JP2022089807A - Computer-implemented method and computer program product (estimating attributes of running workloads on platforms in system including multiple platforms as service) - Google Patents

Abstract

To provide techniques of allowing optimal placement of workloads on platforms in a system including multiple platforms as a service.SOLUTION: The invention provides a computer-implemented method and a computer program product for estimating attributes of running workloads on platforms in a system including multiple platforms as a service. A computer receives definitions of respective workloads and definitions of respective platforms that are eligible to run a set of the respective workloads. The computer maps the respective workloads and the respective platforms to attributes of running the respective workloads on the respective platforms. The computer estimates the attributes and storing the attributes in a matrix. The computer updates the attribute in the matrix, in response to a triggering event for modifying the matrix.SELECTED DRAWING: Figure 2

Description

The present invention relates to a platform as a service (PaaS) model in general, and more specifically, estimates attributes when a workload is executed on a platform in a system including a plurality of platforms as a service. Regarding that.

In a platform as a service (PaaS) model, workloads are placed to run on a platform rather than a virtual or physical machine. The platform can contain any combination of underlying resources, which is usually not exposed to the users of the platform. Platforms typically expose interfaces that allow workloads to be placed, run, monitored, and controlled on the platform.

Infrastructure as a service, such as virtual machine (VM) type selection, VM image creation and management, VM and cluster connectivity and disconnection, VM return, VM provision waiting, VM tracking, etc. Some processing elements that are normally required when working on the (IAaS) platform are no longer needed when working on the PaaS platform. At PaaS, there are two main concepts: workloads and platforms that run workloads.

The PaaS platform can be local or remote. The user creates an account on the PaaS platform. You can then upload and run workloads on these PaaS platforms via the accounts you create. Being able to run workloads on both local and remote platforms offers several benefits. Cost reduction is a typical example. Offload workloads from local to remote platforms by leveraging the hybrid cloud mechanism without spending money to build and maintain resources to handle the occasional spike in resource usage ( Since it can be offloaded, you only have to pay for additional resources when you need it, and you can reduce your total cost of ownership. Another advantage is the flexibility to use multiple cloud providers and platforms with different attributes, taking into account workload requirements and costs. Another advantage is the improvement of security. Cloud providers can provide enhanced security, isolation, and communication over private networks, thus addressing security and compliance aspects. The hybrid cloud mechanism can provide scalability by dynamically allocating workloads on the cloud platform. You can also minimize the risk of failure and downtime by leveraging the resources of your cloud provider.

The present disclosure is intended to provide a technique that enables optimal placement of workloads on platforms in a system that includes multiple platforms as a service.

According to one aspect, in a system including a plurality of platforms as a service, a computer implementation method for estimating attributes when executing a workload on the platforms is provided. The computer implementation method involves receiving the definition of each workload and the definition of each platform that is eligible to run each set of workloads. Computer implementation methods further include mapping each workload and each platform to the attributes of running each workload on each platform. Computer implementation methods further include estimating attributes and storing them in a matrix. Computer implementation methods further include updating the attributes in the matrix in response to trigger events that modify the matrix.

According to another aspect, in a system including a plurality of platforms as a service, a computer program product for estimating attributes when executing a workload on the platforms is provided. Computer program products include computer-readable storage media with program instructions. Program instructions can be executed by one or more processors. Program instructions receive the definition of each workload and the definition of each platform that is eligible to run each set of workloads, and each workload and each platform on each platform. Mapping to the attributes when running each workload, estimating the attributes and storing the attributes in the matrix, and updating the attributes in the matrix in response to the trigger event that modifies the matrix. , Is feasible.

FIG. 1 is a diagram showing a system including a plurality of platforms as a service according to an embodiment of the present invention. FIG. 2 is a flowchart showing an operation step of arranging a workload on a plurality of platforms according to an embodiment of the present invention. FIG. 3A is a flowchart showing an operation step for optimizing workload placement on a plurality of platforms according to an embodiment of the present invention. FIG. 3B is a flowchart showing an operation step for optimizing workload placement on a plurality of platforms according to an embodiment of the present invention. FIG. 4 is a diagram showing components of a computer device according to an embodiment of the present invention. FIG. 5 is a diagram showing a cloud computing environment according to an embodiment of the present invention. FIG. 6 is a diagram showing an abstraction model layer in a cloud computing environment according to an embodiment of the present invention.

The issues considered in this disclosure are defined as follows. Multiple workloads are runnable or already running. Multiple platforms are available to run these multiple workloads. Some of the platforms may be running workloads. Each workload may have a required time to complete its processing. Each workload may show different performance on each platform in terms of resource consumption and time to completion. Each platform incurs different costs, which can change over time. The cost per platform may vary depending on the resource usage, time interval, and other factors specified. Each platform may have different capacity limits. Some of the platforms may be local and some may be remote. The purpose of this technique is to discover mappings between multiple workloads and multiple platforms, which map the time to completion of the workload and the total cost of using the platforms. It is to be minimized.

Existing commercial systems for performing cloud bursting have some limitations. In a platform as a service (PaaS) model, systems typically identify resource overs and shortages through monitoring tools, monitor and determine which applications to move, monitor and regulate budget usage, It also relies on system administrators to manage cloud resource preparation and releases. The PaaS cloud is a relatively new model, with limited or no support in terms of service level and budget management in current PaaS cloud systems.

To overcome the above limitations, embodiments of the present invention focus on the PaaS cloud model. Embodiments of the invention support any type of workload and workload environment and make no assumptions about the virtualized environment. The embodiments of the present invention also consider service levels, budgets, costs, resources, workloads, and platforms. Embodiments of the invention support multiple local and remote clouds with different attributes. Embodiments of the present invention automatically optimize service levels of workloads and cloud utilization costs in the face of budget, resource, and platform constraints. Embodiments of the present invention provide a scalable solution.

