Einstein 2017-09-07T10:48:21+00:00

Einstein College of Medicine

Einstein College of Medicine is a world renowned research facility that works to advance human health, innovative research and discovery. The college specializes in executing a full range of lifescience research including full genome sequencing. DST manages the a large file storage repository ensuring performance, efficiency and availability to the researchers. DST helps Einstein do more science.

DST provides remote system management for the scientists who do their research at Einstein College of Medicine as well as On Demand Services for the technologists who work to keep the environments applications always available and always current for the research community. We provide a wide range of skill sets to the University via On Demand Services which helps address whatever the scientists throw at them. Our On Demand Services have helped with projects such as: help my dissertation is on an old ZFS array that was shut down and no one knows the password anymore, help we need a Nagios instance to monitor infrastructure, help we need an NGINX stood up right away for professors, help we need an instance of MATlabs for a new PI to name a few.

The design offers support for any native Linux virtual infrastructure such as KVM, Xen or OpenStack. DST is recommending the use of OpenStack and the automated OpenStack deployment software that makes implementing and on-going care and feeding of an OpenStack environment less cumbersome. Bright Cluster Manager for OpenStack provides a wizard-based GUI for deploying a complete virtual environment and provides resources to effectively manage processing, memory allocation and load balancing across the cluster. This is essential in the proposed environment due to the fact the VMs are running on the Quobyte Controller nodes.

Bright OpenStack is ideal for users seeking a fast and easy path to OpenStack.  Bright deploys and provisions not only the OpenStack instance, but also all of the underlying software ecosystem on which OpenStack relies.  This includes associated databases, message queueing services (AMQP), load balancers (HAProxy), various clustered Linux servers, and the Linux operating system itself.

Bright’s installation wizard enables users to choose between an “express” and a “standard” deployment.  In express mode, users will be prompted to answer only a few questions during the install process, with the possibility to alter the values automatically determined for them by the wizard.  In standard mode, which allows for more customization (i.e. in terms of the storage and networking back end), the install process is still very streamlined, due to Bright’s ability to choose the best defaults and to detect pre-existing installations.

Bright OpenStack’s Cluster Management GUI (CMGUI) is optimized for performing daily OpenStack Management operations.  Several capabilities distinguish Bright’s GUI from open source and other versions.

Single pane of glass.  Bright OpenStack’s GUI is a single pane of glass solution. It gives the cloud admin full management and monitoring access to OpenStack resources as well as all of the components around OpenStack. This includes physical hardware, switches, PDUs, operating system services, hypervisors, virtualized hardware, and virtualized operating systems, various auxiliary software (HAProxy, RabbitMQ, etc.), etc. All from within a single window.

Lifecycle management.  With CMGUI, you can easily modify the underlying OpenStack deployment as your requirements change. This can be done using the concept of roles.

Need more nova-api nodes? Simply assign the “OpenStack Compute API” role to a node or a group of nodes, maybe change some defaults if needed, and you’re done.  The Bright cluster management daemon will take care of writing out the nova.conf configuration file, starting (and monitoring) the openstack-nova-compute service, and registering the new endpoint with HAProxy.

Native OpenStack management.  Managing OpenStack objects (networks, subnets, routers, VMs, users etc), is straightforward with CMGUI.  Admins always have direct access to all of the OpenStack objects associated in one way or the other with the currently selected object. This makes for very intuitive navigation.

Global search.  OpenStack users frequently cite the need to conduct searches across all OpenStack objects, e.g. “all OpenStack objects (regardless of type) that have been disabled,” or, “VMs that were created 2 weeks ago between 1pm and 4pm.”  This type of search is a snap with CMGUI’s built-in search function.

Managing complexity with filtering.  In production, cloud admins often have to deal with thousands of objects.  With CMGUI, users can easily filter out objects of interest by applying filters to their individual fields. It’s easy to show, for example, all volumes of a specific size, owned by specific tenant, and created on specific date.

Monitor the entire stack.  Bright’s unique CMGUI monitoring view gives admins full monitoring access to the entire stack. Starting with monitoring physical hardware, physical switches, operating system, OS services, the hypervisor service, all the way through virtualized hardware, virtualized operating system and services, up to virtualized “userspace” programs running inside of those VMs. All from within the same window.

The Brocade ICX 7700 Series switches are among the most performant and easiest to deploy and manage for high throughput requirements.

