Abstract: Embodiments for performing rolling software upgrades in a disaggregated computing environment. A rolling upgrade manager is provided for upgrading one or more disaggregated servers. A designated memory area is used for storing an updated software component, and a disaggregated server is switched to the designated memory area from a currently assigned memory area when performing the software upgrade. A process state and program data is maintained in the currently assigned memory area while maintaining the updated software component in the designated memory area such that the process state and program data are read from the currently assigned memory area and the updated software component is read from the designated memory area during currently executing operations of the disaggregated server.

Abstract: Embodiments for optimizing memory placement in a disaggregated computing environment. A new workload is assigned to a subset of a plurality of processors, the subset of processors assigned a subset of a plurality of memory devices. In some embodiments, a determination is made as to whether the new workload is categorized as a memory-dependent workload having a memory need which can be met primarily by the subset of the memory devices. If the new workload is categorized as a memory-dependent workload, a determination is then made as to whether the subset of the memory devices is meeting the memory need of the new workload. When the subset of the memory devices is not meeting the memory need of the new workload, a memory related action is taken.

Abstract: An embodiment of the invention includes applying a first partition to a plurality of LPs, wherein a particular LP is assigned to a first set of LPs. A second partition is applied to the LPs, wherein the particular LP is assigned to an LP set different from the first set. For both the first and second partitions, lookahead values and transit times are determined for each of the LPs and related links. For the first partition, a first system progression rate is computed using a specified function with the lookahead values and transit times determined for the first partition. For the second partition, a second system progression rate is computed using the specified function with the lookahead values and transit times determined for the second partition. The first and second system progression rates are compared to determine which is the lowest.

Abstract: Embodiments are provided herein for component composition of a disaggregated computing system. A plurality of general purpose links connecting a computing element to other hardware elements are provided within the disaggregated computing system. Each of the plurality of general purpose links comprise a point-to-point connection to at least one of the other hardware elements such that the plurality of general purpose links conform to a configuration used by the other hardware elements regardless of a type of data being transferred through the plurality of general purpose links.

Abstract: Direct inter-processor communication is enabled with respect to data in a memory location without having to switch specific circuits through a switching element (e.g., an optical switch). Rather, in this approach a memory pool is augmented to include a dedicated portion that serves as a disaggregated memory common space for communicating processors. The approach obviates the requirement of switching of physical memory modules through the optical switch to enable the processor-to-processor communication. Rather, processors (communicating with another) have an overlapping ability to access the same memory module in the pool; thus, there is no longer a need to change physical optical switch circuits to facilitate the inter-processor communication. The disaggregated memory common space is shared among the processors, which can access the common space for reads and writes, although particular locations in the memory common space for reads and writes are different.

Abstract: MapReduce processing is carried out in a disaggregated compute environment comprising a set of resource pools that comprise a processor pool, and a memory pool. Upon receipt of a MapReduce job, a task scheduler allocates resources from the set of resource pools, the resources including one or more processors drawn from the processor pool, and one or more memory modules drawn from the memory pool. The task scheduler then schedules a set of tasks required by the MapReduce job. At least one particular task in the set is scheduled irrespective of a location of data required for the particular task. In association with a shuffle phase of the MapReduce job, and in connection with the particular task, at least one connection between a processor and at least one memory module is dynamically rewired based on the location of the data required for the particular task, thereby obviating network transfer of that data.

Abstract: Photonic circuits are disclosed having an efficient optical power distribution network. Laser chips (InP) having different wavelengths are flip-chip assembled near the center of a silicon photonic chip. Each InP die has multiple optical lanes, but a given die has only one wavelength. Waveguides formed in the photonic chip are optically connected to the lanes, and fan out to form multiple waveguide sets, where each waveguide set has one of the waveguides from each of the different wavelengths, i.e., one waveguide from each InP die. The waveguide network is optimized to minimize the number of crossings that any given waveguide may have, and no waveguide having a particular wavelength crosses another waveguide of the same wavelength. The unique arrangements of light sources and waveguides allows the use of a smaller number of more intense laser sources, particularly in applications such as performance-optimized datacenters where liquid cooling systems may be leveraged.

Abstract: Server resources in a data center are disaggregated into shared server resource pools, which include a pool of secure processors. Advantageously, servers are constructed dynamically, on-demand and based on a tenant's workload requirements, by allocating from these resource pools. According to this disclosure, secure processor modules for new servers are allocated to provide security for data-in-use (and data-at-rest) in a dynamic fashion so that virtual and non-virtual capacity can be adjusted in the disaggregate compute system without any downtime, e.g., based on workload security requirements and data sensitivity characteristics. The approach herein optimizes an overall utilization of an available secure processors resource pool in the disaggregated environment. The resulting disaggregate compute system that is configured according to the approach cryptographically-protects workload data whenever it is outside the CPU chip.