FIG. 1 is a diagram showing a system 100 including a plurality of platforms as a service according to an embodiment of the present invention. The system 100 including the plurality of platforms as a service is composed of a workload 120, a platform 130, and a system 110 for mapping the workload 120 and the platform 130. The workload 120 includes M pending workloads (waiting workload 1 (121-1), waiting workload 2 (121-2), ..., Waiting workload M (121-). M)) is included. Waiting workloads are waiting to be deployed on the platform. The workload 120 further includes N running workloads (running workload 1 (122-1), running workload 2 (122-2), ..., Running workload N (122-1). -N)) is included. A running workload is a workload that is currently running on the platform. Platform 130 includes K local platforms (local platform 1 (131-1), local platform 2 (131-2), ..., Local platform K (131-K)) and L remote platforms (remote). Platform 1 (132-1), remote platform 2 (132-2), ..., Remote platform L (131-L)). Platform 130 can be provided by one or more cloud providers.

The user requests that the waiting workload run on the platform. Then, the mapping system 110 calculates the optimum placement of the waiting workload on the platform, and places the waiting workload on the platform according to the optimum placement. In response to any changes (eg, adding new workloads to system 100, changing workload requirements, etc., detailed description below), the mapping system 110 is optimal for the platform of running workloads. Recalculate the placement. According to the optimal placement of the running workload, the mapping system 110 will select one or more of the running workloads from the current platform if the current platform is not the calculated optimal platform. Move or migrate to the optimal platform calculated for your workload.

The mapping system 110 may exist on a computer device or a server. Details of the computer device or server will be described later with reference to FIG. The mapping system 110 between the workload 120 and the platform 130 is implemented in a cloud computing environment. Details of the cloud computing environment will be described later with reference to FIGS. 5 and 6.

<Mapping of workload and platform to execution attributes>

ここで、Ｃｏｓｔは、特定のプラットフォーム上で特定のワークロードを実行する際の推定コストである。Ｅｓｔｉｍａｔｅｄ Ｄｕｒａｔｉｏｎ ｔｏ Ｃｏｍｐｌｅｔｅ（ＥＤＣ）は、当該特定のプラットフォーム上で実行される当該特定のワークロードの推定完了時間であり、プラットフォーム上でワークロードが実行開始する時点から測定され、実行開始までのワークロードの待機時間は含まれない。プラットフォームに関するＲｅｓｏｕｒｃｅ Ｒｅｑｕｉｒｅｍｅｎｔｓ（ＲＲ）は、当該特定のプラットフォーム上で実行される当該特定のワークロードの推定リソース要件であり、当該特定のプラットフォームによって課金（charge）されるメインリソースに関する。各プラットフォームでは、使用料（usage charge）の計算の元となるリソースが定義されている。ＣｏｓｔはおよびＥＤＣは、マッピング用システム１１０によって学習され、推定される。ＲＲは、プラットフォームが提供するインタフェースに応じて、マッピング用システム１１０によって推定されるか、またはユーザ入力によって提供することができる。 The mapping system 110 receives definitions for a plurality of workloads and definitions for a plurality of platforms that are eligible to execute the plurality of workloads. The mapping system 110 is specified by using a mapping function that maps each pair of one workload and one platform to the attributes when the workload is executed on the platform as follows. Model the attributes of running a particular workload on your platform.

Here, Cost is the estimated cost of running a particular workload on a particular platform. Estimated Duration to Complete (EDC) is the estimated completion time of a particular workload running on that particular platform, measured from the time the workload starts running on the platform to the start of execution. Waiting time is not included. Resource Requirements (RR) for a platform are estimated resource requirements for that particular workload running on that particular platform and relate to the main resources that are charged by that particular platform. Each platform defines the resources from which the usage charge is calculated. Cost and EDC are learned and estimated by the mapping system 110. The RR can be estimated by the mapping system 110 or provided by user input, depending on the interface provided by the platform.

<Matrix WkPl containing attributes when running workloads on the platform>

The mapping between each workload and each pair of platforms and the attributes of running each workload on each platform can be stored in a matrix. For example, each workload is represented by a row in the matrix, each platform is represented by a column in the matrix, and the attributes for running a particular workload on a particular platform are stored in the cells of the matrix. This matrix is referred to as WkPl. An example of the matrix WkPl is shown in Table 1.

The cells of the matrix WkPl store attributes for executing a specific workload WK _i on a specific platform PL _j . This cell is referred to as WkPl [WK _i , PL _j ]. Of the attributes Cost, EDC, and RR, one attribute related to executing a specific workload WK _i on a specific platform PL _j is referred to as Attribute (WkPl [WK _i , PL _j ]) (for example, , Cost (WkPl [WK _i , PL _j ])).

<Update of matrix WkPl>

The mapping system 110 updates the matrix WkPl when the following trigger event occurs.

(1) A new workload is added to the system 100. In this case, a new row is added to the matrix WkPl.

(2) One or more requirements of the workload are changed. In this case, each cell in the row associated with the workload is modified.

(3) The workload completes the process on a specific platform. In this case, the actual running attributes of the workload on the platform may be updated in the cells associated with the workload and platform. The actual execution attributes of the workload on the platform are also added to the cost, EDC, and RR metrics estimation mechanisms. You may also update the attributes of other workloads that may be running on the same platform.

(4) When the processing of the workload on the platform is completed, the workload is deleted from the matrix WkPl periodically or immediately. In this case, the row related to the workload to be deleted is deleted from the matrix WkPl. If the deletion is done on a regular basis, the deletion will be applied based on the deletion criteria (eg, criteria based on the time attribute of the workload).

(5) A new platform will be added. In this case, a new column is added to the matrix WkPl and the estimated values of the attributes of each workload are stored.

(6) The cost charged by the platform will be changed. In this case, the information stored in the columns of the matrix associated with the platform is updated.

(7) The amount and / or type of resources available on the platform will change. In this case, the information stored in the column of the matrix related to the platform is updated.

(8) The platform is deleted. In this case, the column associated with the deleted platform is deleted.

To discover any of the above trigger events, the mapping system 110 holding the matrix WkPl periodically scans and monitors the workload to determine the current state of the workload and periodically the platform. Scan and monitor to determine the current state of the platform.

The mapping system 110 initiates the construction of a matrix WkPl that includes columns that define the available platforms but no rows. When the workload is added, the mapping system 110 adds rows to the matrix. The mapping system 110 then adds to each cell in this row an estimate of the workload attributes on the available platforms.

