accesses
previously sent information from each instance that contains information (such as
performance data, failures, and so forth) to satisfy the received management tool request.
[0040] Continuing in block 350, the system optionally sends one or more
troubleshooting commands to one or more remote application instances. For example, if
one location is experiencing failures, the administrator may use a management tool to
request additional trace information, to send one or more test requests, or to perform other
types of debugging. The remote application instances carry out the troubleshooting
commands and report requested data back to a central location where the management tool
can access the information.
[0041] Continuing in block 360, the system unifies gathered data to provide a uniform
response to the received management tool request. In this way, management tools need
not be written to include an understanding of the various potential distributions of
applications managed by the tools. The system can thus freely migrate the application
from location to location or to multiple locations as needed to handle application loads,
while still providing administrators with a straightforward management and
troubleshooting experience.
[0042] Continuing in block 370, the system reports the gathered and unified
management data in response to the received management tool request. The system may
send the data through the interface on which the request was received or through a
notification interface or other facility for providing data to the tool. After block 370, these
steps conclude.
[0043] Figure 4 is a flow diagram that illustrates processing of the cloud management
system to report data back from and handle troubleshooting requests at a location of a
remote application instance, in one embodiment. Beginning in block 410, the system
receives management data at a remote application instance handling a portion of load
generated by requests from clients of the application. The management data may include
performance data, log information, error details, statistical information, sales history, or
other indications of application operation useful for management the application.
[0044] Continuing in block 420, the system determines a home location of the
application where an administrator can access management data reported by multiple
instances of the application running at distributed remote locations. The application
instance may receive configuration information from the home location upon creation of
the instance that specifies where the home location can be contacted and that the
application instance is a remote instance of the application. The system may migrate
applications to multiple locations to handle peak loads, perform low priority tasks at
locations where processing is off-peak and thus cheaper, or for other reasons determined
by an administrator. The application may have a home location that is where the
application normally runs and may handle peak or other loads at one or more distributed
remote locations.
[0045] Continuing in block 430, the system sends the received management data from
the remote application instance to the determined home location of the application. The
system may periodically migrate data generated at distributed instances back to the home
location so that management data is available in one place at the home location for the
convenience of administrators and management tools. The system may also migrate data
on demand or as requested by various tools (see, e.g., Figure 3). In some cases, the system
may burst application loads to remote locations for short durations and then collect
information related to the application's execution when the loads are migrated back to the
home location and the remote instances are finished.
[0046] Continuing in block 440, the system optionally receives a troubleshooting
request from a management tool run at the home location to troubleshoot the remote
application instance. The troubleshooting requests may include debug breakpoints, a
request for detailed trace information, or other commands or requests to perform
troubleshooting actions.
[0047] Continuing in block 450, the system performs one or more troubleshooting
actions in response to the received troubleshooting request. The action may include
setting a debug breakpoint, turning up a logging level, sending test data to the application,
or performing any other action specified by the request for determining whether the
application is operating properly.
[0048] Continuing in block 460, the system sends a troubleshooting result to the home
location in response to the received troubleshooting request. By providing a facility for
executing troubleshooting commands remotely, the system allows a troubleshooting tool
operating at the home location to troubleshoot application instances no matter where the
instances are running, and allows the system to seamlessly migrate instances of the
application to various locations without interrupting the ability of an administrator to
manage and troubleshoot the application. After block 460, these steps conclude.
[0049] In some embodiments, the cloud management system migrates application load
by modifying domain name service (DNS) records. The system may modify a DNS server
to point incoming client requests to one or more new destination Internet Protocol (IP)
addresses to direct loads away from a source datacenter over to a target datacenter/cloud.
A global traffic manager (GTM) often points clients to the nearest server for handling their
requests, and these solutions can be modified to redirect traffic based on load or other
conditions. Thus, when one datacenter becomes overloaded or near capacity, the system
may inform the GTM to direct at least some client requests to a new location that can
handle the excess load. Similarly, the system may provide a DNS or other address to
which management tools can address management requests and be connected to
application instances no matter where they reside.
[0050] In some embodiments, the cloud management system migrates log and other data
back from the target computing resources after migration conditions have eased. For
example, following a period of peak load, the system may migrate all application loads
back to the original datacenter, and may pull information generated at the target
datacenter, such as application logs, back to the original datacenter for later analysis. For
some applications, tracking client requests may be a matter of regulatory compliance or
simply useful for debugging and reporting. In either case, consolidating the logs at the
source location may be part of a successful migration back to the source location.
