<div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Sep 30, 2019 at 1:45 PM Kayode Enwerem <<a href="mailto:Kayode_Enwerem@ao.uscourts.gov">Kayode_Enwerem@ao.uscourts.gov</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<div lang="EN-US">
<div class="gmail-m_-7341258934793543404WordSection1">
<p class="MsoNormal"><br></p>
<p class="MsoNormal">I have 32 CPUs on this server and CPU model name is - AMD Opteron(tm) Processor 6386 SE</p></div></div></blockquote><div><br></div><div></div></div><div>People have been having issues with older opterons like that for a long time. They have a lot of cores, but the single core performance is about half that of a more recent CPU.</div><div><br></div><div>With 32 real cores (assuming this is a dual socket system) I'd try running closer to 28 workers which gives you 20% more capacity over 23.</div><div><br></div><div>After that, you need to look at the conn.log to determine where your capture loss is coming from by looking at the missed_bytes column. You may have some elephant flows that are accounting for the majority of that loss.</div><div><br></div>-- <br><div dir="ltr" class="gmail_signature"><div dir="ltr">Justin</div></div></div>