-
-
Notifications
You must be signed in to change notification settings - Fork 10
/
fundraiser.html
144 lines (144 loc) · 6.52 KB
/
fundraiser.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
<!DOCTYPE html>
<html xmlns="http://www.w3.org/1999/xhtml" lang="" xml:lang="">
<head>
<meta charset="utf-8" />
<meta name="generator" content="pandoc" />
<meta
name="viewport"
content="width=device-width, initial-scale=1.0, user-scalable=yes"
/>
<title>SRCF Executive meeting - 14 Nov 2019</title>
<link
rel="stylesheet"
href="/_srcf/vendor/css/bootstrap.min.css"
/>
<link rel="stylesheet" href="/_srcf/srcf-bs.css" />
</head>
<body>
<main class="container">
<h1 id="srcf-new-server-fundraiser">SRCF new server fundraiser</h1>
<p>
This document outlines the details of the EPYC SRCF fundraiser. Please
consider supporting our fundraiser to support the SRCF’s mission in
providing free and flexible computing services to the entire University
community. In these trying times, the SRCF can make a difference but
only with your help.
</p>
<h2 id="context">Context</h2>
<p>
With the move to remote work and the SRCF’s offering of free video
conferencing, there is increasing demand for computationally-intensive
jobs which at present the SRCF is at risk of failing to meet with its
current hardware. It could upgrade its current machines up to a point
with diminishing returns but…
</p>
<p>
The SRCF’s machines are growing increasingly old. This means that
hardware needs constant “patching up” to account for their age and this
results in growing maintenance costs. Continuous patching will work in
the short term but it becomes unsustainable in the long term so instead
we have taken this as an impetus to move forward. What we have opted for
instead is to refresh our <code>thunder</code> cluster and bring it up
to date with the most modern technologies in the server world.
</p>
<p>
The last time the SRCF bought a server was in 2006 and at the time it
was meant to replace the single tower server it had called “kern”. Some
would have called that a watershed point in SRCF history. 14 years
later, we want to take another leap forward.
</p>
<h2 id="aims">Aims</h2>
<p>
The aim of the fundraiser is to purchase
<strong>4x CyberServe AMD EPYC EP1-102</strong>, with a calcualted cost
of £25,000. The machines far outperform anything in our current cluster.
We expect them to serve for 5 years (the warranty period) and given
Zen’s compatibility strengths, we could upgrade them without issues when
the time comes.
</p>
<p>
In the short term, they will be deployed to serve an increasing demand
for remote work and video conferencing, namely with the Cambridge
Students’ Union Virtual Freshers’ Fair, resumption of teaching, and
society events.
<i><a id="timeout-disclaimer"></a>
Editor's note: we wound down our first attempt at a video conferencing
platform, called Timeout, in May 2021. A future video conferencing
platform deployment project has not been ruled out.</i>
</p>
<p>
In the longer term, they will replace our <code>thunder</code> cluster,
which currently consists of 7 servers with an averge age of 8 years.
There, they will perform many of the existing tasks the SRCF has
reliably delivered over the last 21 years and additionally allow us to
expand and acommodate an ever-increasing list of new services.
</p>
<p>
A purchase of this size is extraordinary and unprecedented but it will
bring numerous benefits for the SRCF and its long term projects:
</p>
<ul>
<li>lower long term maintenance costs</li>
<li>
ability to meet increased demand for video conferencing in Michaelmas
stemming from the CamSU Virtual Freshers’ Fair and remote teaching
</li>
<li>
better carbon footprint (both more energy efficient machines and in
the reduction of the cluster size from 9 to 4 machines)
</li>
<li>
better performance – these 32 core servers easily outperform any of
our current ones which have at best 8 cores
</li>
<li>
better space efficiency – one of these machines easily accounts for 4
machines in our cluster and space at our Mill Lane rack is full
</li>
<li>
improved web hosting and scalability – there are plans to offer
multiple versions of our current single web server
<code>sinkhole</code> to better host custom web apps and distribute
user load across multiple machines
</li>
</ul>
<p>
Each server will come fully prepared with the necessary components,
including:
</p>
<ul>
<li>an AMD EPYC 7452 CPU with 32 Cores, 64 Threads, 2.35GHz/3.35GHz</li>
<li>4x 64GB 2933MHz DDR4 LR ECC DIMM modules</li>
<li>an extremely compact form factor - 1U only</li>
<li>a quality motherboard with a BMC</li>
</ul>
<p>
To contribute towards the fundraiser, please email
<code>treasurer@srcf.net</code>. If you are a company, we will gladly
offer you sponsorship benefits including a space on our website to
introduce yourselves, inclusion of your logo on our website, access to a
mailing list for advertising, and more benefits. In addition, we are
negotiating to have your logo on the Virtual Freshers’ Fair platform in
addition to Timeout, and would lead to massive exposure for your
company.
<i>[Editor's note: see <a href="#timeout-disclaimer">note above</a>
regarding the closure of Timeout since this document was written.]</i>
</p>
<h2 id="terms">Terms</h2>
<ul>
<li>
All donors to the fund will be mentioned on this page after the
fundraiser is over. A plaque will also be made and kept safely in the
SRCF server room for celebration purposes.
</li>
<li>
If the SRCF does not meet its fundraiser goal, it will extend the
fundraising period past October but purchase the servers it can afford
with the money obtained. These may or may not be of a lower
specification, depending on the amount raised as a reliable cluster
needs 4 machines to function.
</li>
</ul>
</main>
</body>
</html>