The Brocade ICX 7750 Switch delivers industry-leading 10/40 GbE port density, advanced high-availability capabilities, and flexible stacking architecture, making it the most robust Brocade aggregation and core distributed chassis switch offering for enterprise LANs. In addition to rich Layer 3 features, the Brocade ICX 7750 scales to 12-unit distributed-chassis stacking or Multi-Chassis Trunking (MCT) and is an integral part of Brocade Campus Fabric technology.

Part of the Brocade ICX family of Ethernet switches for campus LAN and classic Ethernet data center environments, the Brocade ICX 7750 Switch is a 1U high-performance, high-availability, and market-leading-density 10/40 GbE solution that meets the needs of business-sensitive campus deployments and classic Ethernet data center environments. With industry-leading price/performance and a low-latency, cut-through, non-blocking architecture, the Brocade ICX 7750 provides a cost-effective, robust solution for the most demanding deployments.

DST proposed two different flavors of 7750 in a single stack; the 7750-26Q and the 7750-48C. The 7750-26Q provides twenty-six (26) host-facing 40GbE ports with six (6) 40GbE stacking ports. This will provide Optimal Geo with eighteen (18) free ports allowing nine (9) additional storage nodes to be deployed for future growth. By connecting the four (4) of the stacking ports to the ICX 7750-48C, Brocade can provide 160Gb of aggregate throughput, sufficient to push well over the 4.5GB/s of throughput that the storage nodes can handle. With nine (9) additional Quobyte nodes, Optimal Geo will be able to push about 20GB/s of throughput to and from the storage environment. The ICX 7750-48C is a 48-port 1/10Gb switch with an additional six (6) 40Gb ports. These six (6) 40Gb ports can be broken out into four (4) 10Gb ports with a special cable for a total of twenty-four (24) additional 10Gb ports.

The following illustrates the Brocade switches as configured for Optimal Geo.

The LB-DH8 Series Data Archiver is a scalable optical disc library system that supports ever-increasing demand for longer-term data storage in data center business. Panasonic has leveraged its optical disc drive, media, and advanced robotics technologies to develop a scalable module expansion configuration that can support an increase of data with high reliability that is necessary for long-term data storage.

The system consists of data archiver magazines which hold twelve optical discs. Each magazine can store 1.2 TB of data. The system has a base module (magazine writer unit) and expansion module (magazine loader) that can house up to 76 data archiver magazines respectively. A maximum of 91.2 TB can be stored in each module. The system consists of three types of modules: base module, bottom module (magazine carrier), and expansion module combined with magazine management software. By installing additional expansion modules on the 19-inch rack, a flexible system configuration can be created according to the data volume. Up to seven modules can be installed to realize high-volume optical disc storage of 638.4 TB max. per rack.

The Data Archiver uses RAID technology. Twelve drive units concurrently operate twelve optical discs housed in a data archiver magazine to provide high-speed data transfer up to 216 MB/s and high reliability to protect data from unforeseen failures. Optical discs that have a data storage life of 50 years*1 are adopted to eliminate regular data migration. Standby power consumption is only 7 W*2, significantly lowering power consumption and reducing running cost of data centers.

DST delivers a full managed service around this environment to include customized coverage times.

Category Description Response SLA Escalation SLA
General System Problem – Hardware This would encompass any hardware related issue that does not significantly impact the user. 120 minutes 120 minutes
General System Problem – OS This would encompass any software or operating system related issue that does not significantly impact the user. 120 minutes 120 minutes
System Outage System outage is defined as a single system not accessible. This may put the client in a degraded or inoperable state 60 minutes 60 minutes
General Storage Problem – Hardware This would encompass any storage hardware related issue that does not significantly impact the user. 120 minutes 120 minutes
General System Problem – Software This would encompass any storage system software related issue that does not significantly impact the user. 120 minutes 120 minutes
Storage Outage Cluster outage is defined as the inability of the cluster to process data 60 minutes 60 Minutes
Workload/Job Issue All workload/job related issues including jobs that saturate storage IO or adversely impact processing and memory on the host 60 minutes 120 minutes
Networking Issue DST will provide escalation of networking and data transfer issues to the appropriate networking team. In an effort to provide accelerated service, DST has network savvy resources to assist. 90 minutes 120 minutes