Abstract: Photonic circuits are disclosed having an efficient optical power distribution network. Laser chips (InP) having different wavelengths are flip-chip assembled near the center of a silicon photonic chip. Each InP die has multiple optical lanes, but a given die has only one wavelength. Waveguides formed in the photonic chip are optically connected to the lanes, and fan out to form multiple waveguide sets, where each waveguide set has one of the waveguides from each of the different wavelengths, i.e., one waveguide from each InP die. The waveguide network is optimized to minimize the number of crossings that any given waveguide may have, and no waveguide having a particular wavelength crosses another waveguide of the same wavelength. The unique arrangements of light sources and waveguides allows the use of a smaller number of more intense laser sources, particularly in applications such as performance-optimized datacenters where liquid cooling systems may be leveraged.

Abstract: Embodiments for optimizing dynamic resource allocations in a disaggregated computing environment. A data heat map associated with a data access pattern of data elements associated with a workload is maintained. The workload is classified into one of a plurality of classes, each class characterized by the data access pattern associated with the workload. The workload is then assigned to a dynamically constructed disaggregated system optimized with resources according to the one of the plurality of classes the workload is classified into to increase efficiency during a performance of the workload.

Abstract: Embodiments for disaster recovery in a disaggregated computing system. A memory pool is allocated including allocated memory elements at a secondary, disaster recovery site for data received from memory pool elements within the memory pool at a primary site. Data is continuously replicated to the allocated memory elements at the disaster recovery site without requiring any compute resources to be attached to the allocated memory elements during the replicating. An orchestration mechanism is used to regulate an available amount of resources to be assigned to the allocated memory elements at the disaster recovery site during a failover operation for performing failover workloads associated with the replicated data upon the primary site becoming inoperable.

Abstract: For measuring component utilization in a computing system, a server energy utilization reading of a statistical significant number of servers out of a total number of servers located in the datacenter is obtained by measuring, at predetermined intervals, a collective energy consumed by all processing components within each server. The collective energy is measured by virtually probing thereby monitoring an energy consumption of individual ones of all the processing components to each collect an individual energy utilization reading, where the individual energy utilization reading is aggregated over a predetermined time period to collect an energy consumption pattern associated with the server utilization reading.

Abstract: For power management in a disaggregated computing system, initial electrical power levels are distributed thereby allocating a voltage and a clock speed to each one of a set of processor cores in the disaggregated computing system. The voltage and the clock speed of respective processor cores within the set of processor cores are adjusted according to a workload priority of respective workloads performed by each respective one of the processor cores, wherein the workload priority is assigned based upon a service level agreement (SLA).

Abstract: Embodiments for preemptive deep diagnostics of resources in a disaggregated computing environment. A set of new resources of a first type is provided to an available resource pool within the disaggregated computing environment. An estimate for an expected time to failure (ETTF) for each one of the set of new resources is computed, and respective ones of the new resources from the available resource pool are provisioned to respective workloads based on the ETTF.

Abstract: A new workload is assigned to a subset of a plurality of processors, the subset of processors assigned a subset of a plurality of cache devices. A determination is made that the new workload is categorized as a cache-dependent workload which would be executed more efficiently were additional data elements associated with the new workload to be held in the subset of cache devices, and pursuant to determining the new workload is the cache-dependent workload, a determination is made as to whether the subset of cache devices is meeting the memory need of the new workload. Responsive to determining the subset of cache devices is not meeting the memory need of the new workload, a cache related action is performed.

Abstract: Embodiments are provided herein for facilitating high link bandwidth utilization in a disaggregated computing system. A plurality of general purpose links are used to connect respective pluralities of computing elements. A traffic pattern between respective ones of a first plurality of computing elements of a first type and respective ones of a second plurality of computing elements of a second type is detected. The first and second pluralities of computing elements are dynamically connected through the respective ones of the plurality of general purpose links according to the detected traffic pattern.

Abstract: Embodiments for efficient resource placement in a disaggregated computing environment. Responsive to receiving an alert of a threshold breach of one or more events associated with a suspicious resource of a first resource type while executing a workload, a known good resource is selected from an available resource pool of the first resource type, where the selecting is performed to optimize a usage of the first resource type. The known good resource from the available resource pool of the first resource type is assigned to the workload such that execution of the workload is transferred from being performed by the suspicious resource to the known good resource.

Abstract: For measuring component utilization in a system having a plurality of subsystems, an energy consumption of each of the plurality of subsystems is monitored whether or not each subsystem performs at least a portion of an overall computation. Respective workloads are classified based upon an energy consumption pattern associated with the monitored energy consumption of each of the plurality of subsystems.

Abstract: Embodiments are provided herein for dynamic memory-based communication in a disaggregated computing system. A pool of similar computing elements is configured as a large address space, the large address space segmented by an identifier. Data travel distances are optimized depending on a historical or expected use of a data object by using a grouping and amortization algorithm to relocate the data object within the pool of similar computing elements at a particular address within the large address space according to the historical or expected use.

Abstract: Embodiments for preemptive deep diagnostics of resources in a disaggregated computing environment. Responsive to detecting a threshold breach of a recurrent event associated with a first resource of a first resource type executing a workload, an alert is generated; and responsive to receiving the alert, the execution of the workload on the first resource is ceased. Health check diagnostics are identified and invoked on the first resource based on the alert and a server telemetry. Results of the health check diagnostics are mapped to a set of learned failure patterns; and a potential failure of the first resource is predicted based on the mapping.