<Estimation of Cost, EDC, RR>

To estimate the Cost, EDC, and RR indicators for each workload and platform pair, the system 110 has previously recorded information, i.e. other workloads of the same type running on a particular platform. Use estimates based on the execution attributes of. If one workload is currently running and estimates of these indicators have been made for that workload, the system 110 may also consider the current state of processing the workload.

The following factors are considered when estimating the Cost, EDC, and RR indicators.

(1) workload type: Each workload can be associated with one workload type. Each workload type is a grouping of workloads that are similar in terms of running characteristics and resource requirements. The association between a given workload and one type can be specified by user input to system 110 as part of the attributes of that given workload. The Cost, EDC, and RR estimation mechanisms used by System 110 are based on the availability of type attributes for each workload, allowing a given workload to be associated with previous workloads of the same type. .. Estimates are calculated for each workload type. For each workload type, aggregated and / or detailed information about the metrics and attributes of running the workload on the platform is determined and stored. This facilitates estimation of the workload associated with the workload type. Aggregate and / or detailed information stored about the workload type associated with the workload is updated with the actual execution attributes of the completed workload.

(2) Additivity of index value: Each index of Cost, EDC, and RR collected must represent each workload independently and have additivity.

(3) PaaS cloud billing model: The PaaS cloud platform usually has one or more resources for which usage and charges are calculated. For example, memory usage or workload processing can be used to calculate usage and charges. These resources are usually defined in software (rather than hardware). Costs Fees are typically proportional to the consumption of these resources and are calculated relative to the user's hourly resource consumption. In addition, there may be an upper limit on the maximum resource usage for each user. The PaaS cloud platform may offer multiple service types. For these service types, the usage cost may be calculated separately or collectively depending on the method of calculating the cost charge. The service type may also be related to the maximum resource usage per user.

(4) Calculation of resource consumption and cost for running or completed workload: If one platform reports information on resource consumption and cost for running a specific workload, the system 110 will Use this reported resource consumption and cost information to update resource requirements and cost information for the workload type that corresponds to that workload (or aggregate or details stored for that workload type). Update or both information). Run a specific workload if one platform reports cost information based on the user's hourly resource usage, or if resource consumption and cost are not reported per workload. Resource consumption and cost for this can be calculated in several ways. 1 The first method is to collect resource consumption and cost information reported from the platform while running the workload independently for a specific user on a specific platform. 2 The second method uses, for example, operating system monitoring information to extract the resource consumption ratio for other workloads when a specific workload is executed, and applies this ratio to be executed by the user. A method of calculating the resource consumption and cost of a specific workload among the resource consumption and cost reported by the platform for multiple workloads. 3 A third way to obtain resource requirements and costs per workload is to receive this information via user input or use user input when automatic estimation is not possible. The reported resource consumption and cost information obtained by these methods is used to update the stored intensive and / or detailed information about the workload type.

(5) Workload migration cost: This cost can be added to the running workload. This is an additional factor for estimated cost, estimated completion time, and estimated resource requirements, and is added on top of estimates that do not take migration into account. This additional cost factor can be either positive or negative, depending on the current state of the running workload and the migration cost of the workload environment. If migration costs are added, the estimates for each platform may differ compared to the estimates for each workload without migration. The migration cost is the cost of migrating one running workload from the current platform to the optimal platform determined by the mapping system 110.

<Overall attributes>

Overall attributes include OverallBudget. An Overall Budget is a monetary overall budget allocated to each user / organizational unit and is used for workloads performed by the user / organizational unit.

<Workload attribute>

The workload (WK _i ) has the following attributes. (1) Original Budget (WK _i ): This is the monetary budget allocated to the workload WK _i . (2) Reminding Budget (WK _i ): This is the current remaining monetary budget (after spending) allocated to the workload WK _i . (3) RDC (WK _i ): This is the request completion time until the workload WK _i is completed on the platform, and is specified from the workload submission time. RDC (WK _i ) is considered a best effort (soft) constraint. (4) SubmissionTime (WK _i ): This is the input time of the workload WK _i . (5) Priority (WK _i ): This is the priority assigned to the workload WK _i . Priority is relative to the priority of other workloads. (6) Eligible Platforms (WK _i ): This is a list of platforms designated as platform indexes in the matrix WkPl, and the platforms in the list are eligible to run workload WK _i . The list of eligible platforms for a particular workload can be, for example, a subset of platforms selected based on platform type and user access to the platform.

<Platform attribute>

The platform attribute of one platform (PL _j ) includes MaxResource (PL _j ). The attribute MaxResource (PL _j ) is the maximum upper limit for the consumption of the main resource charged by the platform PL _j for one user. The main resource can be, for example, the memory size or the number of workload processes.

<Distribution of workload considering cost and service level for multiple platforms as a service>

FIG. 2 is a flowchart showing an operation step of arranging a workload on a plurality of platforms according to an embodiment of the present invention. In step 201, a system that maps the workload to the platform (eg, system 110 shown in FIG. 1) detects a trigger event that modifies the matrix. The matrix pairs the workload and the platform, and contains the attributes for running the workload on the platform. For example, the matrix is the matrix WkPl described above. To detect the trigger event that modifies the matrix, the mapping system periodically scans and monitors the workload and platform to determine the current state of the workload and platform. The trigger event is at least one of the following events. (1) New workloads are added to systems that include multiple platforms as a service, (2) one or more requirements (budget, requirement completion period, priority, eligible platforms) for each workload change (3) The workload completes processing on one of multiple platforms, (4) a new platform is added to a system containing multiple platforms as a service, (5) one of multiple platforms One cost changes, (6) the availability, quantity, or type of resources available to one of the multiple platforms changes, (7) the maximum resource utilization per user for one of the multiple platforms. Is changed, (8) one of the multiple platforms is removed from the system containing the multiple platforms as a service, (9) the overall budget is changed.

In step 202, the mapping system recalculates the matrix when a trigger event is detected. For example, the matrix WkPl is recalculated. Then, in the matrix, the attributes when executing the workload on the platform are updated.

At step 203, the mapping system calculates the optimal placement of the workload on the platform. The mapping system uses an algorithm that solves the optimization problem. The optimization problem and the algorithm for solving it will be described in detail later with reference to FIG.

