Skip Navigation

Friday August 2nd, 2013

Cluster Support Services

Summary: Description of the Cluster Support Services that OIT provides to UCI departments and researchers.

Introduction

OIT is offering the support services described below for departments and researchers using clusters of compute notes to serve their computational needs: Shared Cluster Support, Dedicated Rocks Cluster Support, and Custom Cluster Support.

With each option, OIT provides initial consultation services, grant writing assistance, and technical expertise to guide the owner through the process of determining their needs, submitting an RFP if necessary, and selecting a vendor and product.

The characteristics for each cluster support option are summarized below. More detailed information is available in a separate "service level agreement" document.

Shared Cluster Support (SCS)

Shared Cluster Support (SCS) follows in the tradition of the OIT MPC cluster by allowing PIs to acquire compute nodes that are then installed in a large, shared cluster and dedicated for their use. This is the most efficient support option as it leverages system administration demands by utilizing a pre-established operating environment, thus spreading support costs among all users.

SCS represents a comprehensive package of support services. Each owner signing up for SCS becomes an equal partner in a shared cluster computing community. OIT provisions and supports all aspects of the computing infrastructure including machine room space, network and UC Grid connectivity, compiler and software licenses, home directory and scratch disk space, programming support, and hardware maintenance. The owner purchases computing and storage nodes; the purchase must include a minimum of 3 years of hardware support and the equipment must be approved by OIT. The nodes are housed within the OIT Academic Data Center, which provides power, enhanced security, climate control, and after-hours monitoring and support. OIT takes delivery of, racks, and configures the owner's equipment to the owner's specification within the limits of the SCS framework. OIT creates user accounts and job queues at the direction of the owner. OIT interacts with equipment venders on the owner's behalf to replace faulty components and trains and assists the owner's programmers with use of the SCS. OIT installs both commercial software purchased by the owner and free open-source software requested by the owner as long as the software does not interfere with the normal operation of SCS and within the confines of available staff time. The owner has the option of purchasing additional storage space to augment the OIT-provided storage space.

Dedicated Rocks Cluster Support (DRCS)

Dedicated Rocks Cluster Support (DRCS) is another support option those requiring cluster computing based on the standard ROCKS environment. In many ways, this service provides an "a la carte" selection of services. Similar to SCS, OIT uses a standardized operating environment to minimize support costs. However, with Dedicated Cluster Support, the PI provides all hardware required to form a separate cluster dedicated to his or her research.

OIT provides system administration and configuration of the Red Hat Enterprise Linux OS and ROCKS clustering software on a recharge basis; system administration includes creating user accounts, configuring the TORQUE or Sun Grid Engine job scheduler, installing pre-packaged open-source and commercial software, and maintaining and patching the base OS and applications. The initial cluster hardware setup and configuration, OIT machine room space, compiling of software, diagnosing and repairing hardware issues, programming support and data backups services are performed by OIT staff at established OIT recharge rates. OIT will advise the owner in advance if charges would be incurred for services to be performed and will provide a detailed contract of the support options selected by the owner.

Custom Cluster Support (CCS)

For users whose computing needs are not addressed with a ROCKS computing cluster, OIT also offers a Custom Cluster Support (CCS) option. Since there are a wide-range of possibilities in terms of computing hardware and clustering software, the CCS option must itself be flexible and adaptive. OIT would allocate a percentage of its system administration staff time to build, configure, and maintain the cluster. The amount of time allocated and specific duties would be negotiated on a case-by-case basis with the cluster's owner and could be adjusted over time as the cluster's support needs change. The costs of housing the equipment in the OIT Academic Data Center are determined by the overall equipment footprint, power consumption, and cooling requirements.

Comparison of Cluster Support Options

Shared Cluster Dedicated Rocks Cluster Custom Cluster
OS & Clustering RHEL, ROCKS RHEL, ROCKS Any [1]
Hosting in OIT Academic Data Center Floor space:
Included [2]
Floor space (monthly):
$11.25/U [3] or $200/rack [4]
Determined by overall floor space, power, and cooling usage.
System Administration $300 annually each node $2,520 annually for head node + $300 annually each device [5] Determined by system usage and maintenance requirements.
Hardware Installation and Configuration $65/hour $65/hour $65/hour
Programming Support Included $65/hour Included

In addition, membership in the Shared Cluster includes access to commercial software, such as Portland Group compilers for FORTRAN, C, and C++ and MathWork's MATLAB Distributed Computing Engine, at no additional charge.

[1] The operating system is subject to approval by OIT; it must be actively supported by its vendor and/or developer.

[2] OIT will provide up to 40 U of rack space.

[3] Assumes that OIT provides the rack enclosure.

[4] Assumes that the equipment owner provides the rack enclosure.

[5] A device is defined as a computing node, storage node, or Ethernet network switch.

Sample Pricing

15 Compute Nodes:

If purchasing 15 computing nodes at $4,000 each (1U each):

$780  —  Initial equipment setup fee, estimated 12 hours at $65/hour
$4,500  —  Annual SCS contract
$0  —  Annual Academic Data Center housing
$14,280  —  3 years of SCS contract plus setup
$74,280  —  Total cost over 3 years, hardware and SCS

As a DRCS contract, you will need additional equipment for a head node and network switch (estimated $5,500 additional hardware costs and 5U of rack space):

$780  —  Initial equipment setup fee, estimated 12 hours at $65/hour
$7,020  —  Annual DRCS contract
$2,700  —  Annual Academic Data Center housing
$29,940  —  3 years of DRCS contract, ADC housing, and setup
$97,440  —  Total cost over 3 years, hardware, contract, housing, and setup

30 Compute Nodes:

If purchasing 30 computing nodes on blades at $5,000 each (10 blades per chassis [$6,000, 7U each]):

$1,560  —  Initial equipment setup fee, estimated 24 hours at $65/hour
$9,000  —  Annual SCS contract
$0  —  Annual Academic Data Center housing
$28,560  —  3 years of SCS contract plus setup
$196,560  —  Total cost over 3 years, hardware and SCS

As a DRCS contract, you will need additional equipment for a head node and two network switches (estimated $6,750 additional hardware costs and 6U of rack space) with a rack enclosure ($2,000 for 42U):

$1,560  —  Initial equipment setup fee, estimated 24 hours at $65/hour
$12,120  —  Annual DRCS contract
$2,400  —  Annual Academic Data Center housing
$45,120  —  3 years of DRCS contract, ADC housing, setup
$221,870  —  Total cost over 3 years, hardware, contract, housing, and setup