[0051] In some embodiments, the cloud management system allocates a dynamically
variable amount of application load between a source computing resource and one or more
target computing resources. For example, the system may dynamically route requests to
keep the source computing resource at or near full capacity and only send requests out to
external computing resources that the source computing resource could not successfully
handle. Such decisions may be a matter of cost, data safety, or other considerations to
migrate out as little application load as needed or to place application loads where they can
be performed cheapest or most efficiently. In some cases, the decisions may be based on
regulatory requirements of applications. For example, applications subject to healthcare or
other recordkeeping laws may have restrictions about the datacenters/clouds in which they
can operate.
[0052] In some embodiments, the cloud management system provides various options
for disaster recovery. In some cases, the system may enlist resources at an external
datacenter to monitor a main datacenter for outages. If the external datacenter becomes
unable to reach the main datacenter, then the external datacenter may determine that a
disaster has occurred and move application loads to the external datacenter. In past
systems, it was typical for an organization to maintain 200% of needed capacity (at
substantial expense) in order to successfully handle disasters. With the cloud management
system, the organization can maintain a lower amount of available capacity at a second
location (e.g., 10%) and can rapidly request more as needed in the event of a failure.
Much like insurance, the likelihood of all clients of a cloud provider failing at the same
time and requesting a high capacity spare is low, such that multiple clients can share a set
of redundant secondary resources to be used in the event of failure of primary resources.
The system may also re-home management tools and troubleshooting resources to point to
the new location following disaster recovery so that management continues uninterrupted.
[0053] From the foregoing, it will be appreciated that specific embodiments of the cloud
management system have been described herein for purposes of illustration, but that
various modifications may be made without deviating from the spirit and scope of the
inventon. Accordingly, the invention is not limited except as by the appended claims.
CLAIMS
1. A computer-implemented method to handle a request from a management tool to
access application management data from distributed application instances, the method
comprising:
receiving from an application management tool a request to access management
data related to an application running instances in one or more datacenters;
identifying one or more types of management data that satisfy the received request;
determining a distribution of the application that includes two or more instances of
the application;
gathering management data to satisfy the request from each distributed application
instance;
unifying gathered data to provide a uniform response to the received management
tool request; and
reporting the gathered and unified management data in response to the received
management tool request,
wherein the preceding steps are performed by at least one processor.
2. The method of claim 1 wherein receiving the management request comprises
receiving the tool request through a programmatic application programming interface
(API) exposed to tools for requesting management data.
3. The method of claim 1 wherein the API comprises a uniform interface for
accessing management data without requiring tools to understand where or at how many
locations application instances are running.
4. The method of claim 1 wherein receiving the management request comprises
receiving a request from a performance monitoring tool to access status information
describing operation of one or more application instances.
5. The method of claim 1 wherein identifying types of management data comprises
determining that the request asks for information that is produced by each instance of the
application.
6. The method of claim 1 wherein identifying types of management data comprises
determining which information to gather from each application instance and whether the
data is already gathered locally from data pushed to a central location by each application
instance.
7. The method of claim 1wherein determining the distribution of the application
comprises determining where the application is running and where the system will find
management data to satisfy the request.
8. The method of claim 1wherein gathering management data comprises accessing at
least one instance in a private datacenter and at least one instance in a cloud computing
facility.
9. The method of claim 1wherein gathering management data comprises contacting
each instance of the application to satisfy the received management tool request.
10. The method of claim 1 further comprising sending one or more troubleshooting
commands to one or more remote application instances, wherein the remote application
instances carry out the troubleshooting commands and report requested data back to a
central location where the management tool can access the information.
11. The method of claim 1wherein unifying gathered data comprises formatting data
so that management tools need not be written to include an understanding of the various
potential distributions of applications managed by the tools.
1 . The method of claim 1wherein unifying gathered data comprises formatting data
so that the system can freely migrate the application from location to location or to
multiple locations as needed to handle application loads, while still providing
administrators with a consistent management and troubleshooting experience.
13. The method of claim 1wherein reporting the data comprises sending the data to
the management tool through an interface on which the request was received.
14. A computer system for cross-cloud computing for cross-cloud management and
troubleshooting, the system comprising:
a processor and memory configured to execute software instructions embodied
within the following components;
a location management component that manages information about multiple
datacenter locations at which instances of an application are running;
a location data store that stores information that describes locations at which
instances of the application are running;
a tool interface component that provides an interface to the system through which
one or more tools can access management and troubleshooting information for the
application;
one or more management tools that connect to the tool interface component to
access management information or perform application troubleshooting;
a data migration component that migrates management information at one or more
remote locations where the application is running back to a home location of the
application; and
a troubleshooting component that performs one or more troubleshooting tasks on
the application at one or more locations.
15. The system of claim 14 wherein the tool interface component provides an initial
connection point for tools to access information related to the application at a cloudcomputing
appliance located within an enterprise's private datacenter.