At step 204, the mapping system determines if each workload has not been executed on each platform. For example, if each workload is unexecuted, the workload is a waiting workload (eg, as shown in FIG. 1, waiting workload 1 (121-1), waiting workload 2 (eg, as shown in FIG. 1). 121-2), ..., Waiting workload M (121-M)), waiting to be mapped (by the mapping system) to a platform. When each workload is running, it is a running workload (eg, running workload 1 (122-1), running workload 2 (122-, as shown in FIG. 1)). 2), ..., Running workload N (122-N)).

If it is determined that each workload has not been executed (YES in step 204), in step 205, the mapping system places each workload on the calculated optimal platform for that workload. do. The optimal platform is calculated in step 203.

If it is determined that each workload is running (NO in step 204), then in step 206 the mapping system is currently on the optimal platform for each workload (calculated in step 203). Determine if it is running. In other words, the mapping system determines if the current platform on which each workload is running is the calculated optimal platform for that workload.

If it is determined that each workload is not currently running on the calculated optimal platform (NO in step 206), that is, the current platform is the calculated optimal platform for each workload. If not, at step 207 the mapping system moves or migrates the workload from the current platform to the calculated optimal platform.

If it is determined that each workload is currently running on the calculated optimal platform (YES in step 206), that is, the current platform is the calculated optimal platform for each workload. If so, the mapping system keeps the workload on the current platform (optimal platform for the workload) without moving or migrating from the current platform to another platform.

If it is determined that each workload is currently running on the calculated optimal platform (YES in step 206), after step 205 or after step 207, the mapping system will go to step 208. Run. At step 208, the mapping system determines if all workloads are arranged according to the optimal placement (calculated in step 203).

If it is determined that all workloads are deployed according to optimal placement (YES in step 208), the mapping system considers the cost and service level for multiple platforms as a service, or follows optimal placement. , Complete the workload placement. If it determines that all of the workloads are not arranged according to the optimal placement (NO in step 208), the mapping system repeats steps 204-208 until all of the workloads are placed according to the optimal placement. ..

<Formulation of optimization problem>

ここで、プラットフォームインデックスＰ_ｉは、ワークロードＷＫ_ｉに最適なプラットフォームを特定し、複数のプラットフォームインデックス｛Ｐ_ｉ｝は、ワークロードのそれぞれに最適なプラットフォームを特定する。ＷＫＮはワークロードの数、ＰＬＮはプラットフォームの数を表す。目的関数には以下の制約（constraints）がある。

Here, the platform index _Pi identifies the optimal platform for the workload WK _i , and the plurality of platform indexes {P _i } identify the optimal platform for each of the workloads. WKN represents the number of workloads and PLN represents the number of platforms. The objective function has the following constraints.

これは、ベストエフォート（またはソフト）制約である。この制約は、各ワークロードについて、選択されたプラットフォーム上での推定完了時間（ＥＤＣ）が、要求完了時間（ＲＤＣ）を超えないことを必要とするものである。 The first constraint is as follows.

This is a best effort (or soft) constraint. This constraint requires that, for each workload, the estimated completion time (EDC) on the selected platform does not exceed the required completion time (RDC).

これは、必須（またはハード）制約である。第２の制約は、各ワークロードについて、選択されたプラットフォーム上で実行するためのコスト（Ｃｏｓｔ）が、残り予算（ＲｅｍａｉｎｉｎｇＢｕｄｇｅｔ）を超えないことを必要とするものである。 The second constraint is as follows.

This is a mandatory (or hard) constraint. The second constraint requires that, for each workload, the cost to run on the selected platform does not exceed the remaining budget (Remaining Budget).

これは、必須（またはハード）制約である。第３の制約は、すべてのワークロードを各々の選択されたプラットフォーム上で実行するための総コストが、全体予算（ＯｖｅｒａｌｌＢｕｄｇｅｔ）を超えないことを必要とするものである。 The third constraint is as follows.

This is a mandatory (or hard) constraint. The third constraint requires that the total cost of running all workloads on each selected platform does not exceed the overall budget.

これは、必須（またはハード）制約である。第４の制約は、一のワークロードについて選択されたプラットフォームが、当該ワークロードを実行する資格のある複数のプラットフォームの１つであることを必要とするものである。 The fourth constraint is as follows.

This is a mandatory (or hard) constraint. The fourth constraint requires that the platform selected for a workload be one of a plurality of platforms eligible to run that workload.

これは、必須（またはハード）制約である。第５の制約は、一のプラットフォーム上で実行されるワークロードによって当該プラットフォームから消費されるリソースが、ユーザに対する当該プラットフォームのリソース消費上限を超えないこと、または、いずれかのプラットフォームのリソース容量またはリソース消費上限を超えないことを必要とするものである。 The fifth constraint is as follows.

This is a mandatory (or hard) constraint. The fifth constraint is that the resources consumed by a workload running on one platform from that platform do not exceed the resource consumption limit of that platform for the user, or the resource capacity or resources of either platform. It is necessary not to exceed the upper limit of consumption.

<Algorithm of optimization problem>

The optimization problem is P ₁ , P ₂ , ... .. .. , Includes objective cost function and constraint functions, with _PWKN (Platform Index) as variables. The objective function and the constraint function are variables P ₁ , P ₂ , ... .. .. , Mathematicatical methods that depend on _PWKN cannot be defined by formulation (although this depends on the values stored in the matrix WkPl based on the estimation method described above). programming solvers) are not applicable to this optimization problem. For the optimization problem defined here, the paradigm of the greedy algorithm that progresses step by step to the global optimum solution can be applied by using the locally optimal selection. Is.

A system for mapping a workload to a platform (eg, system 110 shown in FIG. 1) first places each workload on the platform with the lowest cost associated with that workload. The system then increments the cost stepwise for workloads for which the estimated completion time (EDC) exceeds the required completion time (RDC). At each stage, the mapping system selects the workload with the lowest upgrade cost. Here, the workload upgrade must allow the estimated completion time (EDC) of the workload to be less than or equal to the requested completion time (RDC). The mapping system then upgrades the selected workload. At each stage, mandatory and best effort constraints are checked.

When solving the optimization problem, the latest matrix WkPl is input, and the result of solving the optimization problem is as follows. That is, for each workload WK _i , the related platform index Pi in the matrix _WkPl is calculated. The related platform index _{Pi identifies the best platform for running the workload WK i .}

3A and 3B are flowcharts showing operation steps for optimizing workload placement on a plurality of platforms according to an embodiment of the present invention.

Referring to FIG. 3A, in step 301, the system for mapping the workload and the platform (eg, system 110 shown in FIG. 1) executes each workload for each workload. Set up a platform index that identifies each of the platforms with the lowest cost. For each workload WK _i , the mapping system sets the value of the associated platform index Pi in the matrix _WkPl . The platform identified by the value of such platform index _{Pi has the lowest cost of running the workload WK i .} The platform index _Pi satisfies the following.

Referring to FIG. 3A, in step 302, the mapping system determines whether the required constraints on budget, platform eligibility, and platform resource capacity are met. The mapping system checks whether the required (or hard) constraints are met. The essential (or hard) constraint is the second to fifth constraint described above.

If it is determined that the required constraint is not satisfied (NO in step 302), the mapping system executes step 310 shown in FIG. 3 (B). Referring to FIG. 3B, at step 310, the mapping system determines whether the previous placements of each workload meet the required constraints. The previous placement is the solution of the optimization problem in the previous iteration.

Continuing to refer to FIG. 3 (B), if it is determined that the previous placement of each workload meets the required constraints (YES in step 310), then in step 311 the mapping system will use the mapping system for each workload. Output the previous placement as the optimal placement. The mapping system uses the previous solution as a result of the algorithm. The system then completes the steps of the algorithm.

Continuing to refer to FIG. 3B, if it is determined that the previous placement of each workload does not meet the required constraints (NO in step 310), then in step 312 the mapping system will have its respective workload. Does not output the optimal placement for. There is no feasible solution to the optimization problem, for the following reasons: The algorithm starts with the lowest cost mapping. If the budget constraint is violated in this mapping, then there is no feasible solution under the given constraint. Such a scenario can be classified as invalid user input to the algorithm. In this way, the system completes the steps of the algorithm.

Returning to FIG. 3A, if it is determined that the required constraint is satisfied (YES in step 302), in step 303, the mapping system has the best effort regarding the request completion time of each workload. Check the constraints and determine the set of workloads that do not meet the best effort constraints. The best effort constraint is the first constraint described above. Each workload in the workload group is identified using an index in the matrix WkPl. The workload group is referred to as WKS.

Referring to FIG. 3A, in step 304, the mapping system determines whether the workload group (WKS) is empty. An empty workload group (WKS) indicates that the best effort constraint on request completion time (RDC) is met. If it is determined that the workload group (WKS) is empty (YES in step 304), the mapping system executes step 313 shown in FIG. 3 (B). Referring to FIG. 3B, at step 313, the mapping system outputs the current placement as the optimal placement for each workload. The current placement is the solution to the optimization problem in the current iteration. The system uses the current solution as a result of the algorithm. The system then completes the steps of the algorithm.

Returning to FIG. 3A and determining that the workload group (WKS) is not empty (NO in step 304), in step 305, the mapping system incurs costs for the workload group (WKS). Determines the set of candidate platforms that are the lowest and can meet the best effort constraints. For each workload in the workload group (WK _i in WKS), the mapping system determines the lowest cost platform. This platform meets best effort constraints. The candidate platforms {PC _i } are as follows.

ワークロード群ＷＫＳに含まれるワークロードＷＫ_ｉに関するアップグレード後コストは、以下のように示すことができる。

Referring to FIG. 3 (A), in step 306, the mapping system determines, for the workload group (WKS), the upgraded platform indexes that identify candidate platforms, and the upgraded cost (upgraded). costs) is calculated. For each workload WK _i in the workload group, the post-upgrade platform index PU _i is calculated as follows.

The post-upgrade cost for the workload WK _i included in the workload group WKS can be shown as follows.

If there is no platform in the candidate platforms that can meet the best effort constraint for one workload included in the workload group WKS and none of the platforms in the candidate platforms is available, the system is concerned. It is determined that the workload cannot be upgraded, and the post-upgrade cost of the workload is set to invalid or infinite (for example, UpgradeCost (WK _i ) = ∞). In this way, the system avoids adding the workload to the workload group WKS at the next algorithm iteration. The workload upgrade will not be considered in the next iteration to resolve the optimization problem.

Referring to FIG. 3A, at step 307, the mapping system determines if the workload group (WKS) contains at least one workload for which the post-upgrade cost is valid. If the workload group determines that the post-upgrade cost does not contain any valid workload (NO in step 307), the system performs step 313 shown in FIG. 3 (B). Referring to FIG. 3B, at step 313, the system sets the current placement as the optimal placement for each workload. The current placement is the solution to the optimization problem in the current iteration. The system uses the current solution as a result of the algorithm. In this case, there is no better way to upgrade the solution of the optimization problem. The system then completes the steps of the algorithm.

ＷＫ_ｔは、最適化問題を解く今回の反復で発見された最も費用対効果の高いワークロードである。 Returning to FIG. 3A, if the workload group determines that it contains at least one post-upgrade cost-effective workload (YES in step 307), in step 308, the mapping system Select the workload with the lowest post-upgrade cost from one or more workloads with the lowest post-upgrade cost. From the workload group (WKS), the workload with the lowest post-upgrade cost is selected. The workload with the lowest post-upgrade cost is referred to as WK _t . WK _t satisfies the following equation.

WK _t is the most cost-effective workload found in this iteration of solving an optimization problem.

Referring to FIG. 3A, in step 309, the mapping system upgrades the workload by setting the post-upgrade platform index for the workload with the lowest post-upgrade cost. The system upgrades WK _t by setting P _t = PU _t . Here, PU _t is the post-upgrade platform index and P _t is the latest platform index selected for WK _t .

After step 309, the mapping system iterates step 302 as the next iteration of the algorithm to discover another most cost-effective workload. The mapping system performs the next iteration of solving the optimization problem until it finds the optimal placement of each workload on each platform.

An upgrade of one workload can only be done once in the algorithm of the optimization problem, and the post-upgrade workload will not be reviewed in the next iteration of the algorithm. Therefore, the number of algorithm iterations is limited to the number of workloads. At each iteration, a linear scan of the workload can be performed. The optimization problem algorithm then selects the workload to upgrade and proceeds to the next iteration. All calculations per workload need to be done once, and only constraints that take into account all workloads need to be recalculated per iteration.

The algorithm of the optimization problem can be done in parallel by performing all the independent calculations for each workload at the same time for all the associated workloads. As a way to further optimize the algorithm, all basic data and counters for various constraints can be maintained for both workloads and platforms, and these data and counters can be updated incrementally as changes are made. .. This allows the constraint calculation to be performed faster.

FIG. 4 is a diagram showing a component of a computer device or a server 400 according to an embodiment of the present invention. Note that FIG. 4 is merely an example of one embodiment and does not imply any restrictions on the environment in which different embodiments can be implemented.

Referring to FIG. 4, the computer device or server 400 includes one or more processors 420, memory 410, and one or more tangible storage devices 430. In FIG. 4, the communication between the above-mentioned components of the computer device or the server 400 is indicated by reference numeral 490. The memory 410 includes one or more ROMs 411, one or more RAMs 413, and one or more caches 415. One or more operating systems 431 and one or more computer programs 433 reside on a computer-readable tangible storage device 430.

The computer device or server 400 further includes one or more I / O interfaces 450. The I / O interface 450 allows data to be input and output to and from a computer device or one or more external devices 460 that can be connected to the server 400. The computer device or server 400 further includes one or more network interfaces 440 for communication between the computer device or server 400 and the computer network.

The present invention can be an integrated system, method or computer program product or a combination thereof at any possible level of technical detail. The computer program product may include a computer-readable storage medium that stores computer-readable program instructions for causing the processor to perform aspects of the invention.

The computer-readable storage medium can be a tangible device capable of holding and storing the instructions used by the instruction executing device. The computer-readable storage medium may be, for example, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or an appropriate combination thereof. More specific examples of computer readable storage media include portable computer diskettes, hard disks, RAMs, ROMs, EPROMs (or flash memories), SRAMs, CD-ROMs, DVDs, memory sticks, floppy disks, punch cards or grooves. Examples include mechanically coded devices that record instructions in raised structures and the like, and appropriate combinations thereof. Computer-readable storage devices as used herein are radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (eg, optical pulses through optical fiber cables). Or it should not be interpreted as a transient signal itself, such as an electrical signal transmitted over a wire.

The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computer device / processing device. Alternatively, it can be downloaded to an external computer or external storage device via a network (eg, the Internet, LAN, WAN or wireless network or a combination thereof). The network can include copper transmission cables, optical transmission fibers, wireless transmissions, routers, firewalls, switches, gateway computers or edge servers or combinations thereof. A network adapter card or network interface in each computer device / processing device receives computer-readable program instructions from the network and stores the computer-readable program instructions in a computer-readable storage medium in each computer device / processing device. Forward.

The computer-readable program instructions for performing the operations of the present invention are assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcodes, firmware instructions, state setting data, integrated circuit configuration data, or , Source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as SmallTalk and C ++, and procedural programming languages such as C programming languages and similar programming languages. It can be either. Computer-readable program instructions can be executed entirely on the user's computer or partially on the user's computer as a stand-alone software package. Alternatively, it can be run partially on the user's computer and partially on the remote computer, or entirely on the remote computer or server. In the latter case, the remote computer may connect to the user's computer via any type of network, including LAN and WAN, or to an external computer (eg, over the Internet using an Internet service provider). You may connect. In some embodiments, electronic circuits including, for example, programmable logic circuits, field programmable gate arrays (FPGAs), programmable logic arrays (PLAs) are used to customize the electronic circuits for the purposes of carrying out aspects of the invention. By using the state information of the computer-readable program instruction, the computer-readable program instruction can be executed.

Each aspect of the invention is described herein with reference to a flow chart and / or block diagram of a method, apparatus (system), and computer program product according to an embodiment of the invention. Each block in the flow chart and / or block diagram, and a combination of multiple blocks in the flow chart and / or block diagram, can be executed by computer-readable program instructions.

The computer-readable program instructions described above may be provided to the processor of a computer or other programmable data processing device to produce a machine. Thereby, these instructions executed through the processor of such a computer or other programmable data processing device perform a function / operation specified by one or more blocks in a flowchart, a block diagram, or both. Form the means of. The computer-readable program instructions described above may be further stored in a computer-readable storage medium that can be instructed to function in a particular manner for a computer, programmable data processing device or other device or a combination thereof. Thereby, the computer-readable storage medium in which the instructions are stored constitutes a product containing instructions that execute a function / operation mode identified by one or more blocks in a flowchart, a block diagram, or both.

Also, a computer execution process by loading computer-readable program instructions into a computer, other programmable device, or other device and performing a series of operating steps on that computer, other programmable device, or other device. May be generated. Thereby, an instruction executed on the computer, another programmable device, or another device performs a function / operation specified by one or more blocks in a flowchart, a block diagram, or both.

Flow charts and block diagrams in the drawings show the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the invention. In this regard, each block in a flowchart or block diagram can represent a module, segment, or portion of an instruction that contains one or more executable instructions to perform a particular logical function. In some other implementations, the functions shown in the blocks may be performed in a different order than shown in each figure. For example, two blocks shown in succession may actually be achieved as one step, simultaneously or substantially simultaneously, or partially or wholly, depending on the functions involved. It may be executed in a time-overlapping manner, or may be executed in the reverse order as the case may be. It should be noted that each block in the block diagram and / or flowchart, and the combination of multiple blocks in the block diagram or flowchart or both, may be by a dedicated hardware-based system performing a specific function or operation, or with dedicated hardware. It can be executed in combination with computer instructions.

Although this disclosure includes a detailed description of cloud computing, the implementation of the teachings described herein is not limited to cloud computing environments. Rather, embodiments of the invention can be practiced with any other type of computer environment currently known or developed in the future.

Cloud computing enables easy and on-demand network access to shared pools of configurable computing resources (eg networks, network bandwidth, servers, processing, memory, storage devices, applications, virtual machines and services). It is a model of service provision to ensure that it can be quickly provisioned and released with minimal administrative effort or interaction with a minimal service provider. This cloud model may include at least 5 characteristics, at least 3 service models, and at least 4 implementation models.

The characteristics are as follows.

On-demand self-service: Cloud consumers unilaterally prepare computing power such as server time and network storage as needed, without the need for human interaction with service providers. Can be done.

Broad network access: Computing power is available over the network and can be accessed through standard mechanisms. This facilitates utilization by heterogeneous thin or thick client platforms (eg mobile phones, laptops, PDAs).

Resource pooling: Provider's computing resources are pooled and offered to multiple consumers using a multi-tenant model. Various physical and virtual resources are dynamically allocated and reassigned on demand. Consumers generally do not manage or know the exact location of the resources provided, so they have a sense of location independence. However, consumers may be able to locate at higher levels of abstraction (eg, country, state, data center).

Rapid Elasticity: Computing power can be prepared quickly and flexibly, so that in some cases it can be automatically scaled out immediately, or released quickly and scaled in immediately. .. To consumers, the computing power available for preparation often appears to be unlimited and can be purchased in any quantity at any time.

Services Measured: Cloud systems automatically utilize resource usage by leveraging some level of abstraction-level measurement capabilities appropriate for the type of service (eg storage, processing, bandwidth, active user accounts). Control and optimize. Resource usage can be monitored, controlled, and reported to provide transparency to both providers and consumers of the services utilized.

The service model is as follows.

Software as a Service (Software as a Service): A feature provided to consumers is the availability of provider applications running on the cloud infrastructure. The application can be accessed from various client devices via a thin client interface such as a web browser (eg, webmail). Consumers do not manage or control the underlying cloud infrastructure, including networks, servers, operating systems, storage, or even individual application features. However, this does not apply to user-specific limited application configuration settings.

Platform as a Service (PaaS): The functionality provided to the consumer is to deploy the application created or acquired by the consumer to the cloud infrastructure using the programming languages and tools supported by the provider. .. Consumers do not manage or control the underlying cloud infrastructure, including networks, servers, operating systems, and storage, but they can control the deployed applications and, in some cases, the configuration of their hosting environment.

Infrastructure as a Service (IaaS): The capabilities provided to consumers are processors, storage, networks, and other basic computing resources that allow consumers to deploy and run any software, including operating systems and applications. Is to prepare. Consumers do not manage or control the underlying cloud infrastructure, but can control the operating system, storage, and deployed applications, and in some cases partially some network components (eg, host firewalls). Can be controlled.

The deployment model is as follows.

Private cloud: This cloud infrastructure is dedicated to a specific organization. This cloud infrastructure can be managed by the organization or a third party and can exist on-premises or off-premises.

Community Cloud: This cloud infrastructure is shared by multiple organizations and supports specific communities with common interests (eg, missions, security requirements, policies, and compliance). This cloud infrastructure can be managed by the organization or a third party and can exist on-premises or off-premises.

Public Cloud: This cloud infrastructure is provided to the general public and large industry groups and is owned by organizations that sell cloud services.

Hybrid Cloud: This cloud infrastructure is a combination of two or more cloud models (private, community or public). Each model retains its own entity, but is bound by standard or individual techniques to provide data and application portability (eg, cloud bursting for load balancing between clouds).

A cloud computing environment is a service-oriented environment with an emphasis on statelessness, low coupling, modularity and semantic interoperability. At the core of cloud computing is an infrastructure that includes a network of interconnected nodes.

FIG. 5 shows an exemplary cloud computing environment 50. As shown in the figure, the cloud computing environment 50 includes one or more cloud computing nodes 10. With respect to these, a local computer device used by a cloud consumer (such as a mobile device 54A, a desktop computer 54B, a laptop computer 54C, or an automobile computer system 54N or a combination thereof) can communicate with each other. Nodes 10 can communicate with each other. Nodes 10 can be physically or virtually grouped (not shown) in one or more networks, such as, for example, the private, community, public or hybrid clouds described above, or a combination thereof. This allows the cloud computing environment 50 to provide infrastructure, platforms or software as services or a combination thereof, for which cloud consumers do not need to maintain resources on local computer equipment. The types of computer devices 54A to N are merely examples, and the computing node 10 and the cloud computing environment 50 may use any type of network, a network addressable connection (for example, using a web browser), or both. It should be understood that it is possible to communicate with any kind of electronic device through.

Next, FIG. 6 shows a set of functional abstraction layers provided by the cloud computing environment 50 (FIG. 5). It should be understood in advance that the components, layers and functions shown in FIG. 6 are merely examples, and the embodiments of the present invention are not limited thereto. As shown, the following layers and corresponding features are provided.

The hardware and software layer 60 includes hardware and software components. Examples of hardware components include a mainframe 61, a reduced instruction set computer (RISC) architecture-based server 62, a server 63, a blade server 64, a storage device 65, and a network and network component 66. In some embodiments, the software components include network application server software 67 and database software 68.

The virtualization layer 70 provides an abstraction layer. From that layer, virtual entities such as virtual servers 71, virtual storage 72, virtual networks 73 including virtual private networks, virtual applications and operating systems 74, and virtual clients 75 can be provided.

As an example, the management layer 80 can provide the following functions. Resource preparation 81 enables the dynamic procurement of computing resources and other resources used to perform tasks within a cloud computing environment. Weighing and pricing 82 enables cost tracking as resources are used within a cloud computing environment and billing or invoice delivery for consumption of these resources. As an example, these resources can include licenses for application software. Security allows for identification of cloud consumers and tasks, as well as protection for data and other resources. The user portal 83 provides consumers and system administrators with access to the cloud computing environment. Service level management 84 allows the allocation and management of cloud computing resources to meet the requested service level. Service Level Agreements (SLA) Planning and Implementation 85 will enable the advance arrangement and procurement of cloud computing resources that are expected to be needed in the future in accordance with SLA.

The workload layer 90 provides an example of a function that can be used in a cloud computing environment. Examples of workloads and features available from this layer include mapping and navigation 91, software development and lifecycle management 92, virtual classroom education delivery 93, data analysis processing 94, transaction processing 95, and function 96. Is done. Function 96 is a function for arranging a workload on a plurality of platforms as a service.

Claims (20)

A computer implementation method for estimating attributes when running workloads on platforms in a system that includes multiple platforms as a service.
Receiving the definition of each workload and the definition of each platform that is eligible to run that set of workloads.
Mapping each of the above workloads and each of the above platforms to the attributes of running the respective workloads on the respective platforms, and
Estimating the attribute and storing the attribute in a matrix,
Updating the attributes in the matrix in response to a trigger event that modifies the matrix.
Computer implementation methods, including. One of the attributes further comprises estimating the cost of running each of the workloads on each of the platforms.
The computer mounting method according to claim 1. One of the attributes further comprises determining the estimated time to complete each of the workloads on each of the platforms.
The computer mounting method according to claim 1. One of the attributes further comprises determining the resource requirements of each of the workloads running on each of the platforms.
The computer mounting method according to claim 1. For workloads currently running, it further includes adding a migration cost to the attributes.
The migration cost is the cost of migrating the currently running workload from the current platform to the optimal platform for the currently running workload.
The computer mounting method according to claim 1. To classify each of the above workloads into one or more workload types based on the similarity of execution attributes and resource requirements.
Determining and storing aggregated information about the attributes for each of the one or more workload types.
Using the aggregated information, further comprising estimating said attributes when running a new workload associated with one of said one or more workload types.
The computer mounting method according to claim 1. When resource consumption and cost information for executing a workload is received from one platform, the resource consumption and cost information is used to describe the aggregation for one workload type with which the workload is associated. Including further updating information,
The computer mounting method according to claim 6. If you do not receive resource consumption and cost information to run one workload from one platform,
(1) While executing the workload on the platform, collect the resource consumption and cost information for executing the workload.
(2) The resource consumption and cost information of the workload is extracted based on the resource consumption and cost information of the other platform in the respective platforms by extracting the resource consumption ratio of the workload to the other platforms in the respective platforms. To decide,
(3) Determining the resource consumption and cost information by one of the methods of receiving user input, and
It further comprises updating the aggregated information about one workload type with which the workload is associated with the resource consumption and cost information.
The computer mounting method according to claim 6. Scanning each of the above workloads and
Detecting if new workloads have been added and
Detecting whether one or more requirements for the currently running workload have changed, and
Detecting whether one of each of the above workloads has completed processing,
Scanning each of the above platforms and
Detecting if a new platform has been added and
Detecting whether the costs incurred on one of the above platforms have changed, and
To detect whether one or more resources in said one of each of the said platforms have changed.
Detecting whether or not the one of the respective platforms has been deleted,
The new platform is added, the one or more requirements for the currently running workload are modified, the one of the respective workloads completes the process, the new platform is added. At least one of the costs incurred on the one of the respective platforms is changed, the one or more resources are modified, and the one of the respective platforms is deleted. In the case of further including determining that the trigger event that modifies the matrix has been detected.
The computer mounting method according to claim 1. Based on the information in the matrix, the optimal placement of each of the workloads on each of the platforms is determined.
Based on the optimal placement, each of the above workloads will be placed on each of the above platforms.
The computer mounting method according to claim 1. In a system including multiple platforms as a service, a computer program product for estimating attributes when executing a workload on the platform, the computer program product is a computer-readable storage medium including program instructions. Including, the program instruction can be executed by one or more processors, and the program instruction is.
Receiving the definition of each workload and the definition of each platform that is eligible to run that set of workloads.
Mapping each of the above workloads and each of the above platforms to the attributes of running the respective workloads on the respective platforms, and
Estimating the attribute and storing the attribute in a matrix,
Updating the attributes in the matrix in response to a trigger event that modifies the matrix.
A computer program product that can be executed. As one of the attributes, it further comprises a program instruction capable of estimating the cost of running each of the workloads on each of the platforms.
The computer program product according to claim 11. As one of the attributes, further comprising a program instruction capable of determining the estimated time to complete each of the workloads on each of the platforms.
The computer program product according to claim 11. As one of the attributes, it further comprises a program instruction capable of determining the resource requirements of each of the workloads running on each of the platforms.
The computer program product according to claim 11. For the workload currently running, it further contains program instructions that can be executed to add a migration cost to the attributes.
The migration cost is the cost of migrating the currently running workload from the current platform to the optimal platform for the currently running workload.
The computer program product according to claim 11. To classify each of the above workloads into one or more workload types based on the similarity of execution attributes and resource requirements.
Determining and storing aggregated information about the attributes for each of the one or more workload types.
It further includes program instructions that can use the aggregated information to estimate the attributes when executing a new workload associated with one of the one or more workload types, and to execute. ,
The computer program product according to claim 11. When resource consumption and cost information for running a workload is received from a platform, the resource consumption and cost information is used to the aggregate for one workload type with which the workload is associated. Includes additional program instructions that can be executed to update specific information,
The computer program product according to claim 16. If you do not receive resource consumption and cost information to run one workload from one platform,
(1) While executing the workload on the platform, collect the resource consumption and cost information for executing the workload.
(2) The resource consumption and cost information of the workload is extracted based on the resource consumption and cost information of the other platform in the respective platforms by extracting the resource consumption ratio of the workload to the other platforms in the respective platforms. To decide,
(3) Determining the resource consumption and cost information by one of the methods of receiving user input, and
The resource consumption and cost information is used to update the aggregated information about one workload type with which the workload is associated, and further includes program instructions that can be executed.
The computer program product according to claim 16. Scanning each of the above workloads and
Detecting if new workloads have been added and
Detecting whether one or more requirements for the currently running workload have changed, and
Detecting whether one of each of the above workloads has completed processing,
Scanning each of the above platforms and
Detecting if a new platform has been added and
Detecting whether the costs incurred on one of the above platforms have changed, and
To detect whether one or more resources in said one of each of the said platforms have changed.
Detecting whether or not the one of the respective platforms has been deleted,
The new platform is added, the one or more requirements for the currently running workload are changed, the one of the respective workloads completes the process, the new platform is added. At least one of the costs incurred on the one of the respective platforms is changed, the one or more resources are modified, and the one of the respective platforms is deleted. In some cases, it further comprises a program instruction capable of determining that the trigger event that modifies the matrix has been detected.
The computer program product according to claim 11. Based on the information in the matrix, the optimal placement of each of the workloads on each of the platforms is determined.
Based on the optimal placement, each of the above workloads will be placed on each of the above platforms.
The computer program product according to claim 